read article and answer question
ksssss
EconomicconsequencesoftheSarbanesOxleyActof2002Zhang.pdf
ARTICLE IN PRESS
Journal of Accounting and Economics 44 (2007) 74–115
0165-4101/$ -
doi:10.1016/j
� Tel.: +1 6
E-mail ad $ This pap
committee, B
invaluable gu
Philip Joos, S
Liu, John Lo
referee, partic
HKUST, Mi
errors are my
www.elsevier.com/locate/jae
Economic consequences of the Sarbanes–Oxley Act of 2002
$
Ivy Xiying Zhang �
Carlson School of Management, University of Minnesota, Minneapolis, MN 55455, USA
Available online 3 March 2007
Abstract
This paper investigates the economic consequences of the Sarbanes–Oxley Act (SOX) by examining
market reactions to related legislative events. Using concurrent stock returns of non-U.S.-traded
foreign firms to estimate normal U.S. returns, I find that U.S. firms experienced a statistically
significant negative cumulative abnormal return around key SOX events. I then examine the cross-
sectional variation of U.S. firms’ returns around these events. Regression results are consistent with
the non-audit services and governance provisions imposing net costs. Additional tests show that
deferring the compliance of Section 404, which mandates an internal control test, resulted in
significant cost savings for non-accelerated filers.
r 2007 Elsevier B.V. All rights reserved.
JEL classification: G38; M41
Keywords: Securities legislation; Sarbanes–Oxley Act; Corporate governance; Internal control; Non-audit services
see front matter r 2007 Elsevier B.V. All rights reserved.
.jacceco.2007.02.002
12 626 3118.
dress: [email protected].
er is based on my dissertation completed at the University of Rochester. I am grateful to my thesis
ill Schwert, Charles Wasley, Ross Watts, and especially Jerry Zimmerman (Chairman) for his
idance. I also appreciate comments from Rui Albuquerque, Jim Brickley, Yan Cao, Paul Griffin,
.P. Kothari (the editor), Tao Kuang, Ling Lei, Andy Leone, Christian Leuz (the discussant), Laura
ng, Pervin Shroff, Clara Vega, Ram Venkataraman, Jerry Warner, Joanna Wu, an anonymous
ipants at the 2005 Journal of Accounting and Economics Conference, and workshop participants at
nnesota, MIT, NYU, Penn State, Purdue, Rochester, and the University of Texas at Dallas. All
own.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 75
1. Introduction
The extant literature provides mixed evidence regarding the consequences of securities legislation. Early studies (e.g., Stigler, 1964; Benston, 1973) find no support for the benefits of the Securities Acts of 1933 and 1934, but the methodology and the interpretation of these studies have been challenged subsequently (Schwert, 1981; Friend and Westerfield, 1975). More recently, Bushee and Leuz (2005) document positive externalities but also significant costs of extending disclosure requirements to OTCBB firms in 1999. In contrast, Greenstone et al. (2006) found that firms required by the 1964 Securities Act Amendments to increase disclosures experienced a positive return around the announcement of the law.
While the impact of existing securities legislation is still under debate, Congress passed the Sarbanes–Oxley Act (the Act or SOX hereafter) in July 2002, in response to a number of high-profile scandals starting in late 2001. The Act has been widely considered the most far-reaching securities legislation since the Securities Acts of 1933 and 1934.
1 It not only
imposes additional disclosure requirements, but more importantly, proposes substantive corporate governance mandates, a practice that is unprecedented in the history of federal securities legislation (Romano, 2004).
By requiring more oversight, imposing greater penalties for managerial misconduct, and dealing with potential conflicts of interest, the Act aims to prevent deceptive accounting and management misbehavior. However, despite the claimed benefits of this Act, the business community has expressed substantial concerns about its costs. Whereas the out- of-pocket compliance costs are generally considered significant (Solomon and Bryan-Low, 2004), they are likely swamped by the opportunity costs SOX imposed on business. Executives complain that complying with the rules diverts their attention from doing business (Solomon and Bryan-Low, 2004). Furthermore, the Act exposes managers and directors to greater litigation risks and stiffer penalties. CEOs allegedly will take less risky actions, consequently changing their business strategies and potentially reducing firm value (Ribstein, 2002).
The overall direct and indirect private costs of SOX on businesses could well outweigh its private benefits. The passage of SOX gives rise to a broader concern that SOX could signal a shift to more rigid federal and state regulation of corporations, thereby causing extensive changes in the economy (Holmstrom and Kaplan, 2003). A 2004 Pricewater- houseCooper’s survey of CEOs finds that 59% of the respondents view the risk of overregulation as one of the biggest threats to the growth of firms (Norris, 2004).
Motivated by the ongoing debate on the economic impact of securities regulation, and that of SOX particularly, this paper investigates stock price reactions to the SOX legislative events. A maintained hypothesis is that stock prices unbiasedly incorporate all the expected private costs and benefits of SOX based on available information. As non- U.S.-traded foreign firms are exposed to common global economic news as U.S. firms but are not directly affected by SOX, I examine abnormal returns of the U.S. market around significant SOX legislative events relative to returns of these foreign firms. The cumulative value-weighted (equal-weighted) raw return of the U.S. market amounts to �15:35% (�12:53%) around the key SOX events. The differences in value-weighted (equal- weighted) raw returns between U.S. and foreign markets range from �1:80% to �7:72%
1 See President Bush’s speech at the signing ceremony of SOX (http://www.whitehouse.gov/news/releases/2002/
07).
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11576
(�5:51% to �9:32%) and are not always statistically significant. However, after taking into account different markets exhibiting different response to common economic news in computing U.S. expected returns, the estimated U.S. cumulative abnormal returns around key SOX events range from �3:76% to �8:21% under alternative specifications and are all statistically significant. Further, consistent with Litvak (2005), I find that SOX-complying foreign firms on average experience a negative cumulative abnormal return around key SOX rulemaking events. The results suggest that SOX imposes statistically significant net costs on complying firms. The losses likely reflect the private costs of SOX and/or the expected costs of future federal and state regulation. The cumulative abnormal returns of the U.S. market around all SOX events are not statistically significant. While this finding does not support SOX being costly, it is likely due to the inclusion of insignificant events reducing the power of the test. Next, I explore the sources of private costs of SOX by investigating its major provisions
and their cross-sectional implications. Specifically, I examine the implications of the restriction of non-audit services, the provisions on corporate responsibilities, the provisions on the forfeiture of incentive pay and insider trading, and Section 404 on internal control tests. Greenstone et al. (2006) conjectured that firms in less compliance with new regulatory requirements likely have greater unresolved agency problems and thus benefit more from the regulations. On the other hand, these targeted firms also incur greater costs as a result of rule changes. If the examined SOX provisions impose net private costs on firms, I expect firms that purchased more non-audit services, maintained weaker governance in terms of shareholder rights than optimal, and relied excessively on incentive pay, and those with more complex business experience more negative abnormal returns. The empirical results largely reject the hypothesis that major SOX provisions entail no
net costs. Firms’ cumulative abnormal returns around the significant legislative events decrease with their purchase of non-audit services. Firms with foreign operations or international transactions incur significantly greater costs. Moreover, firms with so-called ‘‘weak’’ shareholder rights relative to their industry peers experienced more negative abnormal returns in the SOX rulemaking period. There is some evidence that firms with more business lines or more incentive pay incur greater compliance costs, but the statistical significance of the result is sensitive to specifications. Overall, the cross-sectional tests are consistent with my event study findings that the market expected SOX to impose statistically significant net private costs on firms. Finally, I examine market reactions to the SECs announcement of deferring compliance
with Section 404. I find that small firms that obtained a longer extension period experienced significantly higher abnormal returns than other firms around the announce- ment. The evidence reveals that investors consider the deferment to be good news and that the compliance costs of Section 404 are statistically significant for small firms. This study contributes to the research of securities legislation by presenting evidence on
the impact of contemporary federal laws that introduce unprecedented substantive changes in firms’ operations (Romano, 2004; Bainbridge, 2003). The market reaction analysis provides evidence on the new rules, as a package, imposing net private costs on businesses and investors considering the message conveyed by the legislative activities to be negative. Further, the cross-sectional analysis attempts to explore the consequences of specific
SOX provisions and documents associations between firm characteristics and returns that are consistent with certain provisions being costly. Yet, the analysis cannot fully unravel the consequences of specific SOX provisions. SOX provisions are correlated, so are the
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 77
examined firm characteristics. One cannot definitively attribute the costs of SOX to a subset of provisions.
This paper also extends the event-study literature by examining changes in stock prices in response to market-wide news. Investigating the impact of market-wide news is more challenging than typical event studies that investigate market reactions to news releases affecting a subset of listed firms. This paper uses foreign returns as a benchmark to evaluate normal market-level returns in the U.S. and examines the impact of other contemporaneous news announcements in the SOX rulemaking period through intraday return tests.
2
This study does not explore the social welfare implication of the Act. An investigation of changes in security prices provides evidence for the private benefits and costs of regulations, but social benefits and costs may not be fully reflected in stock prices (Gonedes and Dopuch, 1974; Watts and Zimmerman, 1986). In addition, this paper is subject to the fundamental limitations of event studies (Leftwich, 1981; Schwert, 1981). First, other contemporaneous news is incorporated in stock prices and cannot be disentangled definitively from the impact of SOX. The finding that foreign firms not subject to SOX also experienced negative returns in the SOX rulemaking period, albeit smaller in magnitude, suggests that other contemporaneous worldwide economic news negatively affected U.S. firms. Second, as investors’ expectations are unobservable, I cannot reject the alternative hypothesis that investors had expected stronger rules and were disappointed by SOX. Third, it is possible that the market over-estimated the net costs of SOX when it was passed and subsequently SOX turned out to be less costly than expected. Future research on these issues would provide additional insight for the evaluation of the Act.
The remainder of the paper is organized as follows. Section 2 discusses the event history of SOX and related research. Section 3 examines the private costs and benefits of the Act and develops hypotheses regarding the cross-sectional variation in market response to the Act. Empirical tests of these hypotheses are then discussed in Section 4. Section 5 concludes.
2. Related research and SOX event history
2.1. Related research
The existing research on securities regulations is inconclusive on the consequences of mandatory disclosure laws. Early studies by Stigler (1964) and Benston (1973) find no evidence of changes in returns or the variability of returns after the Securities Acts of 1933 and 1934. They argue that mandatory disclosures imposed by the Securities Acts generated few benefits and considerable costs. However, the methodology and the interpretation of these studies have been challenged subsequently (Schwert, 1981; Friend and Westerfield, 1975). Chow (1983) improves on this line of research by examining market reactions to the rulemaking progress of the 1933 Securities Act. He finds stock price declines for complying firms in the rulemaking period and concludes that the Act was costly. Coffee (1984) calls for research on more contemporary securities regulations and points out that the benefits of mandatory disclosures may not be reflected in stock returns.
More recently, Bushee and Leuz (2005) document positive externalities as well as significant costs of increased disclosure by investigating the regulation to mandate OTCBB
2 I thank the editor and the referee for suggesting the use of foreign returns.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11578
firms to comply with the 1934 disclosure laws. They find that a large portion of firms delisted in response to the regulatory change and the delisting decision was associated with significant reductions in their market value. In contrast, Greenstone et al. (2006) examine the impact of the 1964 Securities Acts Amendments to extend the 1934 disclosure requirements to large firms traded over the counter. They find that the most affected firms experienced a positive abnormal return around the announcement of the law and that they outperformed less affected firms (i.e., those in partial compliance with the disclosure mandates) in the rulemaking period. They argue that mandatory disclosures can benefit shareholders by deterring diversions of profits by insiders. While the existing literature produces mixed evidence regarding whether mandatory
disclosures are beneficial, SOX imposes additional disclosure requirements, and more importantly, it further proposes substantive corporate governance mandates, which is unprecedented in the history of federal securities legislation. Romano (2004) argues that by enacting SOX, Congress substitutes its judgments for that of the board in real corporate activities and that SOX can be a first step in federalizing corporate laws. Thus, this study adds to the literature by investigating the consequences of a contemporary and potentially significant change in federal securities legislation. Many concurrent studies examine the consequences of the Sarbanes–Oxley Act; however,
there is no consensus on how SOX changes business practices, nor whether the changes are value increasing for firms. Regarding the SOX provision to require firms have a financial expert on the board, DeFond et al. (2005) document that firms voluntarily appointing a financial expert to the board prior to SOX experienced positive returns at the announcement of the appointment. Yet, they also find that the impact of such an appointment is contextual and point out it is unclear that requiring all firms to appoint a financial expert is beneficial on average. Bhattacharya et al. (2002) examine the market response to CEO and CFO certification required by SOX. They find no significant price reactions. Engel et al. (2006) examine firms’ going private decisions and suggest that some small firms deregister after SOX as a result of the compliance costs imposed by SOX. Investigating firms’ going dark decisions, Leuz et al. (2006) find that SOX drives some poor performing firms out of the stock market as insiders are exposed to greater legal liabilities. Several studies investigate the impact of SOX on U.S.-listed foreign firms that are
required to comply with this legislation. Berger et al. (2005) find that foreign private issuers experience more negative returns in the U.S. market than U.S. firms around certain SOX events. They argue that the incremental bonding benefits provided by SOX are exceeded by the costs. Litvak (2005) compares the abnormal returns of U.S.-listed foreign firms with their non-U.S.-traded counterparts around the SOX events. She finds that U.S.-listed foreign firms experienced negative abnormal returns around events that increased the likelihood of SOX passage. While these studies further the understanding of various aspects of SOX, the average
wealth implication of this Act on U.S. firms is not yet explored, which is the focus of this paper. There are two concurrent studies closely related to this paper. Both Rezaee and Jain (2002) and Li et al. (2004) investigate market reactions to SOX rulemaking activities and conclude that SOX is beneficial. Appendix A discusses the problems with these two studies. This paper reports conflicting evidence after taking into account the problems. First, I find no evidence supporting the conjecture that SOX is beneficial. Second, I examine major SOX provisions and find that the investigated provisions likely impose net private costs on firms.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 79
2.2. Event history
The Sarbanes–Oxley Act, which combined the accounting reform bills of Sen. Sarbanes and Rep. Oxley, was passed in Congress on July 25, 2002. The two bills, together with other legislative proposals towards corporate reforms, were triggered directly by the collapse of Enron in late 2001, which exposed an unprecedented accounting scandal and an allegedly seriously corrupted governance system. I identify the legislative events leading to the passage of SOX by a keyword search of ‘‘accounting’’ through the Wall Street Journal (WSJ hereafter) and the Washington Post (WP hereafter) via Factiva, from November 2001 to July 2002. To identify related post-SOX rulemaking events, I search the WSJ and WP for ‘‘Sarbanes–Oxley’’ from August 2002 to December 2003 and examine press releases of the SEC and the PCAOB during this period. The WSJ is widely considered the most influential and timely business journal and its news filtering system is likely to extract legislative activities that are most relevant to the business community. The WP closely follows significant movements in Congress and provides information supplementary to the WSJ articles.
There was no significant development in securities rulemaking in 2001 (Hilzenrath, 2001). The first signal of a regulatory overhaul was reported on January 16, 2002 (Day and Crenshaw, 2002) when SEC Chairman Pitt announced a reform plan to create an independent regulatory organization. Legislative activities progressed slowly from February to May 2002. The Bush Administration unveiled their response to the Enron scandal in February and March, while Congress moved ahead with several proposals towards accounting reforms. Republican Rep. Oxley’s reform bill, considered business- friendly by the media, was introduced in the House on February 13 (Schroeder, 2002a). Meanwhile, Democratic senators reportedly drafted bills that went beyond Oxley’s proposal (Schroeder, 2002b, c).
Although Sen. Sarbanes’ tough reform bill passed in the Senate Banking Committee on June 18, the media did not expect it to become law at that time (Hilzenrath et al., 2002). However, the exposure of the WorldCom scandal in late June boosted securities rulemaking activities (Hamburger et al., 2002). The rulemaking process accelerated after President Bush delivered a speech regarding accounting reforms on July 9 (Cummings, 2002). The Senate started debate on Sarbanes’ bill on July 8 and the bill was passed 97 to 0 in the Senate on July 15 (Hilzenrath et al., 2002).
House GOP leaders allegedly sought to dilute Sarbanes’ bill after its passage (VandeHei, 2002a). However, they soon retreated from such efforts and offered minor changes to complete the legislation (Murray, 2002a). The House and Senate formed a conference committee and started final negotiations to merge the bills on Friday, July 19 (Hilzenrath, 2002). The negotiation continued over the weekend of July 20 and in a radio address on Saturday, July 20, President Bush urged Congress to pass a final bill before the fall recess (Melloan, 2002). The final rule was agreed upon on July 24 (VandeHei and Hilzenrath, 2002), passed in Congress on July 25, and signed into law on July 30 (Hitt, 2002).
The implementation of SOX started soon after its passage and the rulemaking activities continued in 2003. The SEC adopted rules on management report of internal controls on May 27 (Solomon, 2003). The PCAOB audit standard of internal controls was approved by the SEC in June 2004, which completed the major rulemaking activities directed by SOX.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11580
3. Hypotheses
3.1. U.S. market reaction to SOX
The Sarbanes–Oxley Act established the Public Company Accounting Oversight Board (PCAOB), prohibited auditors from performing certain non-audit services for their audit clients, and imposed greater criminal penalties for corporate fraud. Further, Section 404 of the Act required that management assess internal controls and that auditors report on their clients’ internal controls. The specifics of SOX are discussed in greater detail in the next subsection. Lawmakers expect SOX to enhance corporate controls of public firms and prevent
accounting misrepresentations. Mahoney (1995) points out that mandatory disclosures are a means of controlling agency costs. Greenstone et al. (2006) further argue that as private contracts cannot prevent all expropriations of firm profits by insiders, regulations can increase the expected penalties for expropriation and thereby reduce misbehaviors and increase shareholder value. In other words, if the exposed accounting scandals that led to the rulemaking activities suggest a pervasive market failure, the regulations could increase firm value and improve efficiency. As Watts and Zimmerman (1986) indicate, regulations could improve social welfare in a Pareto sense, if government regulations and enforcement lower the private contracting costs. However, it is unclear that the government’s remedies are always less costly than the
private contracting process. SOX rushed through Congress in a very short period of time. Romano (2004) argues that electoral concerns were aided at the cost of a comprehensive consideration of the implications of the legislation. She explains that Democrats planned to charge pro-business Republicans for being soft on corporate scandals in the mid-term election of November 2002, and Republicans responded to show their determination to punish corporate wrongdoings. The business community and certain academics criticize SOX more specifically. While the
out-of-pocket compliance costs are generally considered substantial (Solomon and Bryan- Low, 2004), the indirect opportunity costs imposed by SOX are likely even greater. The Wall Street Journal cited the chief accounting officer of General Motors: ‘‘The real cost isn’t the incremental dollars, it is having people that should be focused on the business focused instead on complying with the details of the rules’’ (Solomon and Bryan-Low, 2004). Moreover, as SOX increases the litigation risks of management, Ribstein (2002) argues that it could discourage CEOs from value-increasing risky investment. Wallison (2003) expects the change in management’s risk-taking behavior to reduce the growth of firms and even deter economic growth. The change in the behavior of lawmakers, especially pro-business Republicans, towards tighter government controls is likely to give rise to rigid federal and state regulation in the future. The change is therefore expected to have long-lasting and far- reaching influence on business practices (Holmstrom and Kaplan, 2003). If investors consider the Act on net costly and/or the information conveyed by the
passage of the Act bad news for business, the cumulative U.S. return around the rulemaking events adjusted for the impact of contemporaneous economic news would be negative. H1: If SOX imposed net costs on U.S. firms, firms’ cumulative returns adjusted for the
impact of contemporaneous economic news around the SOX rulemaking events would be
negative.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 81
The market response to individual events was determined by the value implication of the Act and how the news changed investors’ expectations of the probability of passing tough rules (Leftwich, 1981). If the proposed regulation imposed net private costs on firms, events that increased the probability of passing tough rules would be associated with significant negative abnormal returns.
3.2. Specific provisions of SOX and their cross-sectional implications
This section examines various SOX provisions in four areas, namely, auditor independence, corporate responsibilities, internal controls, and incentive pay and insider trading.
3.2.1. Provisions on non-audit services
SOX prohibits accounting firms from providing eight categories of non-audit services contemporaneously with audit services, thereby leaving tax services as the primary non- audit service available to audit clients (Section 201).
3 The restriction of non-audit services
reflects lawmakers’ concern that the provision of non-audit services compromises auditor independence by increasing the economic bond between the auditor and the client. However, auditors adamantly opposed this position and argued that performing non-audit services helps them gain competencies and capabilities that are essential to the audit process (Schroeder and Hamburger, 2002).
Empirical evidence on the relation between non-audit services and audit quality is mixed. For example, Frankel et al. (2002) document a positive relation between non-audit fees and their measures of earnings management. In contrast, Ashbaugh et al. (2003) find no systematic evidence to support the claim that non-audit services impair auditor independence. They point out that the results of Frankel et al. (2002) are sensitive to research design choices. Romano (2004) reviews the research on non-audit services and concludes that the majority of studies find no connection between the provision of non- audit services by the auditor and audit quality.
The restriction on non-audit services assumes that it is optimal for a firm not to purchase the restricted non-audit services from its auditor. If indeed firms used non-audit services opportunistically to bribe the auditor, the provision can help correct the misbehavior and increase shareholder value, especially for those clients purchasing more and thus deviating more from the ‘‘optimal’’. As Greenstone et al. (2006) suggest, firms in less compliance with new rules likely face more diversion by insiders and therefore benefit more from regulations. On the other hand, if auditor-provided non-audit services are cost efficient, the restriction would destroy value by eliminating such cost-efficiency. Ribstein (2002) argues that the restriction could block ‘‘knowledge spillovers’’ that give auditors access to valuable information. Auditors will incur greater costs to gain the institutional knowledge that they could have obtained while performing consulting services for their client. New business consultants will also make greater initial investments than the auditor. These
3 The prohibited services include: bookkeeping or other services related to the accounting records of the audit
client; financial information system design and implementation; appraisal or valuation, fairness opinions, or
contribution-in-kind reports; actuarial services; internal audit outsourcing; management functions or human
resources; broker or dealer, investment advisor, or investment banking services; legal and expert services unrelated
to the audit.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11582
incremental start-up costs will be borne by the client. This argument predicts firms that purchased more non-audit services will lose more as a result of the rule change. If the restriction is costly on net, firms purchasing more non-audit services would experience more negative returns. H2: If the restriction on auditor’s provision of non-audit services imposed net costs on firms,
firms purchasing more non-audit services prior to SOX would incur greater costs and
experience more negative cumulative abnormal returns around the SOX rulemaking events.
3.2.2. Provisions on corporate accountability and criminal penalties
SOX prohibits firms from arranging executive loans (Section 402), requires CEOs and CFOs to certify annual and quarterly reports to the SEC (Section 302), and raises criminal penalties for corporate fraud and white-collar crimes. The statute of limitations for security lawsuits is extended from three to five years after the misconduct (Section 804). Moreover, SOX directs the U.S. Sentencing Commission to tighten sentencing guidelines for fraud and white-collar crimes (Sections 805, 905, and 1104). Section 305 lowers the standard to bar individuals from serving as an officer or director. Although the latter changes are not very specific, they indicate a change in attitude towards corporate crimes, and can be expected to influence subsequent legislation and enforcement. These provisions intend to protect shareholders better by improving the accountability
of executives and directors, thereby increasing shareholder value. It is implicitly assumed that at least some firms had not established an optimal governance structure and the new laws can tighten shareholder controls and deter management opportunism in these firms. If the provisions are effective, a firm with weak shareholder protection would benefit more than firms with optimum shareholder protection. However, these provisions also give rise to significant private costs. Some commentators
allege that SOX increased the difficulty of finding qualified directors, compensation of directors, and costs of directors’ and officers’ insurance policies (Francis, 2003). Further, Harris and Raviv (2005) argue that managers’ information can be better exploited when they are in control. Tightening shareholder controls would thus impose costs by reducing management’s flexibility in decision making. Also, Core et al. (2005) argue that while weak governance inhibits the removal of incompetent managers, it can encourage managers to behave in a less risk-averse manner by shielding them from the consequences of lower-tail outcomes of good projects. This suggests that tightening shareholder controls can result in managers foregoing valuable risky investment projects. Thus, firms with governance weaker than optimum are expected to make greater changes after SOX and incur greater costs. If the costs of SOX outweigh the potential benefits, firms with shareholder control weaker than optimum would exhibit more negative returns in the SOX rulemaking period. H3: If the governance provisions of SOX imposed net costs on firms, firms with corporate
governance structure weaker than optimum would incur more costs and experience more
negative cumulative abnormal returns around the SOX rulemaking events.
3.2.3. Provisions on internal controls
SOX directs the SEC to set rules that require management to document and assess the effectiveness of internal controls. It also directs the PCAOB to prescribe rules requiring the auditor to attest to and report on management’s assessment (Section 404). Section 404 is considered one of the most significant changes in financial reporting and a major direct cost driver of SOX. The provision could benefit firms by tightening internal controls and
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 83
reducing opportunistic behavior, but it is also associated with substantial initial implementation and subsequent continuing costs. A.R.C. Morgan examines the actual Section 404 cost disclosures and finds that the average initial compliance costs range from $1.56 million for firms with annual sales less than $250 million to $10 million for firms with annual sales between $7 billion and $10 billion.
4
The PCAOB auditing rules regarding Section 404 require first the firm and then its auditor to sample each class of transactions and perform a walk-through when evaluating management’s assessment of controls. This suggests that firms with more business lines will incur greater compliance costs. Also, firms with foreign operations or transactions will likely incur greater costs, as their transactions are more complicated, and the documentation and evaluation of internal controls are more costly. On the other hand, Ge and McVay (2005) document that firms with complicated business are more likely to report material control weaknesses. If investors expect the complexity of a firm’s transactions increases the likelihood of a material weakness, the benefits of Section 404 would increase with the complexity of firms’ business. Thus, if the costs of complying with Section 404 outweigh the benefits of discovering and remedying control weaknesses, firms with more business lines or foreign operations would experience more negative returns in the rulemaking period.
H4: If Section 404 imposed net costs on firms, firms with more complex business would incur greater costs and experience more negative cumulative abnormal returns around the
SOX rulemaking events.
The SEC adopted rules regarding management’s report on internal controls on May 27, 2003. In the original proposal, firms were required to comply with Section 404 for fiscal years ending on or after September 15, 2003; in the final rule, accelerated filers are required to comply for fiscal years ending on or after June 15, 2004 and non-accelerated filers are required to comply from the fiscal years ending on or after April 15, 2005.
5 Further, as the
compliance date was not extended by one full year or two full years, firms with different fiscal year ends obtained different extension periods. If firms’ internal controls are weak on average, then postponing the compliance date of Section 404 constitutes bad news to investors. In this case, late adopters would experience lower abnormal returns than early adopters. Alternatively, deferring the compliance date reduces additional costs that would be incurred to meet a tight deadline. In particular, late adopters avoid competing with early adopters for auditing resources to comply with Section 404. Also, early adopters will develop techniques and procedures in their control tests that will be useful for late adopters. Thus, late adopters would incur lower compliance costs and experience higher abnormal returns around this announcement, if the cost savings from postponing Section 404 are significant.
H5: If the deferment of Section 404 compliance was beneficial, firms that obtained a longer extension period would experience more positive abnormal returns than firms that were
required to comply with Section 404 earlier around the announcement of postponing the
compliance dates.
4 See A.R.C. Morgan’s report on Sarbanes-Oxley implementation costs at http://www.auditnet.org/articles/
Sarbanes-Oxley_Implementation_Costs.pdf. 5 A firm is considered an ‘‘accelerated filer’’ if: its common equity public float was $75 million or more as of the
last business day of its most recent second fiscal quarter; the firm has been subject to the requirements of
Section 13(a) or 15(d) of the Exchange Act for at least 12 calendar months; the firm has previously filed at least
one annual report pursuant to the Exchange Act; and, the firm is not eligible to use Forms 10-KSB and 10-QSB.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11584
3.2.4. Provisions on incentive pay and insider trading
If there is an accounting restatement as a result of material misconduct, SOX requires CEOs and CFOs to reimburse any incentive-based compensation or profits from the sale of stock received in the 12 months after the misreporting (Section 304). Additionally, executives are prohibited from selling company stock during pension blackout periods and are required to file a more timely disclosure of sales or purchases of company stock (Section 306).
6
There has been a general concern that stock option grants and other incentive-based compensation provide managers with incentives to manipulate accounting numbers and eventually stock prices in order to maximize their personal wealth (e.g., Kristof and Flanigan, 2002). The new rule on incentive pay and insider trading thus aims to reduce such incentives and deter misreporting. If the provisions are effective, firms that used incentive-based compensation excessively would benefit. However, these provisions could impose private costs on firms. Incentive-based
compensation is designed to align managers’ interest with that of shareholders and reward managers based on their performance. Reducing incentive pay in the compensation package would weaken the motivation of managers. Holmstrom and Kaplan (2003) also argue that the disgorgement provision increases the risk to CEOs or CFOs of selling a large amount of stock or options, as ‘‘misconduct’’ is not clearly defined. Putting the incentive pay at stake can cause CEOs to act more conservatively than shareholders would prefer (Ribstein, 2002). Consistent with this view, Cohen et al. (2004b) document a significant decline in both the ratio of incentive compensation to salary and the outlays in R&D and capital expenditures after SOX. These arguments suggest that firms compensating their CEOs with incentive pay excessively would incur more adjustment costs as a result of the rule change. If the net effect of the change is negative, these firms would experience more negative abnormal returns in the SOX event period. H6: If the provisions on incentive pay and insider trading imposed net costs on firms, firms
that use incentive-based compensation excessively would incur greater costs and experience
more negative cumulative abnormal returns around the SOX rulemaking events.
4. Empirical tests and results
This section discusses empirical tests of the implications of SOX. First, I examine the overall U.S. market reaction to events leading up to SOX passage. Returns of major foreign markets are used to evaluate abnormal U.S. returns in the rulemaking period. Contemporaneous confounding events are also investigated in detail. Second, cross- sectional regressions are estimated to test H2–H6. Third, I examine market reactions to related SOX rulemaking events after its passage and discuss alternative explanations and sensitivity analyses.
4.1. Overall U.S. market reaction to the legislative events leading to SOX
4.1.1. Research design
Since SOX provisions affect every listed U.S. firm, I investigate changes of the U.S. market index around the legislative events. However, changes of the U.S. market index
6 Pension blackout period is defined in Section 306 (a) (4) of the Act as any period during which the majority of
plan beneficiaries are prohibited from trading on any equity of the firm held in the plans.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 85
capture not only the impact of SOX events but other contemporary economic news as well. It is thus necessary to control for normal U.S. market returns absent the impact of SOX.
I use foreign firm returns in the SOX rulemaking period to evaluate the impact of other news on the U.S. market in that period.
7 Extant finance research suggests that U.S. and
foreign firms are exposed to substantial common economic news (e.g., Eun and Shim, 1989; Hamao et al., 1990). Meanwhile, foreign firms, except for those listed on U.S. exchanges, do not comply with SOX.
8 Returns of non-U.S.-traded foreign firms in the
SOX period thus reflected the impact of other economic news common to the U.S. and foreign markets but were not directly affected by SOX.
I examine stock returns of the U.S. and 22 major developed foreign stock markets during the SOX rulemaking period. U.S. return data are obtained from CRSP and foreign stock price data are from Datastream. Stocks with a price of less than one unit of local currency at the beginning of each year are deleted to avoid the small price effect (Blume and Stambough, 1983).
U.S.-exchange-listed foreign firms are excluded in computing the foreign returns, as they are required to comply with SOX. The cross-listing information is obtained from the Bank of New York Depositary Receipt Directory and cross-checked with data obtained from the NYSE, AMEX, and Nasdaq. U.S.-traded foreign firms that do not comply with SOX are also excluded. Numerous foreign firms have started with pink sheets trading (Level I ADR) and then upgraded to exchange listing (Level II or Level III ADR, see the Bank of New York’s overview of ADRs). Thus, if SOX is costly, the expected costs of the option to cross-list would increase and non-complying U.S.-traded foreign firms, which are more likely to exercise the option than non-U.S.-traded firms, would be negatively affected.
9
The list of Level I ADRs is also obtained from the Bank of New York Depositary Receipt Directory. The National Quotation Bureau’s pink sheets are used to identify Canadian firms traded by pink sheets. Table 1 Panel A reports the distribution of non- U.S.-traded foreign sample by country. Japan and the U.K. have the largest representation in the sample.
The comparison of returns across countries is complicated by time zone differences. The morning trading of day t in the U.S. overlaps the late afternoon trading in most European countries for one and a half hours, while the U.S. market opens on day t U.S. time after the Asian markets close on day t local time. As a result of the trading hour differences and complicated lead-lag structure between U.S. and foreign returns, the following regression is estimated to model the relation between U.S. and foreign returns,
US_Rett ¼ a0 þ a1 CAN_Rett þ a2 EU_Rett þ a3 EU_Rettþ1
þ a4 AS_Rett þ a5 AS_Rettþ1 þ et. ð1aÞ
7 In a concurrent independent study, Litvak (2005) uses returns of non-U.S.-traded foreign firms to estimate
abnormal returns of U.S.-listed foreign firms in the SOX rulemaking period. She finds that U.S.-listed foreign
firms experienced significantly negative returns around certain SOX events. 8 Foreign firms reporting under Section 13(a) or 15(d) of the Exchange Act are required to comply with SOX.
See the SEC final rule release, ‘‘Certification of Disclosure in Companies’Quarterly and Annual Reports’’, for an
example of how SOX applies to foreign private issuers. 9 Inclusion of these firms in estimating U.S. normal returns will reduce the estimated cumulative abnormal
return by about 0.5%, but the inference does not change.
ARTICLE IN PRESS
Table 1
Relation between returns of U.S. firms and non-U.S.-traded foreign firms
Panel A: Distribution of non-U.S.-traded foreign firms by their home country
Non-U.S.-traded
# Of firms % Of total
Europe
Austria 73 0.90
Belgium 139 1.71
Denmark 155 1.91
Finland 114 1.40
France 643 7.90
Germany 805 9.90
Ireland 20 0.25
Italy 196 2.41
Luxembourg 33 0.41
Netherlands 104 1.28
Norway 110 1.35
Portugal 56 0.69
Spain 96 1.18
Sweden 261 3.21
Switzerland 207 2.54
U.K. 953 11.71
North America
Canada 706 8.68
Asia
Hong Kong 214 2.63
Japan 2811 34.55
Singapore 58 0.71
Oceania
Australia 333 4.09
New Zealand 48 0.59
Total 8135 100.00
Panel: B Correlation between returns of U.S. and non-U.S.-traded foreign firms
US_Rett CAN_Rett EU_Rett EU_Rettþ1 AS_Rett
CAN_Rett 0.7812
EU_Rett 0.6854 0.6693
EU_Rettþ1 0.3690 0.2467 0.2858
AS_Rett 0.2466 0.3332 0.3609 �0.0537
AS_Rettþ1 0.3931 0.2703 0.2645 0.3505 0.1312
Panel: C Regressions of returns of U.S. firms on returns of non-U.S.-traded foreign firms
Model (1a) Model (1b)
Estimate t-stat Estimate t-stat
Intercept 0.0000 0.07 Intercept �0.0002 �0.21
CAN_Rett 0.6243 7.17 ***
CAN_Rett 0.8751 12.39 ***
EU_Rett 0.2773 3.29 ***
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11586
ARTICLE IN PRESS
Table 1 (continued )
Model (1a) Model (1b)
Estimate t-stat Estimate t-stat
EU_Rettþ1 0.1021 1.64
AS_Rett �0.0529 �0.74
AS_Rettþ1 0.1650 2.30 **
Adj. R2 68.18% 60.63%
This table reports the distribution of non-U.S.-traded foreign firms by their home country, the correlation
between U.S. and foreign returns in the estimation period, and the estimation results of models (1a) and (1b) on
the relation between U.S. and foreign returns.
Panel A reports the distribution of non-U.S.-traded foreign firms by their home country. The U.S.-listed or U.S.-
traded foreign firms and firms with a price of less than one unit of local currency are excluded. The U.S.-listing
information is obtained from the Bank of New York Depositary Receipt Directory and cross-checked with data
obtained from the NYSE, AMEX, and Nasdaq. Non-Canadian firms are classified based on their cross-listing
status as of the end of 2002. The National Quotation Bureau’s pink sheets are used to identify Canadian firms
traded over the counter by pink sheets. The classification of Canadian firms is based on their status as of January
2006.
Panel B reports the correlation between value-weighted U.S. and foreign returns in the estimation period.
US_Rett represents the value-weighted raw return of the U.S. portfolio on day t U.S. time. CAN_Rett, EU_Rett,
and AS_Rett denote value-weighted returns of non-U.S.-traded firms in Canada, Europe, Asia (including
Australia and New Zealand) on day t local time. The market capitalization of each stock in U.S. dollars on day
t �1 is used to compute the weight of returns of day t. Correlation coefficients that are significantly different from
zero at less than 5% level are presented in bold.
Panel C presents the estimation results of the following regressions.
US_Rett ¼ a0 þ a1 CAN_Rett þ a2 EU_Rett þ a3 EU_Rettþ1 þ a4 AS_Rett þ a5 AS_Rettþ1 þ et, (1a)
US_Rett ¼ a0 þ a1 CAN_Rett þ et. (1b)
US_Rett represents the value-weighted raw return of the U.S. portfolio on day t U.S. time. CAN_Rett, EU_Rett,
and AS_Rett denote value-weighted returns of non-U.S.-traded firms in Canada, Europe, Asia (including
Australia and New Zealand) on day t local time. The market capitalization of each stock in U.S. dollars on day
t �1 is used to compute the weight of returns of day t. The regressions are estimated using daily return data in the
100 days prior to December 28, 2001 to capture the relation between U.S. returns and returns of non-U.S.-traded
foreign firms.
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 87
US_Rett represents returns of the U.S. market on day t U.S. time. CAN_Rett, EU_Rett, and AS_Rett denote returns of non-U.S.-traded firms in Canada, Europe, Asia (including Australia and New Zealand) on day t local time.
Table 1 Panel B reports the correlation of value-weighted daily returns of stock markets in different zones in the estimation period. The market capitalization of each stock in U.S. dollars on day t �1 is used to compute the weight of returns of day t. The exchange rates are obtained from the Compustat Global Currency file. CAN_Rett is most highly correlated with US_Rett with a correlation coefficient of 0.7812. Both EU_Rett (AS_Rett) and EU_Rettþ1 (AS_Rettþ1) are positively correlated with US_Rett, suggesting that it is necessary to include both returns in predicting US_Rett. However, as EU_Rett (EU_Rettþ1) reflects news in U.S. returns of day t �1 (t þ1) and day t, if significant economic news arrives consecutively in the examined event period but not in the estimation period, the estimate of US_Rett based on the returns of two consecutive days could be
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11588
biased. As an alternative, I also use CAN_Rett alone, which exhibits the highest correlation with US_Rett, to estimate normal U.S. returns,
US_Rett ¼ b0 þ b1 CAN_Rett þ et. (1b)
Regressions (1a) and (1b) are estimated using daily return data in the 100 days prior to December 28, 2001.
10 The estimation results based on value-weighted returns are reported
in Table 1 Panel C. The adjusted R2 is 68.18% for regression (1a) and 60.63% for regression (1b). The high R2s are consistent with prior findings that national stock markets are highly correlated. The inclusion of the European and Asian market returns moderately improves the explanatory power of the model.
4.1.2. Cumulative market returns in the U.S. over the event period
The parameter estimates in Table 1 Panel C are then used to predict U.S. normal returns absent in the impact of SOX in the SOX event period. The abnormal returns are computed as the prediction errors (e.g., ARt ¼ US_Rett �ðb̂0 þ b̂1 CAN_RettÞ). I first conduct t-tests based on the standard deviation of prediction errors in the estimation period. As the variability of prediction errors in 2002 can differ from that in the estimation period and returns of consecutive event days can be autocorrelated, I also test the statistical significance of U.S. abnormal returns using the bootstrapping method.
11
Table 2 reports value-weighted raw and abnormal returns around the legislative events leading to the passage of SOX. No legislative events prior to July 2002, except for event 2, were associated with a significant revision of market’s expectations, consistent with what news reports revealed: the probability of passing laws was generally considered remote before July (e.g., Cummings et al., 2002). The most momentous SOX rulemaking activities occurred in July 2002 and were
associated with the most significant market reactions. President Bush’s speech on July 9 (event 14) emphasized the importance of government-mandated corporate reforms.
12
Romano (2004) noted that the President unexpectedly proposed tougher rule changes, such as the prohibition of corporate loans to executives, which even Sen. Sarbanes did not include in his bill. His speech was viewed by the media as signaling a change of attitude in Washington and a change in the balance of power between the federal government and corporations (Cummings et al., 2002). Cummings et al. (2002) also argue that President Bush’s expressed determination towards reforms significantly increased the likelihood of law passage. Meanwhile, the Senate debated Sen. Sarbanes’ bill from July 8–12. Several amendments, such as the amendment to impose tougher penalties for corporate wrongdoings, were passed to add teeth to the bill (Romano, 2004). The abnormal return estimated using models (1a) and (1b) is �1:55% and �3:10%, respectively. While the
10 In order to avoid mismatching market reactions, a trading day is excluded if foreign markets open on the
prior week-day while the U.S. market closes. For example, the U.S. market stopped trading in the week after the
September 11th terrorist attack while other markets continued trading. The impact of the event was reflected in
U.S. returns on September 17th, when the stock market resumed trading, while it had been incorporated in stock
prices in other countries. Thus, relating the returns on September 17th of U.S. and foreign markets would cause a
mismatch. 11 Following Lo (2003) and Busse and Green (2002), the bootstrapped p-values are calculated as follows: (1) for
each event, returns of the same number of consecutive non-event days in 2002 are computed (sample A); (2) a
sample (sample B) of 1000 returns is drawn with replacement from sample A; (3) one-tailed p-values are computed
as: p ¼ the number of observations in sample B with values greater (lower) than Rt=1000, if Rt4 ¼ 0 (if Rto0). 12 See the content of the speech at http://www.whitehouse.gov/news/releases/2002/07/20020709-4.html.
ARTICLE IN PRESS T a b le
2
R a w
a n d a b n o rm
a l U .S . m a rk e t re tu rn s a ro u n d le g is la ti v e e v e n ts
le a d in g to
th e p a ss a g e o f th e S a rb a n es – O x le y A c t
P a n el
A : R a w
a n d a b n o rm
a l re tu rn s a ro u n d in d iv id u a l S O X
e v en ts
E v e n t
D a te
D e sc ri p ti o n
E v e n t
w in d o w
V W
R E
T t- st a t
B o o ts tr a p
p -v a lu e
A R 1
t- st a t
B o o ts tr a p
p -v a lu e
A R 2
t- st a t
B o o ts tr a p
p -v a lu e
1 1 /1 7 /2 0 0 2
S E C
C h a ir m a n p ro p o se d a n
a c c o u n ti n g o v e rh a u l p la n
1 /1 5 – 1 8
� 0 .0 1 0 3
� 0 .4 4
0 .4 0 0 0
� 0 .0 1 3 2
� 1 .0 2
0 .1 5 7 0
� 0 .0 1 0 2
� 0 .6 9
0 .2 8 5 0
2 2 /2 /2 0 0 2
T re a su ry
S e c re ta ry
c a ll e d fo r
c h a n g es
in ru le s g o v e rn in g
c o rp o ra ti o n s
2 /1 – 4
� 0 .0 3 0 3
� 1 .8 2 *
0 .0 4 6 0
� 0 .0 2 4 8
� 2 .7 1 * * *
0 .0 2 5 0
� 0 .0 2 9 7
� 2 .8 7 * * *
0 .0 2 5 0
3 2 /1 3 /2 0 0 2
O x le y in tr o d u c e d a n
a c c o u n ti n g re fo rm
b il l in
th e
H o u se
F in a n c ia l S e rv ic e s
C o m m it te e
2 /1 1 – 1 4
0 .0 1 7 3
0 .7 4
0 .1 4 0 0
0 .0 0 2 6
0 .2 0
0 .3 6 4 0
0 .0 0 8 2
0 .5 6
0 .2 1 8 0
4 2 /2 8 /2 0 0 2
H o u se
D e m o c ra ts
in tr o d u c e d
le g is la ti o n th a t w o u ld
im p o se
m o re
re st ri c ti o n s th a n O x le y ’s
p ro p o sa l
2 /2 7 – 3 /1
0 .0 2 0 3
1 .0 0
0 .1 5 1 0
� 0 .0 0 1 4
� 0 .1 3
0 .4 2 5 0
0 .0 1 1 2
0 .8 8
0 .1 9 5 0
5 3 /7 /2 0 0 2
B u sh ’s fi rs t re sp o n se
to
a c c o u n ti n g sc a n d a ls u n v ei le d
3 /6 – 8
0 .0 1 7 3
0 .8 5
0 .1 6 3 0
0 .0 0 0 9
0 .0 8
0 .4 6 8 0
0 .0 0 8 2
0 .6 5
0 .2 0 7 0
6 3 /2 6 /2 0 0 2
G re e n sp a n w a rn e d a g a in st
to o m u c h re g u la ti o n
3 /2 5 – 2 7
� 0 .0 0 3 0
� 0 .1 5
0 .5 0 9 0
� 0 .0 0 2 3
� 0 .2 1
0 .3 7 5 0
� 0 .0 0 0 5
� 0 .0 4
0 .5 5 9 0
7 4 /1 1 /2 0 0 2
H o u se
F in a n c ia l S e rv ic e s
C o m m it te e sc h e d u le d to
v o te
O x le y ’s b il l, b u t th e v o te
w a s
p o st p o n e d
4 /1 1 – 1 2
� 0 .0 1 3 5
� 0 .8 1
0 .2 4 7 0
� 0 .0 1 4 8
� 1 .6 2
0 .1 0 6 0
� 0 .0 1 5 0
� 1 .4 5
0 .1 1 2 0
8 4 /1 6 /2 0 0 2
O x le y ’s b il l p a ss e d in
th e
C o m m it te e
4 /1 6 – 1 7
0 .0 2 0 3
1 .2 2
0 .1 2 2 0
0 .0 0 6 2
0 .6 8
0 .2 7 6 0
0 .0 0 5 8
0 .5 5
0 .2 6 9 0
9 4 /2 4 /2 0 0 2
O x le y ’s b il l p a ss e d in
th e
H o u se
4 /2 4 – 2 6
� 0 .0 2 1 3
� 1 .0 5
0 .1 8 6 0
� 0 .0 0 5 8
� 0 .5 2
0 .2 7 1 0
� 0 .0 0 7 9
� 0 .6 2
0 .2 7 5 0
4 /2 5 /2 0 0 2
S e n a te
J u d ic ia ry
C o m m it te e
a p p ro v e d le g is la ti o n
b o ls te ri n g c o rp o ra te
fr a u d
la w s
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 89
ARTICLE IN PRESS T a b le
2 (c
o n ti
n u e d )
P a n el
A : R a w
a n d a b n o rm
a l re tu rn s a ro u n d in d iv id u a l S O X
e v en ts
E v e n t
D a te
D e sc ri p ti o n
E v e n t
w in d o w
V W
R E
T t- st a t
B o o ts tr a p
p -v a lu e
A R 1
t- st a t
B o o ts tr a p
p -v a lu e
A R 2
t- st a t
B o o ts tr a p
p -v a lu e
1 0
5 /8 /2 0 0 2
S a rb a n e s c ir c u la te d h is re fo rm
b il l in
th e S e n a te
B a n k in g
C o m m it te e
5 /7 – 9
0 .0 1 6 9
0 .8 3
0 .1 6 3 0
0 .0 0 8 8
0 .7 8
0 .2 3 0 0
0 .0 0 9 8
0 .7 7
0 .1 9 5 0
1 1
6 /1 1 /2 0 0 2
D e m o c ra ts
in S e n a te
B a n k in g
C o m m it te e u n it e d b e h in d
S a rb a n e s’ b il l
6 /1 0 – 1 3
� 0 .0 2 0 1
� 0 .8 6
0 .2 2 0 0
0 .0 0 5 1
0 .3 9
0 .3 0 6 0
� 0 .0 0 1 1
� 0 .0 7
0 .5 7 9 0
6 /1 2 /2 0 0 2
S E C p ro p o se d ru le s to
re q u ir e
e x e cu ti v e s to
c e rt if y fi n a n c ia l
re p o rt s
1 2
6 /1 8 /2 0 0 2
S e n a te
B a n k in g C o m m it te e
p a ss ed
S a rb a n e s’ b il l
6 /1 8 – 1 9
� 0 .0 1 5 5
� 0 .9 3
0 .2 0 6 0
� 0 .0 0 4 2
� 0 .4 6
0 .6 7 4 0
� 0 .0 0 9 6
� 0 .9 2
0 .1 7 8 0
1 3
6 /2 5 /2 0 0 2
W o rl d C o m
a d m it te d th a t th e y
u n d er st a te d e x p e n se s b y $ 3 .8
b il li o n
6 /2 5 – 2 7
� 0 .0 0 1 7
� 0 .0 8
0 .5 2 5 0
0 .0 0 3 3
0 .2 9
0 .3 6 2 0
0 .0 1 2 8
1 .0 1
0 .1 8 1 0
1 4
7 /8 – 7 /1 2
S e n a te
d e b a te d S a rb a n e s’ b il l
7 /8 – 1 2
� 0 .0 6 4 2
� 2 .4 4 * *
0 .0 2 5 0
� 0 .0 1 5 5
� 1 .0 7
0 .0 3 6 0
� 0 .0 3 1 0
� 1 .8 9 *
0 .0 0 0 0
7 /9 /2 0 0 2
B u sh
d el iv e re d a sp e e ch
o n
c o rp o ra te
re fo rm
s; p a ss a g e o f
S a rb a n e s’ b il l li k e ly
7 /1 0 /2 0 0 2
S e n a te
p a ss e d a to u g h
a m en d m e n t to
st re n g th en
c ri m in a l p e n a lt ie s 9 7 to
0
1 5
7 /1 5 /2 0 0 2
S e n a te
p a ss e d S a rb a n e s’ b il l
7 /1 5 – 1 7
� 0 .0 1 2 9
� 0 .6 3
0 .3 5 3 0
0 .0 0 0 8
0 .0 7
0 .4 6 8 0
0 .0 0 3 0
0 .2 3
0 .3 3 1 0
7 /1 6 /2 0 0 2
H o u se
p a ss e d b il l to
st re n g th e n c ri m in a l p e n a lt ie s
1 6
7 /1 8 /2 0 0 2
H o u se
R e p u b li ca n le a d e rs
re p o rt e d ly
re tr e a te d fr o m
e ff o rt s to
d il u te
th e S e n a te ’s
to u g h b il l
7 /1 8 – 2 3
� 0 .1 2 0 0
� 5 .1 1 * * *
0 .0 0 0 0
� 0 .0 4 9 4
� 3 .8 3 * * *
0 .0 0 9 0
� 0 .0 6 3 8
� 4 .3 4 * * *
0 .0 0 9 0
7 /1 9 /2 0 0 2
C o n fe re n c e c o m m it te e st a rt e d
n e g o ti a ti o n s to
m e rg e b il ls
a n d S e n a te ’s b il l b e ca m e th e
fr a m ew
o rk ; n e g o ti a ti o n
c o n ti n u e d o v e r th e w e ek e n d
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11590
ARTICLE IN PRESS 7 /2 0 /2 0 0 2
B u sh
p u sh ed
to sp e e d u p
ru le m a k in g in
a ra d io
a d d re ss
7 /2 3 /2 0 0 2
L o b b y is ts re p o rt ed ly
lo st th e ir
im p a ct
1 7
7 /2 4 /2 0 0 2
S e n a te
a n d H o u se
a g re e d o n
th e fi n a l ru le
7 /2 4 – 2 6
0 .0 6 1 1
3 .0 0 * * *
0 .0 2 7 0
0 .0 4 5 1
4 .0 3 * * *
0 .0 1 0 0
0 .0 4 2 4
3 .3 3 * * *
0 .0 1 0 0
7 /2 5 /2 0 0 2
S e n a te
a n d H o u se
p a ss e d
S O X
C V
W R
E T
(1 )
t- st a t
B o o ts tr a p
p -v a lu e
U .S .- (1 )
t- st a t
B o o ts tr a p
p -v a lu e
C E
W R
E T
(2 )
t- st a t
B o o ts tr a p
p -v a lu e
U .S .- (2 )
t- st a t
B o o ts tr a p
p -v a lu e
P a n e l: B C u m u la ti v e ra w
re tu rn s a ro u n d th e si g n ifi c a n t S O X
e v e n ts
(e v en ts
2 , 1 4 , 1 6 , a n d 1 7 )
U .S .
� 0 .1 5 3 5
� 3 :4 9 � � �
0 .0 0 0 0
— —
— � 0 .1 2 5 3
� 3 :2 7 � � �
0 .0 0 0 0
— —
—
C a n a d a
� 0 .0 7 9 1
� 2 :0 1 � �
0 .0 5 1 0
� 0 .0 7 4 4
� 2 :6 7 � � �
0 .0 0 0 0
� 0 .0 6 0 7
� 2 :4 1 � �
0 .0 0 2 0
� 0 .0 6 4 5
� 2 :2 7 � �
0 .0 3 1 0
E u ro p e
C u m u la te d o v e r th e sa m e
c a le n d a r d a te s a s th e U .S .
� 0 .1 2 3 3
� 2 :9 7 � � �
0 .0 2 0 0
� 0 .0 3 0 2
� 0 .8 9
0 .0 1 3 0
� 0 .0 5 6 3
� 1 :9 1 �
0 .2 2 9 0
� 0 .0 6 8 9
� 2 :4 8 � �
0 .0 0 0 0
C u m u la te d o v e r th e n e x t d a y
a ft e r th e U .S . d a te s
� 0 .1 3 5 5
� 3 :0 2 � � �
0 .0 0 3 0
� 0 .0 1 8 0
� 0 .3 6
0 .0 5 2 0
� 0 .0 7 0 1
� 2 :2 5 � �
0 .0 9 8 0
� 0 .0 5 5 1
� 1 .6 1
0 .0 0 0 0
A si a
C u m u la te d o v e r th e sa m e
c a le n d a r d a te s a s th e U .S .
� 0 .0 7 6 2
� 1 :9 7 � �
0 .0 0 1 0
� 0 .0 7 7 2
� 1 .5 2
0 .0 5 6 0
� 0 .0 3 2 1
� 1 .3 7
0 .1 0 7 0
� 0 .0 9 3 2
� 2 :3 3 � �
0 .0 0 4 0
C u m u la te d o v e r th e n e x t d a y
a ft e r th e U .S . d a te s
� 0 .0 9 8 1
� 2 :5 8 � � �
0 .0 0 0 0
� 0 .0 5 5 3
� 1 .2 2
0 .1 4 8 0
� 0 .0 4 9 4
� 2 :1 2 � �
0 .0 1 2 0
� 0 .0 7 5 9
� 2 :1 4 � �
0 .0 1 9 0
P a n e l: C
C u m u la ti v e ra w
re tu rn s a ro u n d a ll S O X
e v e n ts
U .S .
� 0 .1 5 9 6
� 1 .8 7 *
0 .1 4 1 0
— —
— � 0 .1 2 9 3
� 1 .7 3 *
0 .2 5 9 0
— —
—
C a n a d a
� 0 .0 9 6 1
� 1 .2 6
0 .2 9 8 0
� 0 .0 6 3 4
� 1 .1 7
0 .2 0 6 0
� 0 .0 2 7 2
� 0 .5 5
0 .1 8 1 0
� 0 .1 0 2 1
� 1 .8 4 *
0 .5 9 9 0
E u ro p e
C u m u la te d o v e r th e sa m e
c a le n d a r d a te s a s th e U .S .
� 0 .1 2 7 3
� 1 .5 8
0 .5 7 7 0
� 0 .0 3 2 3
� 0 .4 9
0 .0 5 7 0
� 0 .0 7 5 8
� 1 .3 2
0 .7 7 0 0
� 0 .0 5 3 5
� 0 .9 9
0 .0 7 6 0
C u m u la te d o v e r th e n e x t d a y
a ft e r th e U .S . d a te s
� 0 .1 3 2 9
� 1 .5 2
0 .4 6 5 0
� 0 .0 2 6 7
� 0 .2 7
0 .0 9 3 0
� 0 .0 8 5 2
� 1 .4 0
0 .6 8 6 0
� 0 .0 4 4 0
� 0 .6 6
0 .1 0 5 0
A si a
C u m u la te d o v e r th e sa m e
c a le n d a r d a te s a s th e U .S .
� 0 .0 6 5 6
� 0 .8 7
0 .0 3 9 5
� 0 .0 9 4 0
� 0 .9 5
0 .3 4 9 0
0 .0 0 1 7
0 .0 4
0 .3 8 2 0
� 0 .1 3 1 0
� 1 .6 9 *
0 .2 1 6 0
C u m u la te d o v e r th e n e x t d a y
a ft e r th e U .S . d a te s
0 .0 1 2 8
0 .1 7
0 .2 4 8 0
� 0 .1 7 2 4
� 1 .9 5 *
0 .0 6 2 0
0 .0 0 9 9
0 .2 2
0 .3 0 9 0
� 0 .1 3 9 1
� 2 .0 2 * *
0 .1 6 7 0
V a lu e -w
e ig h te d
t- st a t
B o o ts tr a p
p -v a lu e
E q u a l- w e ig h te d
t- st a t
B o o ts tr a p
p -v a lu e
P a n e l: D
C u m u la ti v e a b n o rm
a l re tu rn s
C u m u la te d o v e r si g n ifi c a n t S O X
e v e n ts
C A
R 1
� 0 .0 4 4 6
� 1 :8 5 �
0 .0 0 1 0
� 0 .0 3 7 6
� 1 :6 9 �
0 .0 1 2 0
C A
R 2
� 0 .0 8 2 1
� 2 :9 9 � � �
0 .0 0 0 0
� 0 .0 5 2 8
� 1 :8 5 �
0 .0 3 5 0
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 91
ARTICLE IN PRESS
T a b le
2 (c
o n ti
n u e d )
V a lu e -w
e ig h te d
t- st a t
B o o ts tr a p
p -v a lu e
E q u a l- w e ig h te d
t- st a t
B o o ts tr a p
p -v a lu e
C u m u la te d o v e r a ll S O X
e v e n ts
C A
R 1
� 0 .0 5 8 9
� 1 .2 5
0 .0 8 0 0
� 0 .0 4 5 5
� 1 .0 5
0 .1 6 1 0
C A
R 2
� 0 .0 6 7 4
� 1 .2 6
0 .1 9 3 0
� 0 .0 6 1 3
� 1 .1 1
0 .6 1 2 0
T h is ta b le re p o rt s th e ra w a n d a b n o rm
a l re tu rn s in
th e U .S . a ro u n d le g is la ti v e e v en ts le a d in g to
th e p a ss a g e o f S O X . F o r e a c h e v e n t, th e e v en t w in d o w is d et e rm
in ed
u si n g th e fo ll o w in g p ro c e d u re : (1 ) F o r e v en ts re la te d to
sp e e c h es , if a n e w s re p o rt th a t c o v e re d m a jo r c o n te n ts o f th e sp e e ch
a p p ea re d in
th e
W S
J o r
W P o n o r b ef o re
th e d a te
th e sp e e ch
w a s d e li v e re d , th e e v e n t w in d o w c o v e rs
o n e d a y b ef o re
th e n e w s re p o rt to
o n e d a y a ft e r th e sp e e ch ; if a n e w s re p o rt a p p ea re d in
th e
W S
J o r
W P
a ft e r a sp e e c h , th e e v e n t w in d o w c o v e rs
o n e d a y b e fo re
th e sp e e ch
to o n e d a y a ft e r th e sp ee c h . (2 ) F o r e v en ts re la te d to
le g is la ti v e v o te s p ri o r to
th e p a ss a g e o f S O X , a s
th e re
is so m e u n c er ta in ty
a b o u t th e v o te
re su lt , th e e v e n t w in d o w
c o v e rs
th e v o ti n g d a y a n d o n e d a y a ft e r th e d a te . P a n el
A re p o rt s th e v a lu e -w
ei g h te d ra w
a n d
a b n o rm
a l re tu rn s a ro u n d in d iv id u a l e v e n ts .
V W
R E
T re p re se n ts v a lu e -w
e ig h te d ra w re tu rn s. A b n o rm
a l re tu rn s (A
R 1 o r
A R
2 ) a re
c o m p u te d a s th e d if fe re n c e b et w e e n
th e ra w v a lu e -w
e ig h te d U .S . re tu rn s a n d th e e x p e ct e d re tu rn s c o m p u te d b a se d o n th e m o d e ls (1 a ) o r (1 b ) in
T a b le 1 P a n e l C . P a n e ls B to
D re p o rt v a lu e - a n d e q u a l-
w e ig h te d ra w
a n d a b n o rm
a l re tu rn s c u m u la te d o v e r si g n ifi c a n t S O X
e v e n ts
o r a ll S O X
e v en ts .
C V
W R
E T (C
E W
R E
T ) re p re se n ts
c u m u la ti v e v a lu e -w
e ig h te d (e q u a l-
w e ig h te d ) ra w
re tu rn s. C u m u la ti v e a b n o rm
a l re tu rn s (C
A R
1 o r
C A
R 2 ) a re
c o m p u te d a s th e su m
o f th e d if fe re n c e s b e tw
e e n th e ra w
v a lu e - o r e q u a l- w e ig h te d U .S .
re tu rn s a n d th e e x p e ct e d v a lu e - o r e q u a l- w e ig h te d re tu rn s c o m p u te d b a se d o n th e m o d e ls (1 a ) o r (1 b ) in
T a b le
1 P a n e l C . T h e
t- st a ti st ic s a re
c o m p u te d u si n g th e
st a n d a rd
d e v ia ti o n o f ra w
re tu rn s o r th e p re d ic ti o n e rr o rs
in th e e st im
a ti o n p er io d . B o o ts tr a p p in g o n e -t a il e d
p -v a lu e s a re
o b ta in e d u si n g th e m et h o d d es c ri b e d in
fo o tn o te
1 0 .
* * * , * * , * in d ic a te
si g n ifi c a n c e a t 1 % , 5 %
a n d 1 0 %
le v e ls , re sp e c ti v e ly , tw
o -t a il e d te st s.
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11592
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 93
estimated abnormal return is smaller in magnitude than the raw return (�6:42%), showing that there was other news around this event, it is significant except in the t-test of the abnormal return of model (1a). The evidence is largely consistent with the hypothesis that these tough amendments were considered costly and/or signaling bad news.
Negative market returns were also observed when House Republicans reportedly gave up efforts to weaken the Senate’s tough bill around July 18 (event 16). Studying the congressional records, Romano (2004) argues that Republicans expected to dilute Sarbanes’ bill in the conference committee when they allowed the bill to go through the Senate, but their strategy failed in the presence of electoral concerns. The media seemed to share the expectation that Republicans would soften the bill, but found the opposite (see, e.g., Murray, 2002a). The media also revealed concerns that the final bill would impose even greater costs than Sarbanes’ proposal. Melloan (2002) commented that the unanimous passage of several proposals in the Senate was scary, as it signaled that most of the members did not care much about the contents of the bill. Hilzenrath et al. (2002) cite the comments of Sen. Gramm after the passage of SOX, ‘‘this bill could have been a lot worse,’’ as ‘‘virtually anything could have passed the Congress.’’ The estimated abnormal return (�4:94% and �6:38%, respectively, under models (1a) and (1b)) is highly significant, though smaller in magnitude than the raw return (�12:00%).
Given these concerns prior to the final ruling, the positive return on July 24 (event 17) is not surprising. The estimated abnormal return around event 17 is over 4% under both models, consistent with the explanation that the final announcement eliminated prior concerns for tougher rules. By July 23, the House and Senate negotiators still could not agree on several issues including the authority of the PCAOB, whether to prosecute executives that misstate financial reports without a criminal intent, and the extension of the statute of limitations for securities lawsuits (Murray, 2002b). The final rule increased the SECs control over the PCAOB and dropped the proposal to punish unintentional misstatement, but kept the extension of the statute of litigation limitations proposed by the Senate (Murray and Schroeder 2002). Thus, business lobbyists won part of the game.
To summarize, Table 2 Panel A shows that events 2, 14, 16, and 17 were associated with significant changes in investors’ expectations of the likelihood of passing tough rules. The market responded negatively to events that signaled increases in the likelihood of passing tough rules and positively to events revealing that no further costs would be imposed. Tests based on the equal-weighted returns produce similar inferences (untabulated).
Table 2 Panels B–D report the cumulative raw and abnormal returns in the SOX period. For completeness and also to show the robustness of the results, both value- and equal- weighted returns are reported. The equal-weighted abnormal returns are computed in the same way as the value-weighted abnormal returns. The daily raw and abnormal returns are then added up to compute the cumulative returns. As measurement errors in estimating regressions (1a) and (1b) could bias the coefficient estimates and subsequently the estimation of U.S. abnormal returns, I also directly compare the raw cumulative returns of the U.S. with those of the foreign markets.
Table 2 Panel B presents raw returns cumulated over the four significant SOX events and the differences in returns between the U.S. and the foreign portfolios. Cumulative value- and equal-weighted raw returns of the foreign portfolios around the key SOX events are negative and mostly significant, suggesting there was news other than SOX in this period. The differences between foreign and U.S. value-weighted (equal-weighted) cumulative returns range from �1:80% to �7:72% (�5:51% to �9:32%). While the
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11594
differences in equal-weighted returns are mostly significant, for value-weighted returns, only the difference between the U.S. and the Canadian portfolio (�7:44%) is statistically significant under alternative testing methods. Foreign portfolios can respond differently to global economic news than the U.S. portfolio. If so, the simple differences between U.S. and foreign returns would be biased estimates of U.S. abnormal returns.
13 Table 2 Panel C
reports cumulative raw returns around all SOX events. Most of these cumulative returns are insignificant. Finally, Panel D reports the cumulative abnormal returns of the U.S. around the key
SOX events and around all SOX events. For both value- and equal-weighted returns, the cumulative abnormal returns over the key SOX events are negative and statistically significant.
14 The cumulative abnormal returns computed based on value-weighted returns
are �4:46% and �8:21% using models (1a) and (1b), respectively, and are larger in magnitude than those based on equal-weighted returns (�3:76% and�5:28%). Small firms likely incur disproportionately greater direct costs to comply with SOX. This conjecture seems to suggest that estimated abnormal returns based on equal-weighted returns should be larger in magnitude than those using value-weighted returns. However, large firms could be subject to greater litigation and political costs as an indirect result of SOX or future legislation. Thus, it is unclear whether small or large firms experience greater losses in total. The cumulative abnormal returns around all SOX events are not statistically significant, ranging from �4:55% to �6:74%. This result does not support SOX being costly. Nevertheless, as Panel A shows that many SOX events did not significantly change investors’ expectations, the lack of statistical significance is likely due to the inclusion of insignificant event days reducing the power of the test. Overall, the results in Table 2 show that a large fraction of the negative returns that the
U.S. market experienced in the SOX rulemaking period could be due to other negative economic news; yet, the impact of SOX is still likely negative and statistically significant. This is consistent with the hypothesis that investors initially considered the new rule and/or the information it conveyed to be bad news. It should be noted in interpreting the results in Table 2 that SOX can indirectly affect
non-U.S.-traded foreign firms in several ways. First, non-U.S.-traded firms that compete with complying firms could obtain a competitive advantage if SOX imposed costs on complying firms. Thus, the return difference between U.S. firms and non-U.S.-traded foreign firms might overestimate the costs of SOX. However, on the other hand, there exists substantial interdependence in trade and financing activities among firms in different countries (Forbes and Rigobon, 2002). Although non-complying firms are not directly affected by SOX, they likely conduct international transactions, establish subsidiaries in
13 I estimate regressions of U.S. daily returns on each of the foreign market daily returns. Untabulated results
show that the coefficient on foreign returns is always smaller than one. It is rejected that the coefficient equals one,
except in the regression of U.S. returns on Canadian returns. Thus, either non-Canadian portfolios exhibit a
different sensitivity to global economic news than the U.S., or, due to local economic news, their returns contain
substantial measurement errors that bias the coefficient estimates. In either case, subtracting foreign returns other
than Canadian returns from U.S. raw returns potentially understates the abnormal returns of the U.S. 14 I also modify model (1a) by adding contemporaneous implied volatility computed from prices of options on
FTSE 100 or S&P/TSX 60 to control for other information captured by changes in the implied volatility. The
inferences remain intact. Furthermore, I add 30 variables capturing U.S. macroeconomic news (see Zhang, 2005
for the list of the variables) and the implied volatility of options on S&P 500. The estimated cumulative abnormal
return around key SOX events is also negative and statistically significant (�3:69% and �6:71% based on value- and equal-weighted returns, respectively).
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 95
the U.S., and have the option of listing on U.S. exchanges to seek additional financing. If SOX retards the economic growth of the U.S., these firms’ U.S. business would be negatively affected. Also, the costs of SOX would reduce the value of the option to list in the U.S. It is an empirical question whether the benefits of gaining a competitive advantage over SOX-complying firms or the losses resulting from the interdependence are greater for non-complying firms.
Second, securities regulation in other countries is likely influenced by that in the U.S. Since the passage of SOX, a number of countries (e.g., the U.K., see Mayer, Brown, Rowe, & Maw LLP, 2003) have adopted rules following the spirit of SOX. If investors in these countries expect the rulemaking progress in the U.S. to trigger costly regulation in their home country, stock prices in foreign markets are likely to decline as SOX developed in the U.S. and the investigated return differences would underestimate the costs of SOX.
Finally, non-U.S.-traded foreign firms also face news unique to their home country or region. Significant local economic news in foreign markets around the SOX legislative events would add noise to the estimation of the impact of SOX.
4.1.3. Returns of U.S.-exchange-listed foreign firms
If SOX is costly, foreign firms complying with SOX would be adversely affected and experience negative returns in the SOX event period. Berger et al. (2005) examine the incremental impact of SOX on U.S.-listed foreign firms relative to U.S. firms. They find that foreign private issuers experience more negative returns than the U.S. firms around certain SOX events. Their results are consistent with the incremental bonding benefits provided by SOX being exceeded by the costs. Litvak (2005) matches U.S.-listed foreign firms with their non-U.S.-traded counterparts by country, industry, and size to examine the abnormal returns of U.S.-listed foreign firms in their home country around the SOX events. She finds that U.S.-listed foreign firms experienced negative abnormal returns around events that increased the likelihood of SOX passage.
I examine returns of U.S.-listed foreign firms in the U.S. relative to returns of non- U.S.-traded firms in the SOX rulemaking period. Consistent with Berger et al. (2005), I find that the value-weighted cumulative raw return of U.S.-exchange-listed foreign firms in the U.S. amounts to �18:46% around the four key events, lower than the cumulative raw return of the U.S. market (�15:35%) and that of foreign markets. I then compute normal returns of U.S.-exchange-listed foreign firms in the U.S., applying a method similar to that used to estimate normal returns of U.S. firms. Specifically, I use data in 2001 and regress each foreign private issuer’s daily returns on CAN_Rett, EU_Rett and EU_Rettþ1, or AS_Rett and AS_Rettþ1, depending on which region it belongs to. Value-weighted market returns are used since U.S.-listed foreign firms are mostly large firms. Abnormal returns in 2002 are computed as prediction errors. Foreign private issuers on average witnessed a negative abnormal return of �9:71% around the four key SOX events and this return is statistically significant (untabulated). This analysis compliments Litvak’s (2005) finding that U.S.-listed foreign firms underperform non-U.S.-traded foreign firms in their home country market around certain SOX events, providing additional support for SOX imposing net costs on complying firms.
4.1.4. Confounding events and intraday returns
One of the fundamental limitations of event studies is that the measured abnormal performance could also capture other information released at the same calendar date. This
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11596
problem particularly stands out in this paper’s setting. The market returns likely incorporate investors’ reactions to other legislative activities, accounting scandals, and economic news. To evaluate the extent to which the abnormal returns documented in Table 2 is affected
by confounding events, I search for the above-mentioned market-wide news via ProQuest from July 8 to July 31, 2002 for WSJ and WP news reports with their subjects including ‘‘federal legislation’’ or ‘‘federal regulation.’’ To capture news about accounting scandals, I search the WSJ for news reports with their subjects including ‘‘scandals’’ or ‘‘securities fraud.’’ Zhang (2005) discusses the potential confounding events in this period and their impact in detail. The discussion here focuses on the events that might cause the significant market returns in the SOX event period. Around event 14, major contemporaneous events include the WorldCom hearing (7/8),
the debate on the Homeland Security bill in House committees (7/10, 7/11), and the rulemaking related to tax shelters (7/9, 7/11). The House Financial Services Committee held a hearing of the WorldCom scandal starting at 1 p.m. July 8. Both the ex-CEO and the ex-CFO of WorldCom invoked the Fifth Amendment and refused to answer questions. While it is unclear that new information about the WorldCom scandal was revealed through testimonies, the refusal of WorldCom executives to testify could signal that they were hiding some bad news. However, former Enron CFO and CEO also invoked the Fifth Amendment at Congressional hearings on February 8 and 12, 2002, respectively. Neither the one day nor the three-day window around each day was associated with significant market reactions. The rulemaking progress regarding the Homeland Security bill and tax shelters around event 14 was largely overridden around July 18 (Calmes, 2002; Johnston, 2002). Thus, the impact of these bills on the cumulative return of the period was likely canceled out by the subsequent reversals in rulemaking. To provide further support for the link between the negative return around event 14 and
SOX rulemaking news, I examine the immediate market reaction to the President’s speech on July 9. The timing of the speech is obtained from the White House website. The intraday results are reported in Table 3.
15 Stock prices began to decline in the middle of the
President’s speech. The market return in the last 10 min of the President’s speech accounts for 18% of the negative return of the day. As the event window is significantly narrowed, the documented return is less likely to reflect the effect of other contemporaneous news. Around event 16, WorldCom announced bankruptcy filings on July 21 (Sunday). If the
negative return of July 22 was driven by this filing, I expect a significant negative return at
15 The method to compute and test intraday returns is summarized as follows. First, transaction data is obtained
from NYSE TAQ database. If a security trades in consecutive intervals, t �1 and t, the return during interval t is
estimated as Rit ¼ðPit � Pit�1Þ=Pit�1, where Pit is the last transaction price in interval t (Barclay and Litzenberger, 1988). The previous-close-to-open return is calculated using the previous close price, the last transaction price
within 10 min before the close of the market, and the open price, the first transaction price within 10 min
after the open of the market. The value-weighted market return is calculated as: Rt ¼ PN
i¼1
Pit�1 � Shroutit�1 � Rit= PN
i¼1Pit�1 � Shroutit�1, where Shroutit�1 denotes the number of shares outstanding of
firm i at day t �1. If a security does not trade in consecutive intervals, the observation is excluded. Second, as
intraday returns are not normally distributed (Busse and Green, 2002), I use non-parametric bootstrapping tests
to determine the statistical significance of intraday returns. Specifically, to test whether Rt differs from zero, I
employ the following algorithm: (1) returns of the same time interval of non-event days in 2001 and 2002 are
computed (sample A); (2) a sample (sample B) of 1000 returns is drawn with replacement from sample A; (3) one-
tailed p-values are computed as: p ¼ the number of observations in sample B with values greater (lower) than
Rt=1000, if Rt4 ¼ 0 (if Rto0).
ARTICLE IN PRESS
Table 3
Other contemporaneous events between July 8 and July 31, 2002 and intraday returns
Date Time of
event
Event Return
interval
Return p-Value Percentage
7/9/2002 11:20 Bush delivered a speech 11:20–11:29 0.0015 0.0970 �7.22
a.m.–11:47 regarding accounting 11:30–11:39 �0.0020 0.0370 9.81
a.m. reforms 11:40–11:49 �0.0038 0.0140 18.28
11:50–11:59 �0.0007 0.2480 3.16
Previous
close - close
�0.0207
7/22/2002 July 21 WorldCom announced
intention to file for
Previous
close - open
�0.0032 0.2490 10.53
Chapter 11 on July 21 Previous
close - 9:39
�0.0027 0.2950 8.88
Previous
close - close
�0.0304
7/24/2002 July 24, 6
a.m.
Adelphia executives were
arrested around 6 a.m.
Previous
close - open
�0.0190 0.0080 �32.82
Previous
close - 9:39
�0.0268 0.0060 �46.29
Previous
close - close
0.0579
7/29/2002 July 26/27 Trade bill passed in the
House on July 27
Previous
close - open
0.0207 0.0010 39.01
The Homeland Security
bill passed in the House on
July 26
Previous
close - 9:39
0.0234 0.0010 44.10
11:00
a.m.–11:31
a.m.
Bush delivered a speech
regarding the welfare
reform, in which he
11:00–11:09 �0.0005 0.3580 �1.01
explicitly indicated that he 11:10–11:19 �0.0009 0.2340 �1.64
would sign the accounting 11:20–11:29 �0.0010 0.1780 �1.91
reform bill into law 11:30–11:39 0.0026 0.0380 4.94
Previous
close - close
0.0531
This table reports the intraday returns around certain contemporaneous news announcements between July 8 and
July 31, 2002. The method to compute and test the intraday returns is discussed in footnote 14. The last column of
the table, percentage, reports the market return in the interval as a percentage of the previous-close-to-close return
of the day.
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 97
the opening of the market on that day. The previous-close-to-open return of July 22 was �0:32%, but it was not statistically significant. Moreover, the negative return reversed within 15 min after the opening of the market (untabulated). The evidence shows that the bankruptcy of WorldCom is unlikely the key driver of the large negative return of the day.
The major contemporaneous event around event 17 is the arrest of Adelphia executives around 6 a.m. of July 24 (Markon and Frank, 2002). The conference committee reached an agreement in the morning and the content of the conference report was revealed during the
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–11598
day, though the exact timing is not available. 16
If positive stock returns on July 24 mainly reflect the effect of the arrest of Adelphia executives, I expect a significant positive return at the opening of the market. However, the opening return of July 24 was negative, inconsistent with this hypothesis. The negative opening return can be explained by the arrest signaling a change of attitude of the government and an increase in the likelihood that a tough rule would pass. I also examine quarterly earnings news in July 2002, as earnings releases could also
provide information regarding business conditions and move stock prices. Quarterly earnings and earnings announcement dates are obtained from Compustat-CRSP merged quarterly file. Analyst forecast data and actual earnings used to compute forecast errors are from I/B/E/S. First, three measures are used alternatively to classify an earnings announcement as
good news or bad news: earnings, unexpected earnings, or analyst forecast error. 17
Non- negative values are considered good news. Second, I compute the net number of earnings announcements with non-negative earnings (Net_NI), unexpected earnings (Net_UE), or forecast error (Net_FE) for each calendar day, deflated by the number of earnings announcements of the day.
18 If bad earnings news drives the negative return in July 2002,
I expect Net_NI, Net_UE, and Net_FE to be negative in this period and more negative than the rest of the year. Third, as the impact of earnings news on the market index conceivably varies with firm size, I also examine the value-weighted net number of non- negative earnings releases (VW_Net_NI, VW_Net_UE, and VW_Net_FE). The weight is calculated as the market capitalization of the announcing firm over the total market value as of the day prior to the earnings release. The plots of the time-series distribution of Net_NI, Net_UE, Net_FE, VW_Net_NI,
VW_Net_UE, and VW_Net_FE can be found in Zhang (2005). There is no evidence that bad news about earnings clustered in July 2002. On the contrary, the most significant SOX rulemaking period witnessed more good earnings news relative to the rest of the year. Binomial tests reject the hypothesis that the percentage of non-negative news is no greater than 50%, except for July 8, 9, and 26, where it cannot be rejected that good and bad earnings news each account for 50% of the earnings announcements. In summary, although there were other rulemaking activities, accounting scandals, and
earnings news reported from July 8 to July 26, 2002, it is not obvious that they convey significant bad news to the market. Furthermore, the economic recession in 2002 and the news announcements regarding accounting scandals are not totally independent of SOX. The scandals could have caused anti-business rulemaking, the immediate consequence of
16 I could search Dow Jones Newswire and find the stamped time of the first news report to test the intraday
market reactions. However, I find the stamped time of reports is inaccurate. For example, the President gave a
speech on July 29 at 11:00 a.m. However, I find a report stamped at 10:07 a.m., discussing the speech as if it had
been given. It is hard to use the inaccurate time to distinguish the impact of confounding events that occur on the
same day. 17 Unexpected earnings are calculated based on the seasonal random walk model. Analyst forecast error is
computed as the difference between actual earnings and the consensus forecast, where the consensus forecast is the
average of analyst forecasts issued within 60 days prior to the earnings release. 18
Net_NI (Net_UE, Net_FEÞ ¼ the number of announcements with non-negative earnings (unexpected
earnings, forecast error)�the number of announcements with negative earnings (unexpected earnings, forecast
error).
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 99
which was SOX. It is likely that future adverse federal and state regulations were also forecasted by the market which further weakened the economy and deepened the recession.
I also search for significant news announcements around event 2. On the same day as the Treasury Secretary’s talk (February 2, 2002, Saturday), the Powers report was released, suggesting that Andersen did not function as an effective check on the disclosure reported by Enron (Chaney and Philipich, 2002). There would have been a negative market reaction to the report if it changed investors’ expectation of the reliability of financial reporting and audit quality. I cannot distinguish which event gave rise to the negative return on February 4, 2002.
4.2. Cross-sectional variation in abnormal returns around the rule-making events
4.2.1. Research design, sample selection, and descriptive statistics
The regression to test the cross-sectional implications of SOX provisions is specified as follows:
CARi ¼ X
j
a0j Ind ij þ a1 Nonauditi þ a2 Gindexi þ a3 Busi_linesi
þ a4 Foreigni þ a5 Incentivei þ a6 Sizei þ a7 Accri þ a8 Shumwayi
þ a9 Acqi þ a10 MTBi þ a11 ROAi þ a12 Pre_Reti þ ei. ð2Þ
CARi is the cumulative abnormal return of firm i. Abnormal returns are estimated using two methods. I first estimate the market model for each firm using their 2001 daily return data and the U.S. market data. Lo (2003) points out that the market model abnormal returns likely underestimate the impact of regulatory events on individual firms, as the whole market is affected by these events. As an alternative, I also estimate the market model replacing U.S. market returns with five variables capturing returns of non-U.S.- traded foreign firms, namely, CAN_Rett, EU_Rett, EU_Rettþ1, AS_Rett, and AS_Rettþ1. Cumulative abnormal returns are then computed as the sum of the prediction errors.
19
H2 examines the impact of the restriction of non-audit services. I collect the non-audit fee data from firms’ fiscal 2001 proxy statements via Lexis/Nexis. The variable Nonaudit is constructed to capture a firm’s purchase of non-audit services from the auditor. It equals a firm’s non-audit fees minus fees paid to the auditor for tax-related services, deflated by the market value of equity of the firm at the end of 2001. Fees paid to the auditor for tax- related services are excluded because they are not subject to the restriction of non-audit services and they constitute a substantial portion of non-audit fees paid to the auditor.
20
H2 predicts a1 to be negative. Bebchuk et al.’s (2005) governance index is employed to capture shareholder protection.
Bebchuk et al.’s index improves on Gompers et al. (2001), who construct a governance
19 Instead of computing CARi as the simple sum of ARit, I also compute the compounded CARi as
expð P
t lnð1þ ARitÞÞ�1. The inferences of the cross-sectional tests remain unchanged if the compounded CARi is
used as the dependent variable. Also, the inferences are not sensitive to using CARi over all SOX events as the
dependent variable. 20 Neither Sen. Sarbanes’ bill nor the final act banned tax-related services. About 45% of the firms in my sample
separately disclosed the amount of fees for tax-related services. For firms that did not disclose the detailed fee
items, I use the total non-audit fees in the tests. Investors could not determine whether the services would be
affected by SOX if detailed fee items were not disclosed. As a result, it is reasonable to use the total fees for these
firms.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115100
index from corporate charters to capture firms’ differential shareholder rights. Bebchuk et al. (2005) argue that not all the provisions included in Gompers et al.’s index are related to governance. Bebchuk et al. consider the following six provisions to be most relevant to shareholder protection: staggered board, limits to shareholder bylaw amendments, supermajority requirements for mergers and charter amendments, poison pills, and golden parachutes.
21 Firms get one point for the existence of each provision. The higher the value
of the index, presumably the weaker are shareholder rights. The analysis in Section 3.2.2 predicts that firms with weak governance would obtain
more benefits and incur greater costs as a result of the governance provisions. Ideally, the optimal level of shareholder control for each firm should be used as the benchmark to define weak governance. However, the literature has yet to produce a well specified expectation model of the optimal level of corporate governance. It is also not obvious what is the optimum considered by regulators. As a result, the industry median level of shareholder protection is used as a parsimonious proxy for the optimum. Bindex thus equals Bebchuk et al.’s index less the median of the index in each two-digit SIC industry.
22
H3 predicts that a2 is negative. The complexity of business is measured by the number of four-digit SIC industries of
each firm (Busi_lines) and a dummy variable to capture whether a firm is involved in transactions with foreign parties or has foreign operations (Foreign). Foreign equals one if foreign currency adjustment is non-zero, and zero otherwise. H4 predicts a3 and a4 to be negative. H6 predicts that firms that compensate CEOs with excessive incentive-based pay incur
larger costs as a result of SOX. Again, the industry median level of incentive-based compensation is used to proxy for the optimum. The variable of interest is thus the ratio of incentive pay, including option grants, restricted stocks, bonus, and long-term incentive payouts, to the total compensation of the CEO, less the industry median ratio (Incentive), measured at the end of 2001. Compensation data is obtained from ExecuComp. The variable Incentive is likely influenced by firm performance. CEOs in poorly performing firms probably lost their bonus and thus the percentage of incentive pay in their total pay would be low. I control for firm performance by adding return on assets (ROA) and the market-adjusted returns (Pre_Ret) both for 2001. SOX increases the litigation costs of firms and executives and thus firms’ abnormal
returns around the legislative events likely decrease with their litigation risks. The key variables in the cross-sectional regression, such as Nonaudit, Bindex, and Incentive can be correlated with litigation risks. For example, distressed firms are likely to purchase more consulting services to improve business operations. They are also more likely to become the targets of lawsuits. To mitigate the omitted correlated variables problem, I include proxies for litigation
risks in regression (2). Lys and Watts (1994) find that the probability of lawsuit is a function of the likelihood of bankruptcy, the likelihood of being acquired, firms’ accruals, and whether a qualified audit opinion is issued. Since only one firm in my sample received a
21 The governance data are obtained from the IRRC database and are available for every other year. Since the
governance structure is likely sticky, I use the data of 2002, assuming that the governance index of 2002 does not
differ significantly from that of 2001. The tenor of the results does not change if I use the governance data of 2000. 22 If there was a massive market failure prior to SOX, the optimal level of governance can exceed the existing
industry norms. Therefore, Bindex is defined as a continuous variable rather than being truncated at zero.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 101
qualified audit opinion in 2001, I use proxies for the other three factors to control for litigation risks. The variable Accr represents abnormal accruals, computed as residuals from the modified Jones model estimated for each two-digit SIC industry. The likelihood of bankruptcy is estimated based on Shumway (2001) (Table 4b). The variable Acq equals one if a firm is subsequently acquired and delisted in 2003. These variables are predicted to have a negative correlation with SOX abnormal returns.
Size and MTB are also included in the regression. Size is measured as the logarithm of the market value of equity of each firm at the end of 2001. Market-to-book ratio (MTB) is defined as the market value of equity over the book value of equity at the end of 2001. A firm’s purchase of non-audit services, compensation structure, and corporate governance are affected by firm size and investment opportunities. Firms with more growth options could lose more investment opportunities as a result of tight controls and higher litigation risks, which predicts a negative relation between CAR and growth options. There is no clear prediction as to the relation between CAR and Size. Although small firms may incur disproportionately greater direct compliance costs, large firms could be subject to greater litigation and political costs as an indirect result of SOX or future legislation. The Ind js are industry dummies and are defined as in Frankel et al. (2002) and Ashbaugh et al. (2003).
23
To be included in the sample for the cross-sectional test, I require that firms have return data from CRSP, financial data from Compustat, compensation data from Execucomp, governance data from IRRC, and non-audit fee data from proxy statements. Execucomp and the IRRC database largely cover S&P 1500 firms. The final sample includes 1224 observations. The extreme 1% of each continuous variable is winsorized.
Table 4 Panel A shows the sample selection procedures. Although the number of firms only accounts for about 20% of the CRSP population, the market value of these firms constitutes over 90% of the total market capitalization of NYSE, AMEX, and Nasdaq. Small firms are under-represented in this sample. As a result, the test of the differential impact of SOX on small and large firms may lack power. Unreported results show that the pattern and magnitude of average raw returns of the sample around the SOX legislative events are comparable to that of the whole market presented in Table 2. Table 4 Panel B presents firm characteristic variables. The average firm size in the sample is significantly larger than an average firm in the CRSP universe. Panel C reports the Pearson correlation between variables. Nonaudit and Shumway are positively correlated, consistent with distressed firms using more consulting services. As expected, Incentive is positively correlated with ROA. ROA is highly correlated with Shumway (�0:7691), suggesting a multicollinearity problem in the regression. Excluding ROA or Shumway from the regression does not affect the main results.
4.2.2. Regression results
The estimation results of regression (2) are reported in Table 5. As the event dates are clustered, cross-sectional correlation of returns may result in biased standard errors and potentially incorrect inferences (Sefcik and Thompson, 1986; Bernard, 1987). Moreover,
23 Industry membership is determined by SIC code as follows: agriculture (0100–0999), mining and construction
(1000–1999, excluding 1300–1399), food (2000–2111), textiles and printing (2200–2799), chemicals (2800–2824,
2840–2899), pharmaceuticals (2830–2836), extractive (1300–1399, 2900–2999), durable manufacturers (3000–3999,
excluding 3570–3579 and 3670–3679), transportation (4000–4899), utilities (4900–4999), retail (5000–5999),
services (7000–8999, excluding 7370–7379), and computers (3570–3579, 3670–3679, 7370–7379).
ARTICLE IN PRESS
Table 4
Descriptive statistics of the sample used in the cross-sectional tests
Panel: A Sample selection procedures
Sample selection procedures
Number of firms with return data from CRSP in 2002 7033
With financial data from Compustat 4844
With compensation data from Execucomp in 2001 1374
With governance index from IRRC in 2002 1241
With non-audit fee data for 2001 1224
Panel: B Descriptive statistics
Variable Number of
observations
Mean Median Std p25 p75
Nonaudit (%) 1224 0.0946 0.0372 0.1754 0.0114 0.0957
Bindex 1224 0.0976 0 1.0832 �1 1
Busi_lines 1224 3.0547 2 1.8142 2 4
Foreign 1224 0.2974 0 0.4573 0 1
Incentive 1224 �0.0776 0 0.2784 �0.2069 0.1196
Accr 1224 0.0497 0.0385 0.1799 �0.0217 0.1213
Shumway 1224 0.0303 0.0057 0.1295 0.0020 0.0128
Acq 1224 0.0229 0.0000 0.1496 0 0
MTB 1224 3.1601 2.2243 3.0214 1.4275 3.8359
Size 1224 7.2886 7.1345 1.5854 6.2391 8.3002
Pre_Ret 1224 �0.2127 �0.3050 0.6797 �0.5283 �0.0241
ROA (%) 1224 0.6819 3.4706 16.1072 �1.0523 7.2661
Panel: C Pearson correlation
Nonaudit Bindex Busi_lines Foreign Incentive Accr Shumway Acq MTB Size Pre_Ret
Nonaudit
Bindex 0.0212
Busi_Lines 0.0527 0.0093
Foreign 0.0244 0.0156 �0.0019
Incentive �0.1280 0.0233 0.0902 0.0368
Accr �0.0406 0.0438 0.0467 0.0412 �0.0183
Shumway 0.1601 �0.0485 �0.0807 0.0028 �0.0797 �0.2915
Acq 0.0339 0.0165 �0.0137 �0.0278 �0.0429 �0.0168 �0.0107
MTB �0.2121 �0.0057 �0.0973 0.0464 0.1554 �0.0075 �0.0260 �0.0383
Size �0.3407 �0.0433 0.2196 0.0537 0.3718 0.0097 �0.1433 �0.0500 0.4070
Pre_Ret �0.0537 �0.0242 �0.0390 �0.0663 0.0147 0.0092 �0.1505 0.0206 0.0179 �0.0890
ROA �0.1910 0.0014 0.0950 �0.0322 0.1162 0.3529 �0.7691 �0.0277 0.1804 0.2522 0.2049
Panel A summarizes the sample selection procedure for the cross-sectional tests. Firms are required to have daily return data from CRSP, financial data from Compustat to compute all variables except for the compensation, non-audit fee, and governance variables, compensation data from Execucomp, governance index from IRRC and non-audit fee data manually collected from proxy statements. Panel B reports the descriptive statistics of explanatory variables in the cross-sectional tests. Nonaudit is calculated as non-audit fees minus fees paid for tax-related services, deflated by the market value of equity of each firm at the end of 2001. Bindex is the governance index of Bebchuk et al. (2005) as of 2002 less the median index value in the same two-digit SIC industry. Busi_lines equals the number of four-digit SIC industries of each firm. Foreign equals one if foreign currency adjustment is non-zero, zero otherwise. Incentive equals the ratio of incentive pay to total compensation of the CEO less the industry median ratio. Accr measures abnormal accruals, computed as residuals from the modified Jones model estimated using 2001 data for each two-digit SIC industry. Shumway captures the likelihood of bankruptcy, estimated based on Shumway (2001, Table 4b). Acq equals one if a firm is subsequently acquired and delisted in 2003. MTB denotes the market-to-book ratio, evaluated at the end of 2001. Size equals the logarithm of market value of equity at the end of 2001. Pre_Ret is cumulative market-adjusted returns in 2001. ROA is return on assets of 2001. Panel C reports the Pearson correlation between variables used in the cross-sectional tests. Correlation coefficients that are significantly different from zero at less than 5% level are presented in bold.
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115102
ARTICLE IN PRESS
Table 5
Cross-sectional test of market reactions to SOX
Panel: A Cross-sectional test of the market reaction to SOX—cumulative returns
Prediction Dependent variable
CAR (US) CAR (Foreign)
Estimate p-Value Estimate p-Value
OLS Bootstrap OLS Bootstrap
Nonaudit – �0.0820 0.0008 0.0440 �0.0652 0.0078 0.0680
Gindex – �0.0066 0.0728 0.0020 �0.0069 0.0641 0.0020
Busi_lines – �0.0046 0.0585 0.0180 0.0015 0.5368 0.2640
Foreign – �0.0223 0.0150 0.0000 �0.0284 0.0021 0.0000
Incentive – �0.0079 0.6057 0.4600 �0.0372 0.0159 0.0000
Accr – �0.0701 0.0036 0.0000 �0.0546 0.0247 0.0120
Shumway – �0.0136 0.7822 0.2310 0.0002 0.9961 0.6960
Acq – 0.0434 0.1007 0.0900 0.0319 0.2324 0.2300
MTB – 0.0031 0.0464 0.0700 0.0019 0.2156 0.1320
Size �0.0030 0.3608 0.4000 �0.0051 0.1281 0.3540
ROA �0.0279 o.0001 0.2620 �0.0106 0.0842 0.8900 Pre_Ret �0.0011 0.0109 0.0250 �0.0001 0.7791 0.4820
Adj. R2 17.38% 7.23%
Panel: B Cross-sectional test of the market reaction to SOX—individual events
Pre-
diction
Event 2 Event 14 Event 16 Event 17
Estimate p-Value Estimate p-Value Estimate p-Value Estimate p-Value
OLS Boot-
strap
OLS Boot-
strap
OLS Boot-
strap
OLS Boot-
strap
Nonaudit – 0.0051 0.3913 0.2940 0.0027 0.8241 0.2300 �0.0561 0.0003 0.0260 0.0337 0.0173 0.0580
Bindex – �0.0008 0.3559 0.2060 �0.0005 0.8002 0.4600 �0.0058 0.0142 0.0000 �0.0005 0.8302 0.3050
Busi_lines – 0.0004 0.5468 0.2440 �0.0040 0.0009 0.0060 �0.0041 0.0088 0.0100 �0.0031 0.0293 0.0100
Foreign – �0.0021 0.3462 0.2010 �0.0056 0.2148 0.0980 �0.0094 0.1062 0.0000 0.0051 0.3360 0.1160
Incentive – �0.0001 0.9769 0.5460 0.0092 0.2203 0.0960 0.0060 0.5351 0.1110 0.0230 0.0096 0.0000
Accr – �0.0058 0.3287 0.4180 �0.0423 0.0004 0.0000 �0.0476 0.0020 0.0000 �0.0255 0.0685 0.0340
Shumway – �0.0238 0.0478 0.2000 �0.0323 0.1811 0.2040 �0.0139 0.6568 0.3460 �0.0565 0.0480 0.0620
Acq – 0.0169 0.0091 0.0590 �0.0077 0.5561 0.0980 0.0168 0.3194 0.1780 �0.0173 0.2606 0.1210
MTB – 0.0001 0.8403 0.5760 0.0003 0.6733 0.4720 0.0013 0.1916 0.1020 �0.0014 0.1212 0.0410
Size �0.0004 0.5900 0.4660 0.0013 0.4210 0.2920 �0.0032 0.1331 0.1980 0.0007 0.7078 0.4660
Pre_Ret �0.0020 0.1745 0.6680 �0.0113 0.0002 0.3720 �0.0206 o.0001 0.0000 �0.0060 0.0910 0.0120 ROA �0.0001 0.2619 0.4340 �0.0013 o.0001 0.0000 �0.0008 0.0045 0.0180 �0.0011 o.0001 0.0000
Adj. R2 2.21% 26.13% 32.37% 24.64%
This table presents the estimation results of regression (2) to examine the cross-sectional variation in market
reaction to the key events leading to SOX.
CARi ¼ X
j
a0j Ind ij þ a1Nonauditi þ a2 Gindexi þ a3 Busi_linesi þ a4 Foreigni þ a5 Incentivei
þ a6 Sizei þ a7 Accri þ a8 Shumwayi þ a9 Acqi þ a10 MTBi þ a11 ROAi þ a12 Pre_Reti þ ei ð2Þ
Nonaudit is calculated as non-audit fees minus fees paid for tax-related services, deflated by the market value of
equity of each firm at the end of 2001. Bindex is the governance index of Bebchuk et al. (2005) as of 2002 less the
median index value in the same two-digit SIC industry. Busi_lines equals the number of four-digit SIC industries
of each firm. Foreign equals one if foreign currency adjustment is non-zero, zero otherwise. Incentive equals the
ratio of incentive pay to total compensation of the CEO less the industry median ratio. Accr measures abnormal
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 103
ARTICLE IN PRESS
accruals, computed as residuals from the modified Jones model estimated using 2001 data for each two-digit SIC
industry. Shumway captures the likelihood of bankruptcy, estimated based on Shumway (2001, Table 4b). Acq
equals one if a firm is subsequently acquired and delisted in 2003. MTB denotes the market-to-book ratio,
evaluated at the end of 2001. Size equals the logarithm of market value of equity at the end of 2001. Pre_Ret is
cumulative market-adjusted returns in 2001. There are 1224 observations included in each regression. ROA is
return on assets of 2001.
In Panel A, CARi is the cumulative abnormal return of firm i over the four key SOX events. Abnormal returns are
estimated using two methods. (i) The market model for each firm using their 2001 daily return data and U.S.
market data (‘‘US’’). (ii) The market model is modified by replacing U.S. market returns with returns of non-U.S.-
traded foreign firms, namely, returns of Canada, Europe, and Asia (including Australia and New Zealand) on the
same calendar date, and European and Asian returns one day after that day (‘‘Foreign’’).
Panel B presents the estimation results of regression (2) with CAR*S as the dependent variable, where CAR
denotes abnormal returns around the individual events that are associated with revisions in expectations (events 2,
14, 16, and 17), and S is an indicator variable which equals one for events 2, 14, and 16, and minus one for event
17. The market model based on U.S. market returns is used to estimate CAR. The coefficients on industry
dummies ðInd jÞ are not reported. For each regression, both the asymptotic two-tailed OLS p-values and the
bootstrapped one-tailed p-values are reported.
Table 5 (footnote continued)
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115104
the association between abnormal returns and firm characteristics could be explained by other documented regularities. For example, a negative correlation between abnormal returns and Bindex could be driven by the regularities found by Gompers et al. (2003). Therefore, I compute both bootstrapped p-values and the asymptotic p-values of the estimated coefficients.
24
Table 5 Panel A reports the estimation results of regression (2) with U.S. firms’ cumulative abnormal returns around the four key SOX events as the dependent variable.
25
As explained in the previous subsection, cumulative abnormal returns are computed from the market model and the modified market model using foreign returns as the regressors. Both the OLS two-tailed p-values and bootstrapped one-tailed p-values show similar inferences. The coefficient on Nonaudit is negative and statistically significant in both regressions,
rejecting the hypothesis that the restriction imposes no net costs on firms. Firms’ abnormal returns decrease with Bindex, rejecting the hypothesis that the requirements to strengthen shareholder rights impose no net costs on firms. Given that SOX is primarily characterized as an act to protect investors, the result casts doubt on the value of the rules, assuming Bindex at least partly captures the extent of shareholder rights. The coefficient on Busi_lines is negative, but not always statistically significant. There is some evidence that firms with more business lines incur greater costs. The coefficient on Foreign is significantly negative in both regressions, providing support for H4. The coefficient on Incentive is not always significant. Although Cohen et al. (2004b) document a significant decline in the
24 Based on Lo (2003)’s method, the bootstrapped p-values are calculated as follows: one-tailed p-values are the
percentage of 1000 repetitions that generate coefficients greater than the OLS coefficients in the table (less than the
OLS coefficient if it is negative). Each repetition uses sample firms’ abnormal returns from random non-event days
selected from 2002 or 2003. If the event days are consecutive, I select consecutive non-event days similarly. I then
add up each firm’s abnormal returns to obtain CAR. 25 Using cumulative abnormal returns around all SOX events as the dependent variable does not change the
inferences.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 105
ratio of incentive pay to salary after SOX, the return test in this paper does not find strong support for the provisions on incentive pay being costly.
Among the three litigation risk measures, only the coefficient on Accr is negative and statistically significant. Accr is often used as a proxy for earnings management. Cohen et al. (2004a) suggest that SOX may constrain earnings manipulation. Thus, the negative coefficient on Accr can also be interpreted as showing that firms managing earnings prior to SOX incur more net costs. However, it cannot be distinguished whether the litigation or the earnings management explanation is more important.
The coefficient on MTB is positive, largely inconsistent with the prediction that firms with more growth options incur greater costs. The coefficient on Size is insignificant. The coefficient on ROA is negative and significant in one specification. It could be that investors expect firms with higher accounting earnings to have manipulated accounting numbers or that investors expect them to face greater litigation risks.
Table 5 Panel B reports the estimation results of regression (2) with CAR�S as the dependent variable, where CAR is firms’ abnormal returns around the events that revised investors’ expectations, and S is an indicator variable that equals one for events 2, 14, and 16 (the events with negative market returns) and minus one for event 17 (the event with a positive return).
26 The sign of the coefficients in regression (2) depends on the cross-
sectional implications of SOX and the direction in which each event revised investors’ expectations. Event 17 was likely associated with a reduction in expected costs of SOX. Consequently, the sign of CAR for this event is reversed so that the estimated coefficients have the same signs as predicted by H2–H6. For brevity of presentation, only results of CAR estimated using the U.S. market model are presented. The other method based on foreign returns generates very similar results, except that the coefficient on Busi_lines is insignificant in the regression for event 17.
The event-by-event analysis does not always reject the null hypothesis that the examined SOX provisions are not costly. First, not surprisingly, little cross-sectional variation is found for event 2. The Treasury Secretary’s talk sent a signal to the market that new regulations were impending, but it was unclear then how laws would be written. Second, in the regression for event 14, only the coefficient on Busi_lines is significant with the expected sign. The coefficients on other variables of interest are insignificant, though those on Bindex and Foreign have the predicted sign. Third, in the last regression of event 17, again, only the coefficient on Busi_lines is significant with the expected sign. While the coefficient on Bindex has the predicted sign, the coefficients on Nonaudit and Incentive are positive and significant. The positive coefficient on Nonaudit is either inconsistent with H2 or reflecting that the market had expected the provisions on non-audit services to be less costly than the final restriction. This latter explanation is plausible as the media reported that accounting firms unexpectedly lost their influence in Congress (Weisman, 2002). The positive coefficient on Incentive is inconsistent with H6. Yet, in Panel A, the coefficient on Incentive is insignificant.
The lack of consistency in the event-by-event analysis can reflect the tests lacking power. In the regression for event 16, the event with market reaction of the largest magnitude, the coefficients on Nonaudit, Bindex, Busi_lines, and Foreign are all statistically significant with the expected sign. Also, except for Nonaudit, the coefficients on Bindex,
26 I also conduct the event-by-event cross-sectional test using the methodology of Sefcik and Thompson (1986).
The inferences remain intact.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115106
Busi_lines, and Foreign have the predicted sign in at least three out of the four event-by- event regressions. In summary, the cross-sectional analysis based on cumulative returns rejects the
hypothesis that three major provisions entail no net costs on firms, providing additional support for the hypothesis that the market initially expected SOX to be costly. However, the event-by-event analysis does not consistently support the hypotheses, potentially due to the tests lacking power. A few issues are worth noting in interpreting the cross-sectional findings. First, the tests
document associations between returns and firm characteristics. One cannot rule out the possibility that the coefficient on Nonaudit, for example, reflects the impact of other or all provisions rather than just the restriction of non-audit services. Second, certain variables that capture firm attributes are controversial. For example, Cremers and Nair (2005) argue that the governance index based on corporate charters is an anti-takeover protection index, representing external governance.
4.2.3. Market reactions to the announcement of postponing compliance with Section 404
H5 predicts that the postponement of Section 404 announced on May 27, 2003 is good news and therefore late adopters would experience higher abnormal returns than early adopters. The prediction is summarized in Table 6 Panel A. If a firm has a market capitalization of less than $75 million by the end of 2002, I classify it as a non-accelerated filer. Note that some of the non-accelerated filers could be misclassified as accelerated filers, if their total market value of equity is greater than $75 million but do not satisfy other criteria for an accelerated filer. I exclude foreign firms and banks, as the expected compliance costs of foreign firms and banks prior to the announcement of the final rule were likely different from those of the other firms.
27
I employ the following regression to examine market reactions to the announcement:
CARi ¼ X
j
a0j Ind ij þ a1 Non_Acci þ a2 Late_Noni þ a3 Late_Acci
þ a4 MTBi þ a5 Sizei þ ei, ð3Þ
where Ind ij are industry dummies as defined in regression (2), Late_Non equals one for non-accelerated filers that obtain two years of extension and zero otherwise, and Late_Acc equals one for accelerated filers that obtain a one year extension and zero otherwise. The variable Non_Acc equals one for non-accelerated filers and zero for accelerated filers. H5 predicts that a2 and a3 are positive if the deferment is good news. I do not have predictions regarding the difference in reaction between accelerated and non-accelerated filers. As accelerated and non-accelerated filers are treated differently under some other SEC rules, investors could have expected the SEC to treat them differently in this case. Therefore, the sign of a1 is indeterminate. The estimation results are reported in Table 6 Panel B. The dependent variable is
abnormal returns cumulated over days ð�1;1Þ, ð�3;1Þ, and ð�5;1Þ around the announcement day, respectively. The regression with CAR ð�1;1Þ as the dependent
27 For example, since large banks had been required to file an internal control report to the FDIC before SOX,
banks were lobbying for a waiver of Section 404. The final rule of the SEC allows large banks to file one report
both to the FDIC and to comply with Section 404. The inclusion of these firms does not affect the tenor of the
results.
ARTICLE IN PRESS
Table 6
Market reactions to the announcement of postponing compliance with Section 404
(A) Summary of predictions
All firms
Accelerated f ilers
2709 firms
Median Mkt Cap: $509 million
Nonaccelerated f ilers
1536 firms
Median Mkt Cap: $22 million
Early
Fiscal year ending
between June 15
and September 14
(unaffected)
A firm is
classified into this
group if its fiscal
year ends in June,
July or August
Late
Fiscal year ending
between September
15 and June 14 (one
year of extension)
A firm is classified
into this group if its
fiscal year ends in
months other than
June, July or August
Early
Fiscal year ending
between April 15
and September 14
(one year of
extension)
A firm is classified
into this group if its
fiscal year ends
between April and
August (inclusive)
Late
Fiscal year ending
between September
15 and April 14 (two
years of extension)
A firm is classified
into this group if its
fiscal year ends
between January and
March, or between
September and
December
Predictions:
H5: Early_AccoLate_Acc H5: Early_NonoLate_Non
Panel: B Regression results
Prediction CAR ð�1;1Þ CAR ð�3;1Þ CAR ð�5;1Þ
Estimate p-Value Estimate p-Value Estimate p-Value
OLS Bootstrap OLS Bootstrap OLS Bootstrap
Non_Acc 0.0060 0.1083 0.4010 0.0057 0.2707 0.4650 0.0092 0.1355 0.4090
Late_Non + 0.0028 0.3872 0.2860 0.0087 0.0504 0.0870 0.0126 0.0190 0.0530
Late_Acc + 0.0023 0.5094 0.2220 �0.0028 0.5683 0.2720 0.0004 0.9422 0.4500
Size �0.0009 0.1635 0.3860 0.0001 0.9503 0.4790 0.0005 0.6033 0.2930
MTB 0.0001 0.7510 0.1170 0.0002 0.6785 0.0340 �0.0002 0.6370 0.6150
Adj. R2 1.08% 2.79% 2.85%
This table reports the test of market reactions to the announcement of postponing the compliance dates of Section
404 on May 27, 2003.
Panel A summarizes the predictions of H5, which hypothesizes that the postponement is beneficial as it reduces
the compliance costs, especially for firms that obtained a longer extension period.
The proposed compliance date was fiscal year ending on or after September 15, 2003. In the final rule, accelerated
filers need to comply with Section 404 for fiscal year ending on or after June 15, 2004, and non-accelerated filers
were required to comply for fiscal year ending on or after April 15, 2005. In the empirical tests, firms with a
market capitalization lower than $75 million by the end of 2002 are classified as non-accelerated filers (see
footnote 4 for the definition of an accelerated filer). Foreign firms and banks are excluded.
Panel B reports the estimation results of the regression that examines the cross-sectional variation in market
reactions to the announcement of postponing compliance with Section 404 on May 27, 2003. The announcement
day is set to be event day zero. Foreign firms and banks are excluded
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 107
ARTICLE IN PRESS
CARi ¼ X
j
a0j Ind ij þ a1 Non_Acci þ a2 Late_Noni þ a3 Late_Acci þ a4 MTBi þ a5 Sizei þ ei.
There are 2709 accelerated filers and 1536 non-accelerated filers included in estimating the regression. Late_Acc
equals one if a firm is an accelerated filer and obtained one year of extension, zero otherwise. Late_Non equals one
if a firm is a non-accelerated filer and obtained two years of extension, zero otherwise. Non_Acc equals one if a
firm is a non-accelerated filer, zero otherwise.
Table 6 (footnote continued)
I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115108
variable does not reject the null hypothesis that the deferment had no significant impact. However, the regressions with CAR ð�3;1Þ and CAR ð�5;1Þ as the dependent variable show that non-accelerated filers that obtained two years of extension realized significantly higher abnormal returns than non-accelerated filers that obtained a one year extension, consistent with H6. The coefficients on Size and MTB are largely insignificant. Unreported results based on CAR ð�4;1Þ and CAR ð�6;1Þ show similar inferences. The findings suggest that there could be news leakage prior to the release of the final rule. It is not surprising that the difference between the early and late adopters in the accelerated filers group is insignificant. Given the relatively large fixed costs of Section 404 compliance, the cost savings as a percentage of the market value of large firms are likely smaller than those of small firms. Table 6 Panel B shows that the cost savings of delaying compliance for one more year
are about 1.26% (from the regression with CAR (�5, 1) as the dependent variable) of the market value of an average firm in the non-accelerated filers group. The average market capitalization of a non-accelerated filer is $27 million, which suggests that the estimated cost savings are about $0.34 million. The average annual revenue of the non-accelerated filers is $91 million. The FEI January 2004 survey (Financial Executive International, 2005) indicates that the direct initial compliance costs for a firm with annual revenue between $25 and $99 million are about $0.74 million. Compared with FEI’s cost estimates, the magnitude of the cost savings appears reasonable. In summary, the results in Table 6 largely support H5. The compliance costs of Section
404 are particularly significant for small firms and delaying compliance appears beneficial for them.
4.3. Subsequent events
Li et al. (2004) examine both the events leading to the passage of SOX and several events related to the implementation of SOX that occurred between August and December 2002. However, SOX implementation activities continued in 2003. To investigate whether the post-SOX implementation events changed the expected costs of SOX, I examine the implementation activities of the SEC and the PCAOB, as directed by SOX, up to the auditing rule of Section 404 internal control test was proposed by the PCAOB. Zhang (2005, Table 3) reports U.S. stock returns around the 16 implementation events
subsequent to SOX passage. Most of the rulemaking events are not associated with significant market reactions, suggesting that the implementation of SOX did not significantly change the market’s expectations.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 109
4.4. Alternative explanations and sensitivity analyses
4.4.1. Alternative explanations
It is likely that some of the abnormal U.S. market returns around the rulemaking event days are not directly attributed to SOX. Both SOX and the message conveyed by SOX about future legislation affected stock prices and their impact cannot be disentangled. As a result, I cannot decisively conclude that SOX itself is costly. Yet, it is likely that the passage of SOX conveys bad news to the market because SOX is costly and signals the likelihood of further anti-business legislation. If SOX were beneficial to business, it is more plausible that investors would expect the government to write similar beneficial legislation in the future. Moreover, the cross-sectional tests in Section 4.2 provide additional support for the conjecture that SOX is costly. It is not clear why the expected costs of future anti-business legislation should vary systematically with firms’ purchases of non-audit services, shareholder rights, and the complexity of firms’ business.
Second, although returns of non-U.S.-traded foreign firms are used to filter out the impact of global economic news, the documented U.S. cumulative abnormal returns can reflect market reactions to other contemporaneous news unique to the U.S. or to the other foreign markets. However, the cross-sectional findings connect SOX-related legislative activities and the negative returns. If most of the market reactions reflected news unrelated to the SOX rulemaking, firms’ abnormal returns would not vary cross-sectionally as predicted.
In addition, as the market expectation of government reaction to accounting scandals is unobservable, I cannot completely rule out the hypothesis that SOX was beneficial but investors had expected different rules to correct the market failures and were dissatisfied with SOX. However, legal researchers, such as Romano (2004), argue that ‘‘congressional initiatives rarely are constructed from whole cloth; rather, successful law reform in the national arena typically involves the recombination of old elements that have been advanced in policy circles for a number of years prior to adoption.’’ This argument suggests that rational investors should not assign much likelihood to rules that differ significantly from available options. Future research that provides additional evidence on this alternative hypothesis will extend our understanding of the Act.
4.4.2. Sensitivity tests
To test the impact of the disclosure requirement of SOX, I examine whether firms’ cumulative abnormal returns around key SOX events vary with their disclosure practices. I use the S&P financial disclosure rating to proxy for disclosure quality and add it to regression (2). The inclusion of this variable reduces the sample to 421 firms. The disclosure rating is not significantly associated with CAR. The coefficient on Nonaudit is still significantly negative. The coefficient on Bindex remains negative but is insignificant, likely due to the smaller sample size. The coefficient remains insignificant for the smaller sample without the inclusion of the new variable.
Furthermore, I estimate regression (2) using alternative definitions of variables. I replace the governance index (Bindex) with measures of institutional investor holding, the percentage of independent members on the audit committee, and the percentage of independent board members. The coefficients on these three variables are negative but insignificant. While this may suggest that the governance results are sensitive to alternative specifications, it is also possible that the audit committee requirement lacks a significant
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115110
impact, as NYSE and Nasdaq had already required listed firms to establish independent audit committees with certain exceptions prior to SOX. Moreover, these variables may not capture the overall level of corporate governance. Bainbridge (2002) argues that a firm can have more insiders on the board if more specific knowledge is necessary in the decision making process. I redefine Nonaudit as the logarithm of total non-audit fees minus fees for tax related
services (Ferguson et al., 2004; Ashbaugh et al., 2003). All inferences remain intact. However, when the ratio of non-audit fees to audit fees is used, the coefficient on the fee ratio is insignificant. In this specification, it cannot be rejected that the restriction does not impose significant costs on firms. I also estimate an aggregate measure for litigation risks, which is calculated based on Shu’s (2000, Table 3) model to predict the likelihood of lawsuits. Again, the inclusion of the aggregate measure does not affect the main results of Table 5. In the cross-sectional tests, I estimate the market model for individual firms using data
prior to the first event to obtain the parameters and then calculate AR as the prediction error. Alternatively, following Schipper and Thompson (1983), I estimate the market model using data prior to and during the event period, allowing the intercept to shift around the event days. The coefficients on event-day dummies are then aggregated to compute CAR. I re-estimate regression (2) with CAR computed in this way as the dependent variable. The inferences are not affected.
5. Conclusion
This paper investigates the economic consequences of the Sarbanes–Oxley Act through a study of market reactions around the legislative events prior and subsequent to the passage of SOX. I find that the cumulative abnormal returns of U.S. firms and foreign firms complying with SOX around key SOX events are negative and statistically significant. The evidence is consistent with SOX imposing net costs on complying firms. The cumulative abnormal returns of U.S. firms around all SOX events are negative but not statistically significant. While this finding does not support SOX being costly, it is likely due to the inclusion of insignificant events reducing the power of the test. Furthermore, I investigate the sources of costs by examining the cross-sectional
implications of major provisions of SOX. I find that firms’ cumulative abnormal returns around key SOX legislative events decrease with their purchases of non-audit services and the complexity of their business. These findings are consistent with the hypotheses that the restriction of non-audit services and Section 404, requiring internal control tests, impose net private costs on firms. I also find that firms with shareholder rights weaker than their industry peers experienced more negative cumulative abnormal returns around SOX events. Finally, the test of market reactions to the announcement of postponing compliance with Section 404 shows that the postponement is particularly beneficial for small firms. The findings should be interpreted with caution as evidence of SOX being costly. First,
the documented negative cumulative abnormal return around key SOX rulemaking events likely reflects the impact of SOX, the expected costs of future anti-business regulation, and the impact of other contemporaneous news not captured by the expectation models (1a) and (1b). Not all of the documented negative cumulative return can be directly attributed to SOX. However, the cross-sectional test of the paper provides support for the argument
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 111
that SOX is likely costly and at least a major part of the negative cumulative abnormal return around key SOX events are related to SOX. Second, as the market expectation is unobservable, I cannot rule out entirely an alternative explanation that the losses reflect investors’ disappointment with SOX. While legal research provides little support for this hypothesis, future research that distinguishes between this alternative explanation and the cost explanation could provide stronger evidence regarding the impact of SOX. Third, SOX provisions are correlated, so are the examined firm characteristics in the cross- sectional analysis. Thus, although the cross-sectional results support certain SOX provisions being costly, SOX abnormal returns cannot be decisively attributed to a subset of provisions. Additionally, this paper focuses on an examination of private benefits and costs of SOX. Research on the social benefits and costs of the Act will help depict a more complete picture of SOX.
Appendix A. Related working papers on market reactions to SOX
Rezaee and Jain (2002) investigate S&P 500 index returns around the events leading to SOX, but the events they examine are largely a subset of the events listed in Table 2. For example, President Bush’s speech on July 9, 2002, which the media considered a signal of a change of attitude in Washington, is not included in their study. They find that the abnormal returns are positive around the final legislative events before SOX passage and negative around prior events. They argue that the market reacts positively as uncertainty is resolved and thus conclude SOX is value increasing. However, the market response captures only the unexpected portion of news. If the final rule reveals lower costs on firms than previously expected, positive abnormal returns can be observed around its announcement, even though investors consider the rule to be costly. Indeed, news reports indicated that lobbyists successfully pushed some of their proposals through at the last minute (e.g., Murray and Schroeder, 2002). Moreover, reports prior to the final ruling revealed concerns that the rules would impose greater costs than Sarbanes’ bill (Melloan, 2002). Thus, an examination of the cumulative abnormal return around all the SOX events provides stronger evidence than Rezaee and Jain’s (2005) focus on the final legislative events.
Rezaee and Jain’s (2005) also examine the relation between the abnormal returns of the S&P 500 firms around the final rulemaking event and firm characteristics. They find that firms’ abnormal returns increased with their S&P disclosure rating and decreased with their purchase of non-audit services in 2002. The findings are inconsistent with their claim that SOX provides net private benefits for firms. The reliability of their results and conclusion is further confounded by methodology issues (e.g., overlapping event windows and omitted correlated variables).
Li et al. (2004) examine the market reaction to the rulemaking events around SOX passage. However, they also do not provide a complete list of the events. For example, they exclude the negotiation of the House–Senate conference committee starting July 19 as an event. They argue that there was no news leakage prior to the issuance of the conference report on July 24. However, the opening statements of major lawmakers at the first conference meeting on July 19 set the tone of the final bill and were made available to the public. Democrat Rep. LaFalce made a public statement about the progress of the conference on July 22. Moreover, Hilzenrath (July 20, 2002, WP), VandeHei (July 21, 2002b, WP), Oppel (July 22, 2002, NYT), Weisman (July 23, 2002, WP), and Murray
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115112
(July 24, 2002b, WSJ) all discussed specific progress details of the negotiation and lobbying activities. The last two articles also revealed major disputes between the two parties by July 23, citing talks given by lawmakers. These articles do not support the argument of Li et al. (2004). Their cross-sectional test provides little support for their hypothesis that SOX is beneficial. Another difference between these two papers and my study is that I do not include the
presidential approval of the Act (July 30) as a major SOX event. The President’s comments in the rulemaking process revealed a strong indication that he would sign the bill into law and thus his approval was well expected. However, both studies find a significant positive cumulative return for the three-day window from July 29 to 31. The value-weighted returns for the three days are 5.31%, 0.47%, and 0.58%, respectively. Only the return of July 29 (Monday) is significantly different from zero. I argue that the positive return of July 29 is likely explained by the passage of the Trade Act or the Homeland Security bill in the House after the close of the stock market on July 26 (Friday). If so, I expect a significant positive return at the opening of the market on July 29. Table 3. Panel B shows that the previous-close-to-open return accounted for almost 40% of the daily return of July 29. Unreported results reveal that the realized market return within 1 h after the opening of the market accounted for 66% of the daily return. The significant positive opening return of July 29 was driven by news released prior to or around the opening of stock market. A search of the Dow Jones Newswire from July 26 to 27 does not find any news about President Bush’s approval of SOX. Thus, most, if not all, of the significant positive return on July 29 was likely unrelated to SOX. The positive return should not be included in the test of market reactions to SOX.
References
Ashbaugh, H., Lafond, R., Mayhew, B.W., 2003. Do non-audit services compromise auditor independence?
Further evidence. Accounting Review 78, 611–639.
Bainbridge, S.M., 2002. A critique of the NYSE’s Director Independence Listing Standards. Securities Regulation
Law Journal 30, 370–400.
Bainbridge, S.M., 2003. The creeping federalization of corporate law. Regulation Spring 2003, 26–31.
Barclay, M., Litzenberger, R.H., 1988. Announcement effects of new equity issues and the use of intraday price
data. Journal of Financial Economics 21, 71–99.
Bebchuk, L., Cohen, A., Ferrell, A., 2005. What matters in corporate governance? Working paper, Harvard Law
School.
Benston, G., 1973. Required disclosure and the stock market: an evaluation of the Securities Exchange Act of
1934. The American Economic Review 63, 132–155.
Berger, P.G., Li, F., Wong, M.H.F., 2005. The impact of Sarbanes–Oxley on cross-listed companies. Working
paper, University of Chicago and University of Michigan.
Bernard, V.L., 1987. Cross-sectional dependence and problems in inference in market-based accounting research.
Journal of Accounting Research 25, 1–48.
Bhattacharya, U., Groznik, P., Haslem, B., 2002. Is CEO certification credible? Regulation 26 (3), 8–10.
Blume, M., Stambough, R., 1983. Biases in computed returns: an application the size effect. Journal of Financial
Economics 12, 387–404.
Bushee, B.J., Leuz, C., 2005. Economic consequences of SEC disclosure regulation: evidence from the OTC
bulletin board. Journal of Accounting and Economics 39, 233–264.
Busse, J.A., Green, T.C., 2002. Market efficiency in real time. Journal of Financial Economics 65, 415–437.
Chaney, P.K., Philipich, K.L., 2002. Shredded reputation: the cost of audit failure. Journal of Accounting
Research 40, 1221–1245.
Chow, C.W., 1983. The impacts of accounting regulation on bondholder and shareholder wealth: the case of the
Securities Acts. The Accounting Review 58, 485–520.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 113
Coffee, J., 1984. Market failure and the economic case for a mandatory disclosure system. Virginia Law Review
70, 717–753.
Cohen, D.A., Dey, A., Lys, T.Z., 2004a. Trends in earnings management and informativeness of earnings
announcements in the pre- and post-Sarbanes Oxley periods. Working paper, Northwestern University and
University of Southern California.
Cohen, D.A., Dey, A., Lys, T.Z., 2004b. The Sarbanes Oxley Act of 2002: implications for compensation structure
and risk-taking incentives of CEOs. Working paper, Northwestern University and University of Southern
California.
Congress, 2002. The Sarbanes–Oxley Act of 2002.
Core, J.E., Guay, W.R., Rusticus, T.O., 2005. Does weak governance cause weak stock returns? An examination
of firm operating performance and investors’ expectations. Journal of Finance 61 (2), 655–687.
Cremers, K.J.M., Nair, V., 2005. Governance mechanisms and equity prices. Journal of Finance 60 (6),
2859–2894.
DeFond, M., Hann, R.N., Hu, X., 2005. Does the market value financial expertise on Audit Committees of
Boards of Directors? Journal of Accounting Research 43, 153–193.
Engel, E., Hayes, R.M., Wang, X., 2006. The Sarbanes–Oxley Act and firms’ going-private decisions. Journal of
Accounting and Economics, in press, doi:10.1016/j.jacceco.2006.07.002.
Eun, C.S., Shim, S., 1989. International transmission of stock market movements. Journal of Financial and
Quantitative Analysis 24 (2), 241–256.
Ferguson, M.J., Seow, G., Young, D., 2004. Nonaudit services and earnings management: U.K. evidence.
Contemporary Accounting Research 21 (4).
Financial Executive International (FEI), March 2005. Section 404 costs survey, hhttp://www.fei.org/
404_survey_3_21_05.cfmi.
Frankel, R.M., Johnson, M.F., Nelson, K.K., 2002. The relation between auditors’ fees for non-audit services and
earnings management. Accounting Review 77, 71–105.
Friend, I., Westerfield, R., 1975. Required disclosure and the stock market: comment. American Economic
Review 65, 467–472.
Forbes, K.J., Rigobon, R., 2002. No contagion, only interdependence: measuring stock market comovements.
Journal of Finance 57 (5), 2223–2261.
Ge, W., McVay, S., 2005. On the disclosure of material weaknesses in internal control after the Sarbanes–Oxley
Act. Accounting Horizons 19 (3), 137–158.
Greenstone, M., Oyer, P., Vissing-Jorgensen, A., 2006. Mandated disclosure, stock returns, and the 1964
Securities Acts amendments. Quarterly Journal of Economics 121 (2), 399–460.
Gompers, P.A., Ishii, J.L., Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of
Economics 118 (1), 107–155.
Gonedes, N., Dopuch, N., 1974. Capital market equilibrium, information production and selecting accounting
techniques: theoretical framework and review of empirical work. Journal of Accounting Research 14,
89–137.
Hamao, Y., Masulis, R.W., Ng, V., 1990. Correlations in price changes and volatility across international stock
markets. Review of Financial Studies 3 (2), 281–307.
Harris, M., Raviv, A., 2005. A theory of board control and size. Working paper. University of Chicago and
Northwestern University.
Holmstrom, B., Kaplan, S., 2003. The state of U.S. Corporate governance: what’s right and what’s wrong?
Journal of Applied Corporate Finance 15, 8–20.
Leftwich, R., 1981. Evidence of the impact of mandatory changes in accounting principles on corporate loan
agreements. Journal of Accounting and Economics 3, 3–36.
Leuz, C., Triantis, A., Wang, T.Y., 2006. Why do firms go dark? Causes and economic consequences of voluntary
SEC deregistrations. Working paper, University of Chicago, University of Maryland, and University of
Minnesota.
Li, H., Pincus, M., Rego, S., 2004. Market reaction to events surrounding the Sarbanes–Oxley Act of 2002.
Working paper, University of Iowa.
Litvak, K., 2005. The effect of the Sarbanes–Oxley Act on non-US companies cross-listed in the US. Working
paper, the University of Texas School of Law.
Lo, K., 2003. Economic consequences of regulated changes in disclosure: the case of executive compensation.
Journal of Accounting and Economics 35, 285–314.
Lys, T., Watts, R.L., 1994. Lawsuits against auditors. Journal of Accounting Research 32, 65–93.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115114
Mahoney, P.G., 1995. Mandatory disclosure as a solution to agency problems. University of Chicago Law Review
62, 1047–1112.
Mayer, Brown, Rome, Maw, LLP, 2003. The U.K.’s answer to Sarbanes–Oxley?—U.K. corporate governance
reforms.
Rezaee, Z., Jain, P., 2005. The Sarbanes–Oxley Act of 2002 and security market behavior: early evidence.
Working paper, University of Memphis.
Ribstein, L.E., 2002. Market vs. regulatory responses to corporate fraud: a critique of the Sarbanes–Oxley Act of
2002. Journal of Corporation Law 28 (1).
Romano, R., 2004. The Sarbanes–Oxley Act and the making of quack corporate governance. Working paper,
Yale University, NBER and ECGI.
Schipper, K., Thompson, R., 1983. The impact of merger-related regulations on the shareholders of acquiring
firms. Journal of Accounting Research 21, 184–221.
Schwert, G.W., 1981. Using financial data to measure effects of regulation. Journal of Law and Economics 24 (1),
121–158.
Sefcik, S.E., Thompson, R., 1986. An approach to statistical inference in cross-sectional models with security
abnormal returns as dependent variable. Journal of Accounting Research 24, 316–334.
Shu, S.Z., 2000. Auditor resignations: clientele effects and legal liability. Journal of Accounting and Economics
29, 173–205.
Shumway, T.G., 2001. Forecasting bankruptcy more accurately: a simple hazard model. Journal of Business 74
(1), 101–124.
Stigler, G.J., 1964. Public regulation of the securities markets. Journal of Business 37 (2), 117–142.
Watts, R.L., Zimmerman, J.L., 1986. Positive Accounting Theory. Prentice-Hall, Upper Saddle River, New
Jersey.
Zhang, I.X., 2005. Economic consequences of the Sarbanes–Oxley Act of 2002. Unpublished thesis, University of
Rochester.
Articles from The Wall Street Journal or The Washington Post to identify the legislative
events
Calmes, J., 2002. Two bills put Coast Guard, FEMA in Security Agency. Wall Street Journal, July 18.
Cummings, J., 2002. Bush to seek tougher penalties in assault on corporate fraud. Wall Street Journal, July 9.
Cummings, J., Schlesinger, J.M., Schroeder, M., 2002. Securities threat: Bush crackdown on business fraud
signals new era. Wall Street Journal, July 10.
Day, K., Crenshaw, A.B., 2002. SEC, Accounting Firms Redrafting Audit Rules; Agency Chairman Draws Fire
for Role in Effort. The Washington Post, January 16.
Francis, T., 2003. It still costs big to insure against a boardroom scandal - despite the Sarbanes-Oxley Bill, ‘D&O’
policy prices rise 30%, and cancellation clauses swell. Wall Street Journal, July 31.
Hamburger, T., Hitt, G., Schroeder, M., 2002. Sorry, wrong number: WorldCom case boosts Congress in reform
efforts. Wall Street Journal, June 27.
Hilzenrath, D.S., 2001. Pitt seeks closer watch on auditors: SEC chief wants improvements in accounting’s self-
regulation. The Washington Post, December 12.
Hilzenrath, D.S., 2002. Hill begins search for compromise on accounting rules. The Washington Post, July 20.
Hilzenrath, D.S., Weisman, J., Vandehei, J., 2002. How Congress rode a ‘storm’ to corporate reform. The
Washington Post, July 28.
Hitt, G., 2002. Bush signs sweeping legislation aimed at curbing corporate fraud. Wall Street Journal, July 31.
Johnston, D.C., 2002. House takes up, but drops, Bermuda Corporation issue. The New York Times, July 19.
Kristof, K.M., Flanigan, J., 2002. Crisis in corporate America; stock options lavished on executives come under
fire. Los Angeles Times, July 12.
Markon, J., Frank, R., 2002. Five Adelphia officials arrested on fraud charges — three in the Rigas family, two
other executives held, accused of massive looting. Wall Street Journal, July 25.
Melloan, G., 2002. Can capitalism survive election year politics? Wall Street Journal, July 23.
Murray, S., 2002a. Leading the news: house GOP moves on oversight bill, with few changes. Wall Street Journal,
July 18.
Murray, S., 2002b. Lawmakers near completion of accounting-regulation bill. Wall Street Journal, July 24.
ARTICLE IN PRESS I.X. Zhang / Journal of Accounting and Economics 44 (2007) 74–115 115
Murray, S., Schroeder, M., 2002. Governance bill has major consequences for many. Wall Street Journal, July 26.
Norris, F., 2004. Too much regulation? Corporate bosses sing the Sarbanes–Oxley blues. New York Times,
January 23.
Oppel, R.A., 2002. Little progress on corporate governance bill. New York Times, July 22.
Schroeder, M., 2002a. Lawmakers plan more financial oversight—accounting industry faces stepped-up
regulation and limits on services. Wall Street Journal, February 12.
Schroeder, M., 2002b. Senate panel seeks sweeping change for auditors. Wall Street Journal, March 7.
Schroeder, M., 2002c. The economy: Congress seeks harsher penalties for violations of accounting laws. Wall
Street Journal, April 23.
Schroeder, M., Hamburger, T., 2002. Accounting reform is left in SEC’s lap—Oxley proposal would place Pitt in
an awkward position. Wall Street Journal, April 17.
Solomon, D., 2003. Fraud detector: SEC sets a new rule aimed at companies’ internal controls. Wall Street
Journal, May 28.
Solomon, D., Bryan-Low, C., 2004. Companies complain about cost of corporate-governance rules. Wall Street
Journal, February 10.
VandeHei, J., 2002a. House GOP’s leaders fight audit plan; lawmakers seek to dilute some senate-passed reforms.
The Washington Post, July 17.
VandeHei, J., 2002b. Conferees to haggle with eye on Nov.; political stakes high in corporate reform. The
Washington Post, July 21.
VandeHei, J., Hilzenrath, D.S., 2002. Hill leaders agree on corporate curbs; attack on fraud includes auditing
control and jail terms. The Washington Post, July 25.
Wallison, P., 2003. Blame Sarbanes–Oxley. Wall Street Journal, September 3.
Weisman, J., 2002. Lobbyists lose clout; business groups find their influence on audit reform legislation shrinking.
The Washington Post, July 23.
- Economic consequences of the Sarbanes-Oxley �Act of 2002
- Introduction
- Related research and SOX event history
- Related research
- Event history
- Hypotheses
- U.S. market reaction to SOX
- Specific provisions of SOX and their cross-sectional implications
- Provisions on non-audit services
- Provisions on corporate accountability and criminal penalties
- Provisions on internal controls
- Provisions on incentive pay and insider trading
- Empirical tests and results
- Overall U.S. market reaction to the legislative events leading to SOX
- Research design
- Cumulative market returns in the U.S. over the event period
- Returns of U.S.-exchange-listed foreign firms
- Confounding events and intraday returns
- Cross-sectional variation in abnormal returns around the rule-making events
- Research design, sample selection, and descriptive statistics
- Regression results
- Market reactions to the announcement of postponing compliance with Section 404
- Subsequent events
- Alternative explanations and sensitivity analyses
- Alternative explanations
- Sensitivity tests
- Conclusion
- Related working papers on market reactions to SOX
- References
- Articles from The Wall Street Journal or The Washington Post to identify the legislative events
