Our study is related to 2 lines of research, one is to ...

9 downloads 1716 Views 491KB Size Report
defines three levels of problems (from the most to the least severe): material ... broad FDICIA but is a major component of the SOX Act. Neither of the two studies.
Introduction SOX is the most far-reaching reform in the financial reporting and corporate governance requirements for public companies since the 1930s (Donaldson, 2005; Li, Pincus, and Rego, 2008). The Act was designed with the goals of improving financial reporting and restoring investor confidence in publicly traded firms (Donaldson, 2005). Section 404 is one of the most visible and tangible changes brought by the Act (Auditing Standard No. 2 [AS2] of the Public Company Accounting Oversight Board [PCAOB], 2004). The pivotal requirements of Section 404 are that management and an external auditor separately assess the effectiveness of the firm’s internal controls over financial reporting and each issue a report of their conclusions as a result of their assessments. The U.S. Securities and Exchange Commission (SEC) set separate compliance schedules for different firms. Accelerated filers (primarily large firms) were required to begin complying with Section 404’s requirements in the fiscal year ending after November 15, 2004. Non-accelerated filers (primarily small firms) and foreign firms were granted a delayed and phased implementation schedule starting in 2006 and ending in 2009 when those two groups reach full compliance. 1 The first goal of our study is to examine whether the implementation of Section 404 has helped to improve the quality of financial reporting. AS2 states that: “effective controls provide the foundation for reliable financial reporting…Internal control over financial reporting enhances a company’s ability to produce fair and complete financial reports” (Paragraph E.5., Appendix E). Therefore if Section 404 prescribes an effective

1

See Table 1 in Gao, Wu and Zimmerman, 2008 for the chronology of compliance schedule for accelerated filer, non- accelerated filers and foreign firms, and Table 1 of this paper for the final implementation dates.

1 Electronic copy available at: http://ssrn.com/abstract=1052541

process to improve Internal Control over Financial Reporting (ICFR), then one should expect to observe improvement in the quality of financial reporting. 2 The second goal of our study is to find if the implementation of Section 404 was followed by an increase in investor confidence in financial reporting as was intended by SOX.

AS2 states the following: “Without reliable financial reports, making good

judgments and decisions about a company becomes very difficult for anyone, including the board of directors, management, employees, investors, lenders, customers, and regulators. The auditor's reporting on management's assessment of the effectiveness of internal control over financial reporting provides users of that report with important assurance about the reliability of the company's financial reporting” (Appendix E, paragraph 2). Therefore an increase in the quality of financial reporting should lead to more reliable financial statements and to an increase in investor confidence in financial reporting. Yet we should observe an increase in investor confidence only if they are aware of the improvement in financial reporting quality. Few studies examine the effect of SOX on various parameters of financial reporting quality. Lobo and Zhou, 2006 find a decrease in the use of discretionary accruals and an increase in conservatism in reporting following the passage of SOX. Similarly Cohen, Dey, and Lys, 2008 report a reduction in accrual-based earnings management in the post-SOX period. They attribute the decline to the passage of the Act, but acknowledge that other factors might have contributed to it as well. The coexistence of complying and non-complying firms with Section 404 due to differential implementation schedule allows us to adopt a difference-in-differences approach to

2

In section 2 we discuss in greater detail how strong ICFR should lead to reliable, unbiased and accurate financial statements.

2 Electronic copy available at: http://ssrn.com/abstract=1052541

compare the changes in reporting quality and investor confidence of the complying firms to those of a control group of non-complying firms. This allows us to better isolate the effects of the regulation while controlling for other confounding events which should mitigate the problem of omitted correlated variables. We use multiple measures of earning quality to ensure that our results are not driven by any single measure and to mitigate the potential effect of correlated omitted variables. In the post-404 period and in comparison to the control group, complying firms had a significantly larger reduction in the magnitude of absolute abnormal accruals, a significantly larger improvement in the ability of earnings to predict future cash flows and future earnings, and a larger reduction in the asymmetry between the use of negative and positive special items. Both groups had a significant reduction in the asymmetry between reporting earnings that slightly top analysts’ forecasts and reporting earnings that slightly miss the analysts’ forecasts. These findings lead us to conclude that Section 404 has brought a significant improvement to the quality of financial reporting. 3 Regarding our second research question, in the post-404 period we find an increase in the association between abnormal stock returns and analysts’ forecast errors in the three days around the earnings announcements for the complying firms relative to the control group.

This suggests that investors increased their reliance on accounting

information in their decision-making process after the implementation of Section 404 and provides evidence of an increase in investor confidence in financial reporting. Our control group comprises of Canadian firms that were either dual listed in the U.S. and in Canada or only listed on a U.S. exchange during our sample period. Canada

3

Due to findings by ACKL, 2008 that firms that report remediation in their internal controls improve their accrual quality, we repeat our tests excluding those firms and our findings still hold.

3 Electronic copy available at: http://ssrn.com/abstract=1052541

implemented regulations similar to SOX except that those regulations did not include requirements similar to those in Section 404. Given that U.S. regulations granted foreign filers a delay in implementation of Section 404, our control firms were not required to comply with Section 404 during our sample period. 4 The implementation of Section 404 has resulted in some companies discovering and reporting material weaknesses in their internal controls. ACKL, 2008 find that firms that remediated deficiencies in their internal controls exhibited an increase in their accrual quality. Their findings on the positive affect of Section 404 are therefore limited only to the small subsample of firms that reported remediation of internal control deficiencies (ICDs). Our view is that the quality of ICFR varies across firms along a continuum rather than being sufficient or not. We believe that if the process of testing and evaluating ICFR is comprehensive and effective then we would expect all firms to improve their ICFR and therefore examine the effect of Section 404 on all complying firms. In support of the view of variation in ICFR with regard to deficiencies, AS2 defines three levels of problems (from the most to the least severe): material weaknesses, significant deficiencies, and control deficiencies. 5 Hammersley, Myers and Shakespeare, 2008 document stronger market reaction to firms disclosing material weaknesses than to firms disclosing significant deficiencies. This suggests that investors also seem to hold this view. We believe that also for firms without severe control problems the quality of ICFR varies from state-of-the-art control systems to systems that are only slightly above the need to report deficiencies. The decision to require all firms to implement Section 404 rather than to limit it to firms reporting ICDs under Section 302, suggests that 4

We discuss the alternative controls in depth in section 4.2. The main difference between the three categories is in the degree of likelihood a misstatement will be prevented or detected on a timely basis. 5

4

regulators believed that all firms could benefit from implementing Section 404. Another reason not to limit the sample to firms that report ICDs is that firms that find flaws in their internal controls and repair those problems before the fiscal year end are not required to report material weaknesses, but are likely to improve their reporting quality. 6 Our study is most closely related to Ashbaugh, Collins, Kinney, and LaFond (ACKL), 2008 that document improvement in accrual quality for firms that remediated previously reported ICDs.

We make several important contributions beyond the

important findings in that study. First we document the improvement in accrual quality not only for the small group of firms that remediated their ICDs, but for all the firms that complied with Section 404.

Second, ACKL, 2008 document only reduction in

unintentional misstatements, Effective control systems can prevent both intentional and unintentional misstatements in financial reporting (ACKL, 2008).

Our findings of

reduction in absolute abnormal accruals and of increase in the power of earnings to predict future cash flows and future earnings can be due to reduction in both types of misstatements. However, our findings of decreases in the asymmetries between reporting negative special items and positive special items and between the propensities to slightly top the analysts’ forecast and slightly miss the analysts’ forecast are likely due to a decrease in intentional misstatements.

Third, we document an increase in investor

confidence over financial reporting following the improvement in the quality of financial reporting. Fourth, ACKL only study the magnitude of abnormal accruals, while our findings rely on much broader set of measures of financial reporting quality. Our study 6

In addition, all accelerator filers bear the cost of complying with Section 404 and improvement in reporting quality by only small subsample of firms might not warrant the high costs that are shared by all the complying firms.

5

contributes to the literature on the effect of increased internal controls regulations on financial reporting in the U.S. (Altamuro and Beatty, 2007) and in a non-U.S. regulatory regime (Brown, Strohm, and Wompener, 2008). Altamuro and Beatty, 2007 examine whether the internal control regulation of the Federal Depository Insurance Corporation Improvement Act (FDICIA) on banks’ earning quality. SOX 404 provides a cleaner setting than FDICIA as the Internal control regulation represents only a small part of the broad FDICIA but is a major component of the SOX Act. Neither of the two studies examines the effect of the improvement in earning quality on investors. Our study also adds to the large body of literature on SOX regulation. Finally, our study contributes to the ongoing debate regarding the costs and benefits of Section 404 by highlighting some of the improvements brought by this regulation and the reestablishment of investor confidence in financial reporting. Our study is subject to some caveats. It is possible that the improvement in earning quality we document in the post-404 period and attribute to this regulation is due to a late response by firms to SOX as a whole rather than to Section 404 specifically. In this regard, the fact that we document significantly larger improvement for our sample group than for the control group should mitigate this possibility. It is also possible that other uncontrolled events or other factors drive our results. To defuse these possibilities, we provide additional robustness tests controlling for the effect of the implementation of Sections 301 and 401 of SOX as well as for other factors that might affect our results and the main findings still hold. Lastly, it is still possible that uncontrolled factors affect the two groups differently and drive the results. Therefore we provide evidence on why we believe that our control group is appropriate for comparison. We conclude that it is

6

reasonable to attribute at least some of the improvement in financial reporting to Section 404. The paper proceeds as follows: Section 2 provides background information on SOX and Section 404. Section 3 discusses our research questions and reviews related literature. Section 4 describes the data, the sample construction, and the sample periods examined. Section 5 describes the empirical tests and their results. In Section 6 we examine alternative explanations for our results, and Section 7 concludes.

2. Institutional Background Internal control over financial reporting is a process designed to provide reasonable assurance for the reliability of financial reporting and the preparation of financial statements for external purposes in accordance with Generally Accepted Accounting Principles (GAAP).

The process includes maintenance of records that

accurately and fairly reflect transactions and dispositions of the company’s assets. The process should also assure prevention or timely detection of unauthorized acquisitions, and use or deposition of assets that could have a material effect on the financial statements (AS2, paragraph 7). An ICD exists when the design of the control or its operation does not allow the prevention or detection of misstatement on a timely basis (AS2, paragraph 8).

ICDs can lead to unintentional misstatements due to lack of

adequate policies, training of personnel or inconsistent application of these policies. Those deficiencies could cause transactions to be recorded incorrectly. ICDs can also lead to intentional misstatements through misrepresentations and omissions by employees or management that are not prevented or detected due to inappropriate controls. Both

7

intentional and unintentional misstatements may cause the financial statements to be less accurate, more biased (in the case of intentional misrepresentations), less reliable and less relevant (See ACKL, 2008, section II for a comprehensive discussion on ICFR and ICDs).

Tighter internal controls can significantly lower the likelihood of such

misstatements (AS2; Donaldson, 2005). SOX was enacted in August of 2002 following high-profile scandals in firms such as Enron, WorldCom, Global Crossing, and amid growing distrust in the integrity of corporate disclosures and financial reporting (Donaldson, 2005). SOX is considered the most comprehensive overhaul of the corporate financial reporting environment since the adoption of the Securities Act of 1934. One very important aspect of SOX is ICFR. SOX Section 302 requires that the CEO and the CFO of a public company certify in the quarterly and annual financial statements that they have reviewed the financial statements and that to the best of their knowledge the financial statements do not contain material misstatements and fairly represent the financial condition and the results of operations. In addition, they must certify that they have: (a) designed internal controls to ensure that any material information is made known to them during the period in which financial statements are prepared; (b) evaluated the effectiveness of internal controls; and (c) presented their conclusions regarding the effectiveness of internal controls in a report. Any significant deficiencies in the design and operations of the internal controls and any fraud they find must be reported to the firm’s audit committee and to the firm’s auditor. Section 404 and the SEC related implementing rules are much more stringent with respect to ICFR than Section 302’s requirements. They require the filing manager to test the internal control structure and procedures and assess their effectiveness. They also

8

require the firm’s external auditor to attest to the accuracy of the assessment done by the filing manager. The auditor then must issue an independent report on the effectiveness of the internal controls. The assessments of management and of the external auditor should be disclosed in a report to be included in the annual financial statements. 7 Evidence that Section 404 process is much stricter than Section 302 process is the fact that 87 percent of the firms that reported control deficiencies in the early months of compliance with Section 404 certified their control systems as effective under Section 302 in previous report (Glass, Lewis & Company, 2005). 8 Therefore it is likely that in those cases control problems already existed but were not discovered before firms started to follow the more rigorous requirements of Section 404. 9 The amount of work and resources to comply with Section 404 are incomparable to those required for Section 302. A survey of 321 companies by Financial Executives International from 2004 found the additional average compliance costs of Section 404 for a company to include $1.3 million in spending on external consulting and software, additional audit fees of $1.5 million, and 35,000 hours of internal manpower. On top of the above costs, firms incur opportunity costs of management time that can be otherwise devoted to other projects (The U.S. Government Accountability Office, 2006).

Iliev, 2008 finds that compliance with

Section 404 more than doubled audit fees for firms just above the threshold for compliance (from $370,700 to $882,300).

He also observes that firms above the

threshold had substantially lower stock returns than firms below the threshold in their 7

Companies are also required to perform quarterly evaluations of changes that have materially affected or are likely to materially affect ICFR. 8 For our sample, only 2.6% of the firms report ICDs in 2002-03 under Section 302, but this figure jumps to 14.2% in 2004-2005 under Section 404. 9 Prior to SOX, internal control requirements were based on the Foreign Corrupt Practices Act of 1977. The Act required the management of public companies to protect corporate assets, but it neither required management to evaluate or certify the effectiveness of ICFR, nor the external auditor to report to the board of directors on any deficiency noted during the audit of the financial statements.

9

first year of compliance. This could be explained by the high compliance costs. Eldrige and Kealey, 2005 report an average increase in audit fees from 2003 to 2004 of $2.3 million, which is primarily attributed to the new Section 404 audit requirements. Nondorf, Singer, and You, 2008 and Gao et al., 2009, provide evidence that firms around the threshold for compliance undertook actions to postpone the implementation of Section 404. 10 All this evidence stresses the significant additional work and cost for complying with Section 404. Implementation of the Act went smoothly except for Section 404 (Cox, 2006), which prompted fierce controversy and was subject to many delays owing to pressure put on the SEC by public firms, especially small businesses concerned with the high implementation costs. The pressure led the SEC to set separate compliance schedules for different firms. Accelerated filers 11 began complying with Section 404’s requirements in the fiscal year ending after November 15, 2004, while non-accelerated filers and foreign firms implement Section 404 in phases starting in 2006 and ending in 2009.

Many

academics and practitioners have debated the merit of Section 404. Its opponents claim that implementation costs are excessively high and exceed their benefits, especially for small firms (AeA, 2005), because the costs are not scaled to company size (Financial Executives International, 2004; Eldridge and Kealy; 2005, Iliev, 2008), are of excess burden for foreign firms (Russell Reynolds Associates, 2006), and cause domestic firm to go private or go “dark” (Engel, Hays and Wang, 2007; Hansen, Pownall and Wang, 2008; 10

It was also admitted by the SEC that actual implementation costs were much higher than the committee’s initial estimate. 11 Accelerated filer is a company that its aggregate worldwide market value of voting and non-voting common equity held by non-affiliates on the last business day of the second quarter of a fiscal year is larger than $75 millions. An affiliate is defined by the SEC in 17 CFR 230.144(a) (1) as “a person that directly, or indirectly through one or more intermediaries, controls, or is controlled by, or is under common control with (the) issuer.”

10

Leuz, Triantis, and Wang, 2008).

Its proponents, on the other hand, maintain that

investors have benefited from stronger controls, greater transparency, and more rigorous accounting quality (Donaldson, 2006; Interim Report of the Committee on Capital Markets Regulation, November 2006).

Morse, 2006 attributes some of the recent

stronger tax compliance norms to SOX Section 404.

3. Research Questions and Extant Literature 3.1 Research Questions Our main research goal is to assess empirically whether the implementation of Section 404 led to a significant improvement in financial reporting as was intended by the regulations.

Section 404 prescribes a rigorous process for management to test and

evaluate ICFR and for the external auditor of the company to assess the quality of internal controls.

While studies show that internal controls regulations bring

improvement to financial reporting (Altamuro and Beatty, 2007; Brown et al., 2008), it is not obvious to expect similar findings for Section 404. Section 404 was implemented more than two years after the passage of SOX. Additionally, Section 302 of the Act already addresses internal control issues, so it is possible that most of the potential effect of SOX on financial reporting was already realized before Section 404 was implemented. Finally, Zhang, 2007, finds that U.S. firms experienced a negative abnormal return around key SOX events, and Litvak, 2007 reports that stock prices of foreign firms subject to SOX (listed on level 2 or 3 ADRs) declined compared to those of foreign firms not subject to SOX (listed on level 1 or 4 ADRs). Those findings suggest that the market

11

had doubts regarding the effectiveness of SOX. 12

We are interested in examining

whether there was an improvement in the quality of financial reporting, and if it was only due to a reduction in unintentional misstatements or also due to reduction in intentional misstatements. ACKL, 2008 document an improvement in accrual quality for firms that remediated their ICDs. They find evidence of a reduction in unintentional misstatements but fail to find evidence for a reduction in intentional misstatements. ACKL reach their conclusion regarding intentional misstatements because their test of signed abnormal accruals shows no significant change after the remediated of ICDs. We believe that this test suffers from low power problems for two reasons. First, accruals reverse over time. Even if a manager misstated earnings in one reporting period, there will be another period where accruals will be misstated in the opposite direction. Therefore a test that measures abnormal accruals over a long enough period might not detect those misstatements because the two misstatements are likely to be cancelled out. Second, if a manager’s reporting objective is to smooth earnings or to avoid missing the average analysts’ forecast, then a manager that is short of meeting the target will manage earnings upward, but during periods where the target is far exceeded, she might choose to intentionally lower the reporting earnings and create reserves. These reserves can be used in the future to meet the target. Because earnings might be upward misstated for some companies and downward misstated for other companies, once again, a signed test of abnormal accruals might not reveal intentional misstatements. To overcome this problem we included tests that are designed to measure changes in intentional misstatements.

12

Jain and Rezaee, 2006 however, provide evidence of positive (negative) market reaction to legislative events that increased (decreased) the likelihood of the passage of the Act.

12

Our second research question is whether investors increased their reliance on accounting data for their investment decisions in the post-404 period. Improvement in financial reporting quality should result in more reliable financial statements that better measure the true performance of firms. Therefore it is reasonable to expect investors to rely more on the accounting data. Yet, investors’ decision is affected by their beliefs regarding the change in earning quality which might or might not coincide with the true change. Therefore it is an empirical question of whether the reliance on accounting data has increased. 3.2 Literature Review Our study is related to three lines of research. The first line of related research examines internal control deficiencies reported under SOX Sections 302 or 404. Studies that examine the characteristics of firms that report internal control weaknesses find that those firms are smaller in size, underperform and have higher betas (Bryan and Lilien, 2005), engage in more earnings management activities (Chan, Farrell, and Lee, 2005), have lower-quality accruals (Doyle, Ge, and McVey, 2007; ACKL, 2008), experience higher cost of equity (Ogneva, Raghunandan, and Subramanyam, 2005; ACKL, 2009), have lower return-earnings association (Chan et al., 2005), are associated with audit delays (Ettredge, Li, and Sun, 2006), and pay higher audit fees (Bedard, Hoitash, and Hoitash, 2006). Cheng, Ho, and Tian, 2006 find a negative abnormal stock return around the earnings announcement for firms that report material weaknesses in their internal control under Section 404.

Hammersley, Myers, and Shakespeare, 2008 find that

characteristics of the material weaknesses problems, such as their severity, the ability of the auditor to audit around the weakness, and the amount of detail about the weakness

13

provided in the disclosure are all informative to investors. Lastly, ACKL, 2008 report that firms that remediated previously reported deficiencies in internal control improved their accrual quality and that the improvement is only due to a reduction in unintentional misstatements. We contribute beyond the findings in those studies by examining the effect of Section 404 on all complying firms and by providing some evidence on reduction in intentional misstatements. The second line of related research examines changes in characteristics of financial reporting in the post-SOX period. Cohen, Dey, and Lys, 2005 find a decline in earnings management and no change in the informativeness of earnings following the passage of SOX, while Cohen, Dey, and Lys, 2008 report a shift from accrual earnings management to real earnings management in the post-SOX period. Bartov and Cohen, 2007 find a significant decline in the propensity to meet or beat analysts’ expectations in the post-SOX period driven by a reduction in expectation management and in accrual management. Relative to those studies, we employ a more comprehensive set of tests to measure changes in financial reporting quality. In addition, the use of a control group and the difference-in differences approach allow us to better isolate the effect of SOX 404. Without a control group, it is not clear whether the results found in the existing research are due to SOX or driven by significant contemporaneous economic events. The last related line of research examines the effect of new internal controls regulations on financial reporting. Altamuro and Beatty, 2007 examine the effect of internal controls reform mandated by the Federal Deposit Insurance Corporation Improvement Act of 1992 on the banking industry. Brown, Strohm, and Wompener, 2008 study the 1998 German legislation on control and transparency (KTG). Both

14

studies find that the reforms brought substantial improvement to financial reporting. Chang, Chen, Liao, and Mishra, 2006 study the SEC order from June 2002 requiring CEOs and CFOs of firms with revenues of at least $1.2 billions to certify their financial results under oath.

They find positive abnormal stock returns on the day of the

implementation (August 14, 2002) and a decline in the bid-ask spread for those firms in the post-certification period, especially for firms that prior to the new order practiced aggressive revenue recognition and for firms that used Arthur Andersen as their external auditor. They conclude that the certification provided assurance to investors regarding the credibility of the disclosure and reduced information asymmetry.

4. Data, Sample Period and Sample Construction 4.1 Data, and Sample Period An analysis of a regulatory change relies on a comparison of the period prior to the regulation to the period afterward.

New regulations are introduced continually,

therefore, the decision regarding the time horizon to examine involves a trade-off between (1) using a longer period that will provide a richer set of information to examine, but will also increase the chances of the inference being influenced by other uncontrolled factors, and (2) using a shorter period that will reduce the effect of other regulations and uncontrolled factors but might use insufficient information. We take an approach that balance this trade-off and compare the first two years of the regulation, 2004-05, to the two prior years, 2002-03. To ensure that results are not driven by changes due to Section 302 or other provisions of SOX, but are due to the implementation of Section 404, for most of our

15

tests we divide the sample period into three sub-periods: the pre-SOX period, January 2001 through July 2002; the post-SOX, pre-404 period, August 2002 through December 2003; and the post-404 period, January 2004 through December 2005. Figure 1 depicts the partitioning of the sample into two and three sub-periods. We identify firms as accelerated filers based on data from Audit Analytics. We use financial data from Standard & Poor’s Compustat. We remove observations of financial institutions (SIC 6000–6999) because some of those firms, in particular, investment companies, are registered under Section 8 of the Investment Company Act of 1940 and are not subject to Section 404 of SOX. 13 We also remove firms in regulated industries (SIC 4910–4939) because those firms might be subject to different reporting incentives. Stock returns data is obtain from CRSP, and data on analysts’ forecast data is from I/B/E/S. 4.2 Sample and Control Groups Construction For our sample group we include companies that were either accelerated filers or accelerated large filers 14 and reported on Section 404 in both 2004 and 2005. For our control group we choose Canadian companies that their shares were either dual listed in the U.S. and in Canada or only listed on a U.S. exchange. Those firms were not required to comply with Section 404 in the first two years of the implementation.

Canada

implemented regulations similar to SOX except that Canada stopped short of adopting requirements similar to those in Section 404, therefore our control firms were not required by either U.S. or Canadian regulations to comply with Section 404. 13

In addition, some federally insured depository institutions are subject to similar regulation under the Federal Deposit Corporation Act. While not exempted from Section 404, they might not be affected in the same way that other firms are affected by Section 404. 14 Because the implementation schedule is identical for large accelerated filers and for accelerated filers we do not distinguish between those two categories for our tests.

16

Additionally, Canada’s economy is highly dependant on the U.S. economy. Around 80 percent of Canada’s export is to the U.S. and close to 70 percent of its import is from the U.S (Statistics Canada, Canada's National Statistics Agency, 2007). We also notice a very high correlation (0.75) in the changes in the gross domestic product (GDP) between the two countries for the years 1998-2007. The Canadian firms in our sample have mean (median) of 41.0 percent (43.6 percent) of their sales to the U.S. They report their results in either U.S. or Canadian dollars according to Canadian GAAP, which is mostly similar to U.S. GAAP. 15 Like other foreign filers, they provide reconciliation to U.S. GAAP. 74.4 percent of the Canadian firms are audited by a Big 4 firm compared with 87.7 percent of the sample firms. We do not include foreign firms from other countries for two reasons. First, the economy of those countries might be very different than that in the U.S. which makes comparison to U.S. companies hard to interpret. Second, while those foreign firms were exempted from Section 404, they might have been subject to other rules and regulations similar to Section 404 in their countries of incorporation (Cox 2007; Tafara, 2006). 16 Our sample group comprises 1,620 complying firms and the control group includes 263 Canadian firms. Because the complying firms are somewhat larger in size than the control firms, to create two groups of closer size, we eliminate the smallest 5 percent of the Canadian firms and the largest 5 percent of our sample firms, which brings down the number of complying firms to 1,540 and the control firms to 249. After this procedure the two groups have similar mean market value of equity (MVE) ($2.04

15

Only during 2006 Canada decided to start a project of convergence to the International GAAP, which will happen in 2011. 16 Tafara, 2006 and Cox, 2006 provide evidence of widespread global adoption of the major provisions of the Sarbanes Oxley Act of 2002 (SOX).

17

billions for our sample group and $1.74 billions for the control group). 17 An alternative control group would be of non-accelerated U.S. firms that were also not required to comply with Section 404. However, those companies are much smaller in size (in the scale of 1 to 60 in terms of average market value) and therefore substantially different than the complying firms. In addition our tests that are related to analysts’ forecasts are problematic for this group as only 10 percent of those small firms are covered by I/B/E/S. In contrast, the delay in compliance for foreign filers, was granted to all firms regardless of their size.

5. Test Procedures and Results 5.1 Measurement of financial reporting quality There is no single agreed upon measure of earnings quality in the literature. Rather, researchers rely on various measures that capture desirable characteristics of reporting systems (e.g. Francis, LaFond, Olsson, and Schipper, 2004; Altamuro, Beatty 200X). We apply several measures to capture changes in earnings quality that might be due to the effect of stronger internal controls on both intentional and unintentional misstatements.

We first examine changes in the absolute magnitude of abnormal

accruals; then test whether there was an improvement in the ability of earnings to predict future cash flows. Next, we check whether some of the asymmetry between reporting earnings that slightly beat analysts’ forecasts and reporting earnings that slightly miss the forecast was dissipated in the post-404 period. Finally we check for changes in the asymmetry between the use of negative and positive special items. Changes in the magnitude of abnormal accruals and the predictive power of earnings over future cash 17

Our results including all the firms are similar.

18

flows may be a result of changes in both intentional and unintentional errors, while the asymmetries in beating and missing analysts’ forecasts and use of special items are more likely due to intentional misstatement. 5.2 Accrual Quality Our first measure of earnings quality is the magnitude of absolute abnormal accruals (see Becker, DeFond, Jiambalvo, and Subramanyam, 1998; and Klein, 2002; among others). We use the absolute abnormal accruals rather than signed abnormal accruals because accruals reverse over time and because each period is of eight quarters. If for example within the first eight-quarter period a company had a large positive and a large negative abnormal accruals, and in the second eight-quarter period it had a small positive and a small negative abnormal accruals, a signed test will not detect a change in the magnitude of abnormal accruals while testing the absolute magnitude will reveal a change. While abnormal accruals can be used by managers to convey private information to investors, they can also be used as a “purposeful intervention in the external financial reporting process with the intent of obtaining some private gain” (Schipper, 1989). Many studies provide evidence of opportunistic (mis)use of abnormal accruals (application of discretion over accruals) by managers. Dechow and Dichev, 2002 suggest that abnormal accruals can also be due to managers’ estimation errors. Therefore we would interpret a reduction in absolute abnormal accruals as a sign of improvement in earnings quality that can be due to a reduction in either intentional or un intentional errors. We develop an accrual expectations model from which we calculate abnormal accruals. The model that we use combines the accrual quality model in Dechow and Dichev, 2002 with the Jones, 1991 model as suggested by McNichols, 2002. The model

19

relates changes in the firm’s working capital (ΔWC) to cash flows from operations from past (CFOq-4, CFOq-3, CFOq-2, CFOq-1), present (CFOq), and future (CFOq+1, CFOq+2, CFOq+3, CFOq+4) periods, and to contemporary changes in sales (ΔREVq). Because accruals are seasonally correlated we also add ΔWCq-4 to the model. 18 All variables are scaled by total assets at the beginning of the quarter. The model is the following: WC i , g , q   0 , g , q   1, g , q CFO i , g ,q  4   2 , g , q CFO i , g , q 3   3, g , q CFO i , g , q  2   4 , g , q CFO i , g , q 1   5 , g , q CFO i , g , q   6 , g , q CFO i , g ,q 1   7 , g ,q CFO i , g , q  2

(1)

  8 , g , q CFO i , g ,q  3   9 , g ,q CFO i , g , q  4   10 , g , q REV i , g , q   11, g , q WC i , g , q  4   i , g ,q We run this model cross-sectionally every quarter q for every industry g, classified by its 3-digit SIC code. We require a minimum of ten observations per regression. We then combine all the observations not satisfying the criterion above at the 2-digit industry level and re-run the model.

To minimize the effect of influential

observations, we remove observations that meet at least one of the following criteria: absolute DFFITS or absolute standardized residual greater than four, COOK parameter greater than one (Cook, 1977). 19

Abnormal accruals are the residuals from the

regressions. If indeed Section 404 improves the quality of financial reporting, then for complying firms we would expect a significant reduction in the magnitude of the absolute abnormal accruals. Results are reported in Table 2.

The average reduction in the

magnitude of absolute abnormal accruals from the pre- to the post-404 period for complying firms is 20.0 percent and is significant. For the Canadian firms the reduction is of 17.0 and is also significant. Overall, the reduction is 17.6 percent higher for the 18

Jones, Krishnan and Melendrez, 2008 provide evidence for the superiority of this model over other earnings management models. 19 We follow this procedure for all of the regression models in this paper.

20

complying firms. To test if the reduction in mean absolute discretionary accrual is larger for the complying firms we run the following model: AVG  WCPOST 404

i

  0   1 AVG  WCPRE 404

  2 AVG  WCPRE 404 i  COMPLY   5 AUDITORCHG

i

i

i

  3 log  TA i   4 log  BM

i

(2)

 i

|AVGΔWCPRE404| and |AVGΔWCPOST404| are the mean absolute abnormal accruals by firm

i

in the pre- and post-404 periods respectively and COMPLY is an

indicator variable that is set to 1 for complying firms and to 0 for control firms. Because larger firms tend to have smaller scaled abnormal accruals we add log TA that is the natural logarithm of the change in the firm’s total asset between the two periods to control for changes in firm size. Future growth might make firm invest in inventory to meet future demand, which will increase its accruals. We use the book-to-market ratio (BM) to proxy for future growth, and add to the model log BM that is the natural logarithm of the change in the firm’s BM between the two periods. A change from a low quality to a high quality auditor might improve the quality of the accruals of the firm and lower their absolute level. We add the indicator variable AUDITORCHG that is set to 1 if a firm switched from a non-Big-4 audit firm to a Big-4 firm, to –1 if the change was in the other direction and to 0 otherwise. The results are reported in Panel B of Table 2 and show that our coefficient of interest β2 is negative and significant (t-value = –3.41) suggesting that complying firms reduced the absolute level of their abnormal accruals significantly more than the control firms. Taken together, the results suggest that the magnitude of absolute abnormal accruals decreased significantly for complying firms and decreased at a lower rate for the

21

Canadian firms. 20 This may be because of less manipulative reporting by managers or because Section 404 procedures brought down unintentional errors and misstatements and in any case should be interpreted as an improvement in earnings quality. 21 Economic growth is somewhat lower in the Pre-404 period than in the post-404 period (2.1% vs. 3.4%). It is possible that firms incur more impairment losses in the earlier period due to the lower growth which might explain the higher accruals at that period. To address this possibility, we recalculate abnormal accruals after adding to the accruals expectation model median ROA for each industry each quarter and pool all four years together. We repeat our tests and the results (untabulated) do not change qualitatively. The pre-404 period includes a sub-period January 2002 through July 2002, which is prior to SOX. To make sure that the results are not driven by changes from the pre-302 to the post-302 period, we repeat our analysis after removing the pre-302 months from the pre-404 period and reach results that are very similar to those in Table 2. 5.3 Prediction of Future Cash Flows and Future Earnings Our second measure of earnings quality is the ability of current earnings to predict future cash flows and future earnings. Predictability is a desired attribute of earnings because it makes them value-relevant for decision-making (FASB SFAC No. 2), and the higher the predictive power the more informative are the earnings. Many studies have established the relationship between earnings and stock value (e.g., Penman and Sougiannis, 1998) reinforcing the importance of the ability to predict future earnings.

20

Results using the standard deviation of the residuals or the median of absolute accruals are qualitatively the same; to preserve space, they are not reported here. 21 Bedard, 2006 also documents a reduction in average absolute abnormal accruals in the post-404 period; however, she does not examine the differential change between complying and non-complying firms.

22

Our models relate seasoned future (four quarter ahead) cash flows and earnings before extra ordinary items (thereafter, earnings) to current earnings and to control variables as follows: CFO i , q  4 / ROAi , q  4   0   1 POST 404 i q   2 ROAi , q   3 ROAi , q * POST 404 i , q   4 LOGTAi , q   5 LOGTAi , q * ROAi , q   6 STDCFO i , q   i , q

(3)

CFO is cash flows from operations scaled by total assets at the beginning of the quarter and ROA is earnings scaled by total assets. POST404 is an indicator variable set to 1 for the post-404 period and to 0 otherwise. Under this specification, we examine if there has been a change in the prediction ability of earnings in the post-404 period regardless of whether firms comply with Section 404. We add to the model a few variables to control for other factors that might affect the predictability of future cash flows or earnings. We control for size by adding LOGTA, the natural logarithm of total assets, and LOGTA*ROA, the interaction of LOGTA and ROA because larger firms have more stable operations than smaller firms. Because we expect lower predictive ability in more volatile industries, we add the variable STDCFO, which is the standard deviation of cash flows scaled by total assets calculated each quarter at the 2-digit SIC industry code level. Our main variable of interest is ROA*POST404. If earnings in the post-404 period have higher prediction ability, then β3 will be positive and significant. The results of this specification are reported in Tables 3 and 4 under model I. In our second specification we add the indicator variable COMPLY which is set to 1 for complying firms and to 0 otherwise and also interact this variable with POST404 and with ROA.

23

CFO i ,q  4 / ROAi , q  4   0   1COMPLY i q   2 POST 404   3 ROAi , q   4 ROAi , q * COMPLY i , q   5 ROAi , q * POST 404 i , q   6 ROAi , q * COMPLY i , q * POST 404 i q (4)   7 LOGTAi , q   8 LOGTAi , q * ROAi , q   9 STDCFO i , q   i , q

Under this specification, β5 measures the change in prediction ability in the post404 period for the Canadian firms, and β6 measures the incremental predictability for the complying firms. Because the effect of Section 404 is likely to be stronger for complying firms, we expect the coefficient β6 to be positive. The results of this specification are reported in Tables 3 and 4 under model II. In the third specification we add POSTSOX, an indicator variable set to 1 for post-SOX period observations (August 2002 on) and to 0 otherwise and interact it with ROA. Under this specification ROA*POST404 measures the change in the predictive ability of earnings in the post-SOX, pre-404 period relative to the pre-SOX period and ROA*POST404 measures the incremental change beyond the change due to the passage of SOX. Model III is the following: CFOi ,q  4 / ROAi ,q  4   0  1COMPLYi q   2 POSTSOXi q   3 POST404i ,q   4 ROAi ,q   5 ROAi ,q * COMPLYi ,q   6 ROAi ,q * POSTSOXi ,q   7 ROAi ,q * POST404i ,q   8 ROAi ,q * POST404i q * COMPLYi ,q   9 LOGTAi ,q  10 LOGTAi ,q * ROAi ,q

(5)

  11 STDCFOi q   i ,q

Because we run a panel data for all our tests throughout the paper we report white standard errors which are robust to time-series cluster correlation (Peterson, 2009). The results from model I suggest that in the post-404 period the ability of earnings to predict future cash flows and future earnings increased by 7.0 percent and 13.0 percent respectively (the coefficients on ROA*Comply). The results from model II suggest that the increase is only due to the complying firms. There was insignificant change in the

24

predictive ability for the control firms (the coefficients on ROA*POST404), while there was a significant increases of 10.1 percent and 13.0 percent for the complying firms (the coefficients on ROA*POST404*COMPLY). The results also show that, prior to the implementation of Section 404, for complying firms, earnings was a better predictor of future cash flows but not of future earnings (the coefficients on ROA*COMPLY). The results from model III show an increase in the ability of earnings to predict future cash flows but not future earnings in the post-SOX, pre-404 period (the coefficients on ROA*POSTSOX have t-values of 3.32 and –0.32 respectively). For the non-complying firms, there is only a significant increase in the predictability of future cash flows but not of future earnings (the coefficients on ROA*POST404 have t-values of –1.09 and 2.08 respectively) and an incremental increase for complying firms that is

statistically significant ((the coefficients on ROA*POST404*COMPLY). The control variables LOGTA and STDCFO have the expected signs and are significant. 22 In summary, the ability of current earnings to predict future cash flows and future earnings increased significantly more for our sample firms than for the control firms. Higher earnings persistence can be due to reduction in either intentional or un intentional misstatements but can also be due to more intense use of discretionary accruals to smooth earnings. The reduction in absolute abnormal accruals we document in Section 5.2 should rule out the possibility that more earnings smoothing drives our results. 5.4 Asymmetry with Respect to Analyst Forecast Errors There is a rich literature documenting a disproportionally high frequency of firms reporting just above a financial threshold and a disproportionally low frequency just 22

We repeated these tests with prediction of one quarter out rather than four quarters out and the results were very similar

25

below the threshold (e.g. Burgstahler and Dichev, 1997). In this section we examine changes in the asymmetry between the propensity to slightly beat and to slightly miss analysts’ forecast errors. We focus on this threshold because of findings by Dechow, Richardson, and Tuna, 2003 that avoiding missing analysts’ forecasts has become the most important reporting threshold. 23 FE is the difference between the actual earnings per share (EPS) and analysts’ forecast of EPS scaled by price at the beginning of the quarter (Bartov, Givoly, and Hayn, 2002; Abarbanel and Lehavy, 2003). We define an observation as “small beat” if FE multiplied by 100 is positive but not larger than 0.1 and as a “small miss” if this number is negative but not smaller than minus 0.1. Panel A of Table 5 reports changes in the propensities of firms to slightly beat and slightly miss the forecasts in the pre- and the post-404 periods.

For complying firms the propensity to slightly beat the forecast

increased from 15.2 percent to 18.4 percent while the propensity to slightly miss the forecast increased from 5.2 percent to 6.9. Overall the ratio of small beat to small miss decreased by 10.1 percent from 2.94 to 2.65. Results are similar for the control firms. The propensity to slightly beat the forecast increased from 8.9 percent to 9.6 percent while the propensity to slightly miss the forecast increased from 5.3 percent to 6.5 and the ratio decreased 9.6 percent from 1.63 to 1.47. In Panel B of Table 5 we report results from a logit model that relates the propensities to report slightly above or slightly below the analysts’ expectations to the type of the reporting firm, and to the periods before and after the implementation of Section 404. We include control variables that might affect those propensities. We run

23

We also visually examined the thresholds of avoiding reporting a loss, and avoiding reporting an earnings decrease and they do not seem to be important benchmarks in a our sample period.

26

three specifications. Model I examines if the ratio of small positive FE to small negative FE has changed in the post-404 period. Model II additionally measures the differential change between the sample and the control firms in the post-404 period. Model III also examines the change from the pre-SOX period, to the post-SOX, pre-404 period. For brevity we present below the third and most comprehensive model. Pr ob( SMALLBEATi , q  1, SMALLMISS  0)  F (  0   1COMPLYi , q   2 POSTSOX i , q   3 POST 404 i , q   4 POST 404 i , q * COMPLYi , q   5 LOGBM i , q   6 LOGMVEi , q   7 LEVi , q   8 CFOCHGi , q   9 AUDITTYPEi , q   10 LOGNUMESTi , q

(6)

  11 SMALLBEATi , q 1   12 SMALLMISS i , q 1   13 DURi , q   14 QTR 4 i , q   i , q )

The dependent variable is set to 1 if the firm reports “small beat” and to 0 if it reports “small miss”.

We use the indicator variables POSTSOX, POST404, and

COMPLY as before. We include the natural logarithm of book-to-market LOGBM because the penalty for firms that miss an earnings forecast is more severe for growth firms (Skinner and Sloan, 2002, Ertimur, Livnat, and Martikainen, 2003), which in turn might affect the likelihood that those firms will slightly beat or miss the consensus analyst forecast. We include the natural logarithm of market value LOGMVE because firms with high market valuation might suffer larger economic loss to their shareholders for missing the earnings forecast. We include the ratio of debt-to-equity LEV because debt covenants for leveraged firms might affect their reporting incentives. LEV is also a proxy for risk, and it might be more costly for high-risk firms to miss an analysts’ forecast than for low-risk firms. CFOCHG is the seasonal change in cash flows from continuing operations. High cash flows might indicate a strong economic performance for the period, which might be a reason to slightly beat the forecast. We include an indicator variable AUDITTYPE that is set to 1 if the audit firm is a Big-4 firm and to 0

27

otherwise because of the possibility that reports of firms audited by a Big-4 firm are of higher quality (e.g., DeAngelo, 1981) and are subject to less earnings manipulation to beat (or at least not miss) the forecast. We include LOGNUMEST, the natural logarithm of the number of analysts providing earnings forecasts because the cost of missing the forecast might increase with the number of analysts following the firm due to the firm’s higher visibility.

We include the indicator variables, SMALLBEATq-1 and

SMALLMISSq-1, which are set to 1 if the firm slightly exceeded, or slightly missed the earnings forecast respectively in the prior quarter and to 0 otherwise to account for the possibility that performance relative to the analysts’ forecast in the prior period will affect the performance in the current period. We include an indicator variable DUR for industry membership in a durable goods industry because Matsumoto, 2002 shows that firms with greater reliance on implicit claims with stakeholders are more likely to take actions to avoid negative earnings surprises and DUR is used in that study to proxy for such claims. Lastly, we include an indicator variable QTR4 that is set to 1 if the observation is of the fourth quarter and to 0 for the other quarters, because the fourth quarter might differ from the other interim quarters. The incentive to achieve reporting goals might be stronger in that quarter (Dhaliwal, Gleason and Mills, 2004, Jacob and Jorgenson, 2007, Das, DShroff, and Zhang, 2007). On the other hand, because the report of the fourth quarter is audited by the firm’s external auditor it might be more difficult to achieve the reporting goals (Heflin and Hsu, 2007). Model I in Panel B shows that the coefficient on POST404 is negative and significant, suggesting a significant reduction in asymmetry between small beat and small miss in the post-404 period. In model II, POST404 measures the change in symmetry for

28

the control group and POST404*COMPLY measures the incremental change for the complying firms. Consistent with the univariate results from Panel A of Table 5, the results show that in the asymmetry is higher for complying firms (the coefficient on COMPLY is positive and significant). It also shows that the change in the asymmetry is considered separately for the two groups it becomes insignificant. Model III shows additionally that there is no significant change when all three periods are considered. Among the control variables, consistent with our expectations, LOGBM is negative and significant, suggesting that firms with a low BM ratio are more likely to slightly beat the earnings forecast. LOGMVE and CFOCHG are positive but significant. Contrary to our prediction, firms with high leverage are more likely to miss the analysts’ forecast, maybe because the greater scrutiny they receive from creditors prevents them from managing their earnings to exceed this threshold. Big 4 clients are less likely to slightly beat the forecast. It seems that slightly beating or slightly missing the forecast is a repeating pattern, as evidenced by the positive (negative) coefficient on LAGSMALL_BEAT (LAGSMALL_MISS). The coefficients on LOGNUMEST and DUR are positive and significant as expected. The indicator variable for fourth quarter observations is insignificant, probably due to the conflicting forces shaping this reporting period (higher incentives and higher constraints). In summary, the results suggest that asymmetry with respect to the consensus analyst forecasts decreased similarly for both groups in the post-404 period. All our

29

results are qualitatively the same if we also consider firms meeting the forecast and compare the ratio of beating or meeting the forecast to missing the forecast. 24

5.5 Asymmetry with Respect to Special Items Firms might attempt to affect investors’ perception of their future prospects by discretionarily re-classifying operating expenses into a “special items” category Elliott and Hanna 1996). Given that special items are viewed as transitory (e.g. Fairfield, Sweeney, and Yohn; Burgstahler, Jiambalvo and Shevlin, 2002), such action will overstate the firm’s permanent earnings and might mislead investors regarding the firm’s future cash flows. McVay, 2006, Myers, Myers, and Skinner, 2007 and Das, Shroff, and Zhang, 2007 among others, provide evidence that firms use special items to manage earnings. We use the asymmetry in special items as another measure of earnings quality and examine the ratio of firms reporting negative special items to firms reporting positive special items. Since special items can be small and immaterial, consistent with Das, Shroff, and Zhang, 2007 we examine only special items with absolute value of more than 1 percent of sales. Panel A of Table 6 reports univariate results. Negative special items are much more frequent than positive special items and the practice appears to be more common among complying firms especially in the pre-404 period. For the complying firms the ratio decreases 15.7 percent from 4.39 to 3.70 in the post-404 period and the decrease is statistically significant. For the control firms the ratio decreases 2.8 percent from 3.17 to

24

In our sample period there is no premium to meeting the forecast. To the contrary we find a negative stock returns around earnings announcement for firms just meeting the forecast which is consistent with Koh, Matsumoto, and Rajgopal (2008).

30

3.09 (statistically insignificant). 25

We interpret the asymmetry reduction as an

improvement in reporting quality mostly for complying firms. Similarly to our analysis of the propensity to beat or miss the analysts’ forecast, we develop a model for the propensity to report negative or positive special items and run the three specifications. All the control variables are the same except that we replace SMALLBEATq-1 and SMALLMISSq-1 with NEGSPIq-1 and POSSPIq-1 and drop LOGNUMEST from the model. NEGSPIq-1 (POSSPIq-1) is an indicator variable that is set to 1 if the company reported negative (positive) special items in excess of 1 percent of sales and to 0 otherwise. We also add to the model CFO q-1 because in general, economic events precede accounting recognition and special items such as write-downs might follow poor past performance (Elliott and Shaw, 1988). For brevity we present below the third and most comprehensive model. Pr ob( NEGSPIi ,q  1, POSSPI  0)  F (  0  1COMPLYi ,q   2 POSTSOX i ,q   3 POST 404i ,q   4 POST 404i ,q * COMPLYi ,q   5 LOGBM i ,q   6 LOGMVEi ,q   7 LEVi ,q   8 CFOCHGi ,q   9 CFOi ,q 1  10 AUDITTYPEi ,q  11 NEGSPIi ,q 1

(7)

 12 POSSPIi ,q 1  13 DURi ,q  14 QTR4 i ,q   i ,q )

Model I shows that the asymmetry between reporting negative and positive special items declined significantly in the post-404 period. Model II shows further a significant decrease in the asymmetry for the complying firms relatively to the control group. When we consider also the pre-302 period in model III, the results remain

25

Scaling by lagged total assets generate similar results. For complying firms the ratio decreases 17.3 percent, while it increases 3.7 percent for the control firms.

31

directionally the same but become insignificant. 26 All the control variables except for LOGBM are statistically significant. In summary, the results in Sections 5.4 and 5.5 provide some evidence for a reduction in asymmetry in reporting for complying firms. Because the asymmetry in reporting is more likely to be due to intentional misrepresentation rather than due to innocent estimation errors or unintentional misstatements, we conclude that the implementation of Section 404 brought down the use of intentional misstatements. 5.6 The Returns-Earnings Relationship The results of the tests reported in sections 5.2-5.5 show a significant improvement in the quality of earnings in the post-404 period for complying firms. If financial statements in the post-404 period are more reliable and accurate and are less subject to manipulation, then they should be a better measure of the firms’ true performance. If investors perceive the financial statements as more informative, it might lead them to place higher weight on the reported earnings in their investment decisions in the post-404 period. We use the stock market reaction to earnings surprises to measure the change in investors’ perception of earnings quality. We examine the relationship between earnings surprises relative to analysts’ forecasts and abnormal stock returns in the three days around the earnings announcement. We use the following model: ABRET i , q   0   1 FE i , q   2 FE * ABSFE i , q   3 POSTSOX i , q   4 POST 404 i , q   5 COMPLY i , q   6 FE i , q * COMPLY i , q   7 FE i , q * POSTSOX i , q   8 FE i ,q * POST 404 i , q (8)   9 FE i , q * COMPLY i , q * POST 404 i , q   10 LOGBM i , q   11 LOGMVE i , q   i , q

26

A potential reason for the weaker results in model III is the fact that the explanatory variable QTR4 is significant, and in the three-period model there is no equal number of fourth quarter across the three periods.

32

ABRET is the 3-day market adjusted stock return around the earnings announcement day. FE is defined as above and we add to the model FE*ABSFE, an interaction term between FE and the absolute magnitude of the error to allow for non-linear relations between FE and ABRET (Freeman and Tse, 1992). To control for risk factors that might affect returns, we include in the model LOGMVE and LOGBM. We report the results in Table 7. Model I tests for the change in the relationships between FE and abnormal stock returns around earnings announcements in the post-404 period regardless of firm classification. Model II examines if there is a differential change between the complying firms and the control firms. In model III we also test for change in those relationships relative to the PRE-SOX period. As expected, model I shows that stock returns are strongly associated with earnings surprises in a non-linear way as both FE and FE*ABSFE are strongly significant. The coefficient on our variable of interest FE*POST404 is positive and statistically significant implying that the relationships between FE and abnormal stock returns are stronger in the post-404 period. Under the specification of model II, the variable FE*POST404 measures the association between FE and abnormal stock returns in the post-404 period for the control group, while the variable FE*POST404*COMPLY measures the incremental association for complying firms. The coefficient FE*POST404 is negative and significant which suggests a weaker association between abnormal returns and earnings surprises for non-complying firms.

In contrast, the coefficient on

FE*POST404*COMPLY is positive and significant. This suggests that relative to our control group, there was an increase in the association between earnings surprises and

33

abnormal stock returns for the complying firms. 27 The results from model III confirm the results from model II and in addition show that there was no significant change in the relations between FE and ABRET in the post-SOX, pre-404 period. In summary the results suggest that the association between earnings surprises and abnormal returns increased for our sample firms in comparison to the control group after the implementation of Section 404, with no significant change in this association in the post SOX, pre-404 period. We conclude that Section 404 contributed to investors’ confidence in financial reporting. 28 An alternative explanation for our findings could be that the financial reports of the complying firms became timelier than those of the control group and therefore investors reacted more strongly to their reports. We check if there was a change in the length of period from the end of the quarter till the earnings announcement for the two groups. We find that for our sample firms the mean length of the period was shortened 7 percent from 41.3 to 38.4 days, but decreased 18 percent from 66.1 to 54.1 days for our control group. This suggests that the control group improved more the timeliness of its reports which rules out the alternative explanation. 6. Robustness Tests In this section we examine alternative explanations for our results. In Section 6.1 we explore the possibility that other rules and regulations rather than Section 404 drive our results. In Section 6.2 we verify that the improvement we document is not only due to firms that remediated previously reported deficiencies in their internal controls, and in

27

When we run the model only for the complying firms, the association is significantly higher in the post404 period which suggests that the stronger association is not only relative to the control group. 28 Results using beta adjusted stock returns are very similar.

34

Section 6.3 we verify that the results are not due to foreign exchange income (loss) that is more common for our control firms. 6.1 Other Rules and Regulations Because other rules and regulations were introduced around the time of our analysis, we review them to ensure that they did not influence our results. Section 401(b) of SOX directs the SEC to establish new rules regulating non-GAAP earnings disclosures.

Accordingly, in March 2003 the SEC implemented a new non-GAAP

earnings disclosures rule. Regulation G, the addition of an Item 12 to Form 8-K requires that whenever a firm, or any person acting on its behalf, discloses a non-GAAP earnings number in any public communication, it also needs to: (a) disclose the most directly comparable GAAP earnings number, (b) disclose a reconciliation of the non-GAAP number to the GAAP number, and (c) furnish, within five days, a Form 8-K containing an explanation of why management believes the non-GAAP number is useful to investors. The rule makes the difference between GAAP and non-GAAP earnings more visible and the use of non-GAAP earnings to achieve reporting goals such as meeting or beating analyst forecasts more apparent. Heflin and Hsu, 2007 find a decline in the probability of meeting or beating the forecast after the implementation of those requirements. They also find a decrease in the use of special items and an increase in the association between FE and abnormal stock returns. Because this rule was introduced in March 2003, we want to make sure that our results are not due to the new rule. We identify all the firms that during the period 2001-03 reported non-GAAP earnings at least half the time (for the process of identifying non-GAAP earnings

35

reporting, see Doyle and Soliman, 2004, pp. 14-17) and remove them from our sample. We then repeat our tests and all our results still hold. Section 301 of SOX requires each member of the audit committee to be independent. It also requires the stock exchanges to develop governance rules regarding the independence of other committees of the board of directors. The requirement to comply with Section 301 took effect in January 2004. Vafaes, 2005 finds that higher degree of independence of those committees is associated with a higher earnings quality. Chhaochharia and Grinstein, 2007 find that SOX’s Section 301 requirements benefited firms that were less compliant with those requirements prior to the passage of the Act. Using data from IRRC, we calculate the change in the percentage of independent audit committee members and of independent directors on the board of directors from the preto the post-404 period. We repeat our tests after removing those firms that increased either the independence of the board or of the audit committee by at least 5 percent and all of the results still hold. 29 6.2 Remediation of Past Weaknesses in Internal Controls ACKL, 2008 find that firms that their external auditor confirmed remediation of previously reported ICDs exhibited an increase in accrual quality relative to firm that did not remediate their control problems. To ensure that our results are not solely driven by this subsample of firms we remove 199 complying firms that improved their internal controls and repeat our tests and the results still hold. 6.3 The Effect of Foreign Exchange Income (Loss)

29

Two important accounting rules, SFAS 142 - Goodwill and Other Intangible Assets, and SFAS 144 Accounting for the Impairment or Disposal of Long-Lived Assets, took effect in fiscal years 2002 and 2001, and therefore are fully reflected in the pre- and post-404 periods.

36

Our control firms have large sales to the U.S. suggesting that they might be more exposed to changes in exchange rates. We find that 52 percent of our control firms report foreign exchange income (loss), which is much higher percentage than for our complying firms (only 12 percent). To ensure that exchange rate fluctuation is not the reason for our results, we repeat the tests of the changes in absolute abnormal accruals and in the prediction ability of current earnings after adjusting for foreign exchange income (loss) and the results remain very similar. 30 7. Conclusions In this paper we examine the effect of Section 404 of SOX on all firms that were subject to this regulation during our sample period, rather than its effect on firms reporting internal control weaknesses as other studies have done. Because of the very high compliance effort both in terms of more comprehensive external audit and additional internal resources devoted to the evaluation and correction of the internal controls over financial reporting, we expect this regulation to affect all complying firms even if their systems are not dysfunctional to the degree that those firms need to report material weaknesses. We find that complying firms improve their reporting quality in the post404 period in comparison to a control group of Canadian firms that their shares are traded on a U.S. exchange and were not required to comply with Section 404. We find the improvement on many dimensions of reporting quality including some evidence of reduction in intentional misstatements. Finally, we find that the association between earnings surprises and abnormal stock returns increased for the complying firms relative

30

COMPUSTAT’s definition of special items (item # 32) excludes Foreign exchange income (loss) and therefore we do not repeat the tests of asymmetry in special items. Similarly, analysts usually do not include exchange income (loss) in their forecasts and therefore we do not repeat the tests of asymmetry in the propensity to slightly beat or slightly miss the forecast.

37

to the control group in the post-404 period. Taken together, we provide evidence that the implementation of Section 404 helped to achieve SOX’s main goal of protecting investors by improving the accuracy and reliability of corporate disclosure. Our study should be of interest to regulators and standard-setting bodies as it provides evidence on the effectiveness of Section 404 of SOX. It also shows that Section 404 helped to restore investor confidence. Those benefits should be weighted against the implementation costs as regulators assess the costs and benefits of Section 404. Given our findings on the benefits of Section 404, Canadian regulators should consider implementing similar regulations in their country. As non-complying firms start to comply with Section 404, it will be of interest to study its effect on their reporting quality. The two-phase implementation will allow researchers to evaluate which part of Section 404 has a bigger effect on the quality of financial reporting and on investors’ perceptions: the requirement to produce a management report, or the attestation by the external auditor. We leave those questions for future study as it is yet too early to examine them now.

38

References AeA. 2005. Advancing the Business of Technology. Sarbanes-Oxley Section 404: The Section of Unintended Consequences and Its Impact on Small Business, February. Abarbanel, J., and R. Lehavy. 2003. Biased Forecasts or Biased Earnings? The Role of Reported Earnings in Explaining Apparent Bias and Over/Underreaction in Analysts’ Earnings Forecast, Journal of Accounting and Economics 36: 105-146. Altamuro, J., and A. Beatty. 2007. Do Internal Control Reforms Improve Earnings Quality? Working paper, Ohio State University. Ashbaugh, H., D. Collins, W. Kinney, and R. LaFond. 2008. The Effect of SOX Internal Control Deficiencies and their Remediation on Accrual Quality. The Accounting Review 83 (1): 217-250. Ashbaugh, H., D. Collins, W. Kinney, and R. LaFond. 2009. The Effect of SOX Internal Control Deficiencies on Firm Risk and Cost of Equity. Journal of Accounting Research 47: 1-43. Bartov, E., D. Givoly, and C. Hayn. 2002. The Rewards to Meeting or Beating Earnings Expectations, Journal of Accounting and Economics 33: 173-204. Bartov, E., and D. Cohen. 2007. The 'Numbers Game' in The Pre-and Post-SarbanesOxley Eras. Working paper, NYU. Becker, C., M. DeFond, J. Jiambalvo, and K. Subramanyam. 1998. The Effect of Audit Quality on Earnings Management. Contemporary Accounting Research 15(1): 1-24. Bedard, J., R. Hoitash, and U. Hoitash. 2009. Audit Pricing and Internal Control Disclosures among Non-Accelerated Filers. The Accounting Review Forthcoming.

39

Brown, N., C. Strohm, and A. Wompener. 2008. The Effect of Internal Control Regulation on Earnings Quality: Evidence from Germany. Working paper, University of Southern California and University of Muenster. Bryan, S., and S. Lilien. 2005. Characteristics of Firms with Material Weaknesses in Internal Control: An Assessment of Section 404 of Sarbanes-Oxley. Working paper, Wake Forest University and City University of New York. Burgstahler, D., and I. Dichev. 1997. Earnings Management to Avoid Earnings Decreases and Losses Journal of Accounting and Economics 24: 99-126. Burgstahler, D., J. Jiambalvo, and T. Shelvin. 2002. Do Stock Prices Fully Reflect the Implications of Special Items for Future Earnings? Journal of Accounting Research 40(3): 585-612 Chan, K., B. Farrell, and P. Lee. 2005. Earnings Management and Return-Earnings Association of Firms Reporting Material Internal Control Weaknesses under Section 404 of the Sarbanes-Oxley Act. Working paper, Pace University. Chang, H., J. Chen, W. Liao, and B. Mishara. 2006. CEOs’/CFOs’ Swearing by the Numbers: Does It Impact Share Price of the Firm? The Accounting Review 81(1): 127. Cheng, C., J. Ho, and F. Tian. 2006. An Empirical Analysis of Value-relevance of Disclosure of Material Weaknesses under Section 404. Working paper, Louisiana State University and University of California, Irvine. Chhaochharia, V., and Y. Grinstein. 2007. Corporate Governance and Firm Value: The Impact of the 2002 Governance Rules. Journal of Finance 62(4): 1789-1810.

40

Cohen, D., E. Dey, and T. Lys. 2005. Trends in Earnings Management and Informativeness of Earnings Announcements in the Pre- and Post-Sarbanes Oxley Periods. Working paper, NYU, University of Chicago, and Northwestern University. Cohen, D., E. Dey, and T. Lys. 2008. Real and Accrual-Based Earnings Management in the Pre- and Post-Sarbanes Oxley Periods. The Accounting Review 83(3): 757-787. Committee on Capital Markets Regulation. 2006. Interim report. November, 30. Cook, R. 1977. Detection of Influential Observations in Linear Regression. Technometrics 19: 15–18. Cox, C. 2006. Testimony Concerning the Impact of the Sarbanes-Oxley Act by SEC Chairman Christopher Cox. September 19. Das, S., P. Shroff, and H. Zhang. 2007. Quarterly Earnings Patterns and Earnings Management. Working paper, University of Illinois at Chicago, University of Minnesota, and the Ohio State University. DeAngelo, L. 1981. Auditor size and audit quality. Journal of Accounting and Economics 3: 183-199. Dechow, P., and I. Dichev. 2002. The Quality of Accruals and Earnings: The Role of Accrual Estimation Errors. The Accounting Review 77 (supplement): 35-59. Dechow, P., R. Richardson, and I. Tuna. 2003. Why Are Earnings Kinky? An Examination of the Earnings Management Explanation, Review of Accounting Studies 8: 355-384. Dhaliwal, D., C. Gleason, and L. Mills. 2004. Last-Chance Earnings Management: Using the Tax Expense to Meet Analysts’ Forecasts. Contemporary Accounting Research 21(2): 431-459

41

Donaldson, W. 2005. Testimony Concerning the Impact of the Sarbanes-Oxley Act. House Committee on Financial Services, U.S. House of Representatives. Government Printing Office. Doyle, J., and M. Soliman. 2004. Do Managers Define Street Earnings to Meet or Beat Analyst Forecasts? Working paper, Utah State University and University of Washington. Doyle, J., W. Ge, and S. McVey. 2007. Accruals Quality and Internal Control over Financial Reporting. The Accounting Review 82(5): 1141-1170. Eldridge, S., and B. Kealey. 2005. SOX Costs: Auditor Attestation under Section 404. Working paper, University of Nebraska at Omaha. Elliott, J., and J. Hanna. 1996. Repeated Accounting Write-Offs and the Information Content of Earnings. Journal of Accounting and Research 34: 135-155. Elliott, J. A. and W. H. Shaw. 1988. Write-offs as Accounting Procedures to Manage Perceptions. Journal of Accounting Research 26 (Supplement): 91-119. Engel, E., R. Hayes, and X. Wang. 2007. The Sarbanes-Oxley Act and firms’ going private decisions, Journal of Accounting and Economics, 44(1-2): 116-145. Ertimur, Y., J. Livnat, and M. Martikainen. 2003. Differential Market Reactions to Revenue and Expense Surprises. Review of Accounting Studies 8: 185–211. Ettredge, M., C, Li, and L. Sun. 2006. The Impact of SOX Section 404 Internal Control Quality Assessment on Audit Delay in the SOX Era. Auditing: A Journal of Practice & Theory 25(2): 1-23. Fairfield, P., R. Sweeney, and T. Yohn. 1996. Accounting Classification and the Predictive Content of Earnings. The Accounting Review 71: 337-355.

42

Financial Executives International. 2006. Financial Executives International survey on Sarbanes-Oxley Section 404 implementation. Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2004. Costs of Equity and Earnings Attributes. The Accounting Review 79(4): 967-1010. Freeman, R., and S. Tse. 1992. A Nonlinear Model of Security Price Responses to Unexpected Earnings. Journal of Accounting and Research 30(2): 185-209. Gao, F., J, Wu, and J. Zimmerman. 2009. Unintended Consequences of Granting Small Firms Exemptions from Securities Regulation: Evidence from the Sarbanes-Oxley Act. Journal of Accounting and Research 47(2): 459-506. Hammersley, J., L. Myers, and C. Shakespeare. 2008. Market Reactions to the Disclosure of Internal Control Weaknesses and to the Characteristics of Those Weaknesses under Section 302 of the Sarbanes Oxley Act of 2002. Review of Accounting Studies, 13(1): 141-165. Hansen, B., G. Pownall, and X. Wang. 2008. The Robustness of the Sarbanes Oxley Effect on the U.S. Capital Market. Working Paper, University of New Hampshire and Emory University. Heflin, F., and C. Hsu. 2007. The Impact of the SEC’s Regulation of Non-GAAP Disclosures. Working paper, Florida State University and Hong Kong University of Science and Technology. Iliev, P. 2008. The Effect of the Sarbanes-Oxley Act (Section 404). Working paper, Pennsylvania State University. Jacob, J., and B. Jorgenson. 2007. Earnings Management and Accounting Income Aggregation. Journal of Accounting and Economics 43: 369-390.

43

Jain, P., and Z. Razaee. 2006. The Sarbanes-Oxley Act of 2002 and Capital-Market Behavior: Early Evidence. Contemporary Accounting Research 23(3): 629-654 Jones, J. 1991. Earnings Management during Import Relief Investigations. Journal of Accounting Research 29: 193–228. Jones, K., G. Krishnan, and K. Melndrez. 2008. Do Models of Discretionary Accruals Detect Actual Cases of Fraudulent and Restated Earnings? An Empirical Analysis. Contemporary Accounting Research 25(2): 499-531. Klein, A. 2002. Audit Committee, Board of Director Characteristics, and Earnings Management, Journal of Accounting and Economics 33: 375-400. Koh, K., D. Matsumoto, and S. Rajgopal. 2008. Meeting or Beating Analyst Expectations in the Post-Scandals World: Changes in Stock Market Rewards and Managerial Actions. Working Contemporary Accounting Research 25(4): 1067-1098. Leuz, C., A. Triantis, and T. Wang. 2008. Why do firms go dark? Causes and economic consequences of voluntary SEC deregistrations. Journal of Accounting and Economics 45(1-2): 181-208. Li, H., M. Pincus, and S. Rego. 2008. Market Reaction to Events Surrounding the Sarbanes-Oxley Act of 2002 and Earnings Management. Journal of Law and Economics 51(1): 111-134. Litvak, k. 2007. The Effect of the Sarbanes-Oxley Act on Non-US Companies CrossListed in the US. Journal of Corporate Finance 13: 195-228. Lobo, G., and J. Zhou. 2006. Did Conservatism in Financial Reporting Increase after the Sarbanes-Oxley Act? Initial Evidence. Accounting Horizons 20(1): 57-73.

44

Matsumoto, D. 2002. Management’s Incentives to Avoid Negative Earnings Surprises. The Accounting Review 77(3): 483-514. McNichols, M. 2002. Discussion of the Quality of Accruals and Earnings: The Role of Accrual Estimation Errors. The Accounting Review 77 (supplement): 61-69. McVay, S. 2006. Earnings management using classification shifting: an examination of core earnings and special items. The Accounting Review 81 (3): 501-531. Morse, S. 2006. The How and Why of the New Public Corporation Tax Compliance Norm. Working paper, Santa Clara University. Myers, J., L. Myers, and B. Skinner. 2007. Earnings Momentum and Earnings Management, Journal of Accounting, Auditing, and Finance 22(2): 249-284. Nondorf, M., H. You, and Z. Singer. 2008. A Study of Firms Surrounding the Threshold of Sarbanes-Oxley Section 404 Compliance. Working paper, University of California, Berkeley, Barclay Global Investors, and McGill University. Ogneva, M., K. Raghunandan, and K. R. Subramanyam. 2007. Internal Control Weakness and Cost of Equity: Evidence from SOX Section 404 Disclosures. The Accounting review 82(5): 1255-1297. Penman, S., and T. Sougiannis. 1998. A Comparison of Dividend, Cash Flow, and Earnings Approaches to Equity Valuation. Contemporary Accounting Research 15(3): 343-383. Petersen, M. 2009. Estimating Standard Errors in Financial Panel Data Sets: Comparing Approaches. Review of Financial Studies 22(1): 435-480.

45

Public Company Accounting Oversight Board (PCAOB). 2004. An Audit of Internal Control over Financial Reporting Performed in Conjunction with An Audit of Financial Statements. Auditing Standard (AS) No. 2. Washington, D.C.: PCAOB. Russell Reynolds Associates, Inc. 2006. Governance for Good or Ill – Europe’s Chairmen Report. Schipper, K. 1989. Earnings Management. Accounting Horizons 3(4): 91-102. Skinner, D., and R. Sloan. 2002. Earnings Surprises, Growth Expectations and Stock Returns or Don’t Let an Earnings Torpedo Sink Your Portfolio. Review of Accounting Studies 7: 289-312. Tafara, E., 2006. Statement by SEC Staff: A race to the top, International regulatory reform post Sarbanes-Oxley. International Financial Law Review 12. United States Government Accountability Office. 2006. Report to the Committee on Small Business and Entrepreneurship, U.S. Senate: Sarbanes-Oxley Act — Consideration of Key Principles Needed in Addressing Implementation for Smaller Public Companies. Vafeas, N. 2005. Audit Committees, Boards, and the Quality of Reported Earnings. Contemporary Accounting Research 22(4): 1093-1122. Zhang, I. 2007. Economic consequences of the Sarbanes–Oxley Act of 2002. Journal of Accounting and Economics 44(1-2): 74-115.

46

Figure 1: Timeline of sub-period analyzed

2-Period Analysis 1/2002

12/2003

12/2005

Post – 404 (2 years)

Pre - 404 (2 years)

3-Period Analysis 1/2001

8/2002

Pre - SOX

12/2003

Post - SOX, Pre - 404

12/2005

Post - 404

47

Table 1 Classification and Industry Membership for complying and non-complying firms This table reports the compliance schedule with Section 404 of the Sarbanes-Oxley Act for various groups of firms.

Compliance Schedule with Section 404 of the Sarbanes-Oxley Act Status Management Assessment* Large Accelerated Filers December 15, 2004 Accelerated Filers December 15, 2004 Non-Accelerated Filers December 15, 2007 Large Accelerated Foreign Firms July 15, 2006 Accelerated Foreign Firms December 15, 2006 Non-Accelerated Foreign Firms December 15, 2007 * The dates refer to the first fiscal year ending after that date.

Auditor Attestation* December 15, 2004 December 15, 2004 December 15, 2009 July 15, 2006 July 15, 2007 December 15, 2009

48

Table 2 Change in Magnitude of Absolute Discretionary Accruals Panel A of this table reports changes in the mean absolute abnormal accruals for complying and Canadian firms from the pre- to the post-404 period based on a model that combines the Dechow and Dichev (2001) model and the Jones (1991) model. Abnormal accruals are the residuals from a regression model that relates current accruals (∆WC) measured as the difference between income before extraordinary items (Compustat item #8) and cash flows from continuous operations (item #108 - item #78) plus depreciation expenses (item #5) to past, present, and future cash flows from continuous operations (CFOq-1, CFOq and CFOq+1) and to change in sales (∆REV, item #2) for each industry g, every quarter q. All variables are scaled by total assets at the beginning of the quarter (# 44). The model is the follow:

WC i , g , q   0 , g , q   1, g , q CFO i , g ,q  4   2 , g , q CFO i , g , q 3   3, g , q CFO i , g , q  2   4 , g , q CFO i , g , q 1   5 , g , q CFO i , g , q   6 , g , q CFO i , g ,q 1   7 , g ,q CFO i , g , q  2   8 , g , q CFO i , g ,q  3   9 , g ,q CFO i , g , q  4   10 , g , q REV i , g , q   11, g , q WC i , g , q  4   i , g ,q Panel B reports results from a regression model that relates the firm’s mean absolute abnormal accruals in the post-404 period (|AVGΔWCPOST404|) to the mean absolute abnormal accruals in the pre-404 period (|AVGΔWCPRE404|), to an interaction variable between that variable and an interaction variable set to 1 for a sample firm and to 0 for a control firm (COMPLY) and to control variables (defined below the table). The model is the following:

AVGWCPOST 404 i   0  1 AVGWCPRE 404 i   2 AVGWCPRE 404 i  COMPLYi   3 log TAi   4 LogBM i   5 AUDITORCHGi   i

Panel A: Changes in Absolute Discretionary Accruals from the pre- to the post-404 period Group and Quarter

Complying Firms Canadian Firms P Value of difference

Pre 404 (2002-03) Mean Absolute Residual 0.0237 0.0347 9.02***

Post 404 (2004-05)

N 9,667 1,383

Mean Absolute Residual 0.0189 0.0288 9.11***

Change

N

Mean

t-value

9,709 1,168

–20.0% –17.0%

–12.84*** –3.69***

Panel B: Regression Model of Mean Absolute Discretionary Accruals Variable Coefficient t-value

Intercept 0.011*** (12.41)

|AVGΔWCPRE404| 0.581*** (14.00)

|AVGΔWCPRE404| *COMPLY –0.140*** (–3.41)

LogΔTA

LogΔBM

AuditorChg

–0.001*** (–3.62)

0.003 (1.45)

–0.003 (–1.44)

29.6% Adjusted R2 N 1,406 * ** *** , , denotes significance at 10%, 5%, and 1%, respectively. LogΔTA = The natural logarithm of the change in total assets (item # 44) from the pre- to the post404 period LogΔBM = The natural logarithm of the change in the ratio of book value of equity to market value of equity (item # 60/ (item # 14 * item # 61)) from the pre- to the post-404 period AUDITORCHG = An indicator variable that is set to 1 if a firm switched from a non-Big-4 auditor to a Big-4 auditor, to –1 if the change was the other way around and to 0 otherwise

49

Table 3 Change in the Ability of Current Earnings to Predict Future Cash This table reports coefficient estimates from a cross-sectional regression relating future seasoned cash flow (CFO) to current earnings before extraordinary items (ROA, item # 8), to an indicator variable set to 1 for complying firms and to 0 otherwise (COMPLY), to an indicator variable set to 1 for observations in the post-Section 404 period and to 0 otherwise (POST404), to interaction variables between those variables, and to control variables (defined below the table). Model I reports the results pooling complying firms and control firms together, while model II separates the two groups. Model III adds an indicator variable that is set to 1 for observations in the post-SOX period and to 0 otherwise. Model III (the most comprehensive) is as following:

CFOi ,q  4   0  1COMPLYi q   2 POSTSOXi q   3 POST 404i ,q   4 ROAi ,q   5 ROAi ,q * COMPLYi ,q   6 ROAi ,q * POSTSOXi ,q   7 ROAi ,q * POST404i ,q   8 ROAi ,q * POST 404i q * COMPLYi ,q   9 LOGTAi ,q  10 LOGTAi ,q * ROAi ,q  11 STDCFOi q   i ,q

Regression Model of Cash flow Prediction Variable Intercept COMPLY POSTSOX POST404 ROA ROA*COMPLY ROA*POSTSOX ROA*POST404 ROA*POST404*COMPLY LOGTA LOGTA*ROA STDCFO

Model I Coefficient t-stat –0.009*** –3.75

Model II Coefficient t-stat –0.013*** –4.46 0.004*** 3.87

–0.003 0.346***

–1.03 22.94

–0.004 0.337*** 0.062**

–1.11 15.54 2.26

0.070***

2.99

0.004*** 0.013** –0.005***

17.09 2.17 –9.16

0.010 0.101*** 0.004*** 0.001 –0.005***

0.39 2.81 16.12 0.20 –9.41

Model III Coefficient t-stat –0.016*** –7.53 0.004*** 3.88 0.006* 1.77 –0.005 –1.29 0.230*** 4.71 0.031 1.27 0.183*** 3.32 –0.026 –1.09 0.158*** 4.83 0.004*** 11.78 –0.012** –2.32 –0.001 –0.94

Adjusted R2 37.7% 37.7% 29.7% N 25,494 25,500 35,932 * ** *** , , denotes significance at 10%, 5%, and 1%, respectively. LOGTA = The natural logarithm of total assets (item # 44) STDCFO = Standard deviation of cash flows from operations scaled by total assets calculated at the 2-digit SIC industry code All other variables are defined at the top of this table.

50

Table 4 Change in the Ability of Current Earnings to Predict Future This table reports coefficient estimates from a cross-sectional regression relating future seasoned ROA to current ROA, to an indicator variable set to 1 for complying firms and to 0 otherwise (COMPLY), to an indicator variable set to 1 for observations in the post-Section 404 period and to 0 otherwise (POST404), to interaction variables between those variables, and to control variables (defined below the table). Model I reports the results pooling complying firms and control firms together, while model II separates the two groups. Model III adds an indicator variable that is set to 1 for observations in the post-SOX period and to 0 otherwise. Model III (the most comprehensive) is as following:

ROAi ,q  4   0  1COMPLYi q   2 POSTSOXi q   3 POST404i ,q   4 ROAi ,q   5 ROAi ,q * COMPLYi ,q   6 ROAi ,q * POSTSOXi ,q   7 ROAi ,q * POST 404i ,q   8 ROAi ,q * POST 404i q * COMPLYi ,q   9 LOGTAi ,q  10 LOGTAi ,q * ROAi ,q  11 STDCFOi q   i ,q

Regression Model of Earnings Prediction Variable Intercept COMPLY POSTSOX POST404 ROA ROA*COMPLY ROA*POSTSOX ROA*POST404 ROA*POST404*COMPLY LOGTA LOGTA*ROA STDCFO

Model I Coefficient t-stat –0.023*** –14.88

Model II Coefficient t-stat –0.027*** –16.09 0.006*** 4.58

–0.002** 0.493***

–2.64 22.25

–0.002*** 0.515*** –0.002

–3.19 17.41 –0.05

0.130***

5.87

0.004*** –0.006 –0.003***

19.57 –1.09 –9.82

0.050 0.130*** 0.004*** –0.014** –0.003***

1.44 3.78 16.27 –2.10 –9.96

Adjusted R2 53.5% 53.6% N 25,453 25,451 * ** *** , , denotes significance at 10%, 5%, and 1%, respectively. All variables are defined in previous tables.

Model III Coefficient t-stat –0.033*** –19.53 0.006*** 4.91 0.006*** 3.61 –0.002*** –3.54 0.542*** 23.27 0.035 1.10 –0.007 –0.32 0.070** 2.08 0.117*** 4.00 0.004*** 21.92 –0.028*** –4.49 –0.001** –2.31 49.8% 35,834

51

Table 5 Changes in Frequency of Reporting Small Positive, and Small Negative Analysts’ Forecast Errors Panel A of this table reports changes in analysts’ small forecast errors from the period before the implementation of Section 404 to the period thereafter. Panel B shows the results of a logit model that relates an indicator variable that is set to 1 for a small positive analysts’ forecast error and to 0 for a small negative forecast error to an indicator variable set to 1 for the complying firms and to 0 otherwise (COMPLY), to an indicator variable set to 1 for observations in the post-Section 404 period and to 0 otherwise (POST404), to interaction variables between those variables, and to control variables (defined below the table). Model I reports the results pooling complying firms and other firms together, while model II separates the two groups. Model III adds an indicator variable that is set to 1 for observations in the postSOX period and to 0 otherwise. Model III (the most comprehensive) is as following:

Pr ob( SMALLBEATi ,q  1, SMALLMISS  0)  F (  0   1COMPLYi ,q   2 POSTSOX i ,q   3 POST 404 i ,q   4 POST 404 i ,q * COMPLYi ,q   5 LOGBM i ,q   6 LOGMVEi ,q   7 LEVi ,q   8 CFOCHGi ,q   9 AUDITTYPEi ,q   10 LOGNUMESTi ,q   11 SMALLBEATi ,q 1   12 SMALLMISS i ,q 1   13 DURi ,q   14 QTR 4 i ,q   i ,q )

Panel A: Frequency of Reporting Small Positive, and Small Negative Analysts’ Forecast Errors Item Before Section 404 Small Beat Small Miss Beat/Miss Ratio After Section 404 Small Beat Small Miss Beat/Miss Ratio

Complying Firms

Canadian Firms

15.2% 5.2% 2.94

8.9% 5.3% 1.63

18.4% 6.9% 2.65

9.6% 6.5% 1.47

Ratio Change from Before to After Section 404 Beat/Miss Ratio –10.1%***

–9.6%*

52

Panel B: Logit Model for the Changes in the Reporting Frequency of Small Positive and Negative Analysts’ Forecast Errors Variable Intercept COMPLY POSTSOX POST404 POST404* COMPLY LOGBM LOGMVE LEV CFOCHG AUDITTYPE LOGNUMEST SMALLBEAT SMALLMISS DUR QTR4

Model I Coefficient P-Value 0.375** 0.03

Model II Coefficient P-Value 0.223 0.41 0.340** 0.03

–0.142**

0.04

–0.598*** 0.035 –0.528*** 0.554 –0.225** 0.248*** 0.183*** –0.174** 0.114** 0.014