Court Stringency and Voluntary Restatements

Transcription

1 Court Stringency and Voluntary Restatements C.S. Agnes Cheng School of Accounting & Finance The Hong Kong Polytechnic University Kowloon, Hong Kong Tel: (852) , Henry He Huang SySyms School of Business Yeshiva University New York, NY Tel: (832) , Zhen Lei School of Accounting & Finance The Hong Kong Polytechnic University Kowloon, Hong Kong Tel: (852) , Haitian Lu * School of Accounting & Finance The Hong Kong Polytechnic University Kowloon, Hong Kong Tel: (852) , haitian.lu@polyu.edu.hk This version: Feb, 2018 Agnes Cheng thanks Hong Kong Government Research Grant Council - General Research Fund (#590213), and Haitian Lu thanks the National Natural Science Foundation of China (# ) for the support of this project. * Corresponding author. Tel.: ; Fax: address: haitian.lu@polyu.edu.hk (Haitian Lu).

2 Court Stringency and Voluntary Restatements Abstract This paper studies how district court stringency affects firms propensity to admit their misreporting using different samples of misstated firms. We find robust evidence that firms headquartered in high dismissal rate (lenient) court jurisdictions are more likely to make voluntary restatements. Exogenous shock from Supreme Court s Tellabs decision confirms this effect. We consider a range of explanations and find the results most easily explained by managers using court dismissal rates as a heuristic when making voluntary restatement decisions. Our evidence shed light on the complicated effect of legal environment on financial reporting quality. Key words: Court stringency, Accounting misreporting, Securities lawsuit, Restatement JEL Classification Code: M41; K22; G39

3 1. Introduction The effect of legal environment on financial reporting quality is an important yet debated question (Baginski, Hassell and Kimbrough 2002; Field, Lowry and Shu 2005; Rogers and Buskirk 2009; Donelson et al. 2012). Prior work shows managers face asymmetric incentives to disclose good news but withhold bad news (Basu 1997; Watts 2003; Kothari, Shu, and Wysocki 2009). Theories of litigation risk suggest that higher perceived legal penalty could, on one hand, deter non-compliance and prompt pre-emptive disclosure (Skinner 1994, 1997). On the other hand, they might be counterproductive, as they instill extreme fear of punishment and make disclosure of wrongdoings unlikely (Pfarrer et al. 2008). Empirically, the average effect of stronger legal environment on corporate (bad news) disclosure remains unclear. To shed light on this question, we study the voluntary restatement decision of misreporting firms. When firms discover accounting misstatements, the federal securities law requires immediate correction through restatements. 1 In practice, however, due to the catastrophic consequences of restatements including lawsuits 2, voluntary restatements are rare, and many companies keep strategic silence, gambling that subsequent events would allow them to conceal the accounting mistakes. The systematic under-correction of accounting mistakes can lower investors confidence in the market. To appreciate its magnitude, Figure 1 shows among the 4,085 defendants of securities lawsuits from 2001 to 2013, only 197 (4.8%) made restatements before 1 The SEC has ruled that there is a duty to correct statements made in any filing if the statements either have become inaccurate by virtue of subsequent events or are later discovered to have been false or misleading from the outset, and the issuer knows or should know that persons are continuing to rely on all or any material portion of the statements (Sec. Act. Rel. 6084, 17 SEC Dock. 1048, 1054 (1979)). The FASB (2005) ASC Topic 250, Accounting Changes and Error Corrections, states, Any error in the financial statements of a prior period discovered after the financial statements are issued shall be reported as an error correction, by restating the prior-period financial statements. See also Accounting Principles Board Opinion 20; Statement of Financial Accounting Standards (SFAS) No. 16; and SFAS No. 154 (issued in May, 2005), among others. 2 These consequences include, for example, negative market responses (Palmrose, Richardson, and Scholz, 2004), increased cost of capital (Hribar and Jenkins, 2004), management turnover (Collins et al. 2009) and the resultant securities lawsuits (Francis, Philbrick and Schipper, 1994).

4 the class-end (truth revelation) date. Another 383 (9.4%) made forced restatements after being sued, and the rest 3,505 (85.8%) never restated their financials despite being sued. On the other hand, the regulators, knowing they are constrained by resources, explicitly encourage and reward self-reporting of wrongdoings (SEC Seaboard Report 2001) 3. In this paper we ask: Does stronger legal environment induce more voluntary restatements? To examine the impact of legal environment on voluntary restatements, we take novel approach by exploiting the variation on the stringency of the district court under which firms are headquartered. In the U.S., the federal securities law requires all private securities litigations to be heard first at a federal district court (USDC), which has jurisdiction over a number of counties. Though plaintiffs can file litigation in any USDC where the defendant firm has a place of business, multiple filings must be consolidated into one case typically heard by the district court where the defendant firm is headquartered. Court decisions on whether to dismiss a shareholder litigation are made by randomly assigned district judges (Bird, 1975; Galasso and Schankerman, 2015), causing exogenous variation in dismissal rates across courts and over the time. 4 Higher dismissal rate means firms headquartered in the jurisdiction of particular USDC are subject to more lenient litigation environment. Our litigation environment measure has advantages over the rule of law indicators used prior cross-country studies based on World Bank data (La Porta et al. 2006; Kaufmann et al. 2003; Srinivasan, Wahid, and Yu 2015) by circumventing country-specific idiosyncrasies that affect our 3 See the SEC s Report of Investigation Pursuant to Section 21(a) of the Securities Exchange Act of 1934 and Commission Statement on the Relationship of Cooperation to Agency Enforcement Decision (Seaboard Report). Oct 23, 2001, which explained how self-reporting, cooperation, self-policing, and remediation factor into SEC decisions when considering enforcement actions. 4 Another indicator of litigation environment is the incidence of litigation (Donelson et al. 2012), defined by the number of filed cases in a district court scaled by the number of headquartered firms under the court s jurisdiction. We controlled for this variable in all our regressions.

5 variables of interest. Specifically, it captures within-country heterogeneities in securities law enforcement, holding the country s regulatory environment constant. Moreover, our measure is at court-level and decided by randomly assigned district judges, which is unlikely affected by individual firm characters. Finally, it is hard to perceive that firms choose or switch their headquartering place based on district court dismissal rates 5. Note that litigation environment can simultaneously affect both firms likelihood to commit accounting misstatement, and their propensity to make voluntary restatements. Our paper focuses on the latter and investigates a sample of firms known to have made financial misstatements (identification method described in Section 3.1). Among these firms, some made voluntary restatement, that is, restatement before revelation by external agencies. Others either made forced restatements, or never make restatements. We then ask whether misstated firms headquartered in more stringent district court jurisdictions are more or less likely to make voluntary restatements. Why should firms care about their home court stringency when making voluntary restatement decisions? We argue that firms litigation risk heightens when prior misstatement came to knowledge of the management. In securities litigations, voluntary restatements can be taken either as evidence of misconduct that are unfavorable to the firm, or as mitigating factors that help to weaken the claim that managers withheld bad news to keep price distorted, and reduce potential damages by shortening the class period and number of affected class shareholders. Central to this discretion is the stringency of the firm s home court. Court s dismissal rate, in this regard, provides managers with a heuristic on the stringency of the court. For example, suppose 5 In our sample, the average years of firms staying in the jurisdiction of one district court since 1996 is 8.2 years, which exceeds the five-year period that we use to calculate court stringency. On firms relocation decision, there are only 1.9% of all the Compustat firm-years relocated out of the jurisdiction of former district court from 1996 to The frequency of headquarter relocations is low and can hardly affect our results.

6 the district court of North Carolina has the reputation of dismissing 80% of shareholder litigations, whilst the district court of South Carolina dismissed only 20%. That indicates firms headquartered in the jurisdiction of district court of South Carolina are under more stringent litigation environment. Our research design exploits this variation on district court stringency to test its effect on firms voluntary restatement propensity. Drawn on the largest sample of shareholder litigations from 2000 to 2014 and restatement records from Audit Analytics (AA), we find misreporting firms headquartered in high dismissal rate (or lenient) courts are more likely to make voluntary restatements, suggesting a negative impact of perceived legal stringency on voluntary disclosure of wrongdoing. The effect is both statistically significant and economically large: One standard deviation increase (decrease) in court dismissal rate leads to % increase (9.6-14% decrease) in the likelihood of voluntary restatements. Using alternative samples of misstating firms, and having controlled for a battery of variables in literature including state fixed effect, the result remains robust. Two pieces of evidence suggests the positive correlation between voluntary restatement and court dismissal rate is not an artifact of measurement error or other randomness. First, we find firms are more sensitive to more recent dismissal rates than remote ones. Second, the effect is more pronounced for firms headquartered in less busy than busy courts. Both evidence are consistent with the availability heuristic theory in social psychology, which states that agents have tendencies to overestimate the likelihood of events with greater availability in memory, which can be influenced by how recent the memories are, or how unusual or emotionally charged they may be (Sherman and Corty 1984; Strack 1985; Schwarz et al. 1991). Our causal evidence comes from the landmark case of Tellabs v. Makor (hereafter Tellabs ). Tellabs represents one of the Supreme Court s first effort to clarify the strong inference standard

7 of Scienter, a core element to plead securities litigations. The effect of Tellabs is that it homogenizes federal courts pleading standards (Choi and Pritchard, 2012), thus tightens the stringency of district courts under previously lenient Circuits. Using difference-in-differences, we first find that after Tellabs, district courts under previously lenient Circuits decreased their dismissal rates relative to the control courts. Next, we show misstated firms under pre-tellabs lenient courts also decrease their voluntary restatement propensity after Tellabs relative to the control firms. This evidence suggests that misstated firms deliberately alter their restatement policy in response to (exogenous) changes in court stringency. Our results suggest that misreporting firms actively consider their home court stringency when choosing their restatement policy. Whether this strategy works, in the sense that it avoids shareholder litigations or helps to reduce the legal and market penalties, is clearly a tougher question to answer. This is because we do not observe what the legal and market consequence would be if the voluntary restating firms chose to keep silent. Instead, we take indirect approach by comparing the outcomes of voluntary restating firms headquartered in lenient and stringent courts, relative to their control firms. Interestingly, our event study on defendant firms reveals no significant differences of market reactions on the class-end date and case-filing date between voluntary restating and control firms headquartered in lenient or stringent courts. On the other hand, we find no evidence that voluntary restating firms in lenient courts enjoy better court advantages in terms of dismissal outcomes, settlement amount, or unconditional cost amount. Taken together, our tests suggest that firms response to court stringency is not more psychological than based on rational economic analysis. We see three contributions of this paper to the literature. First is our study on voluntary restatements as the main construct. The large accounting and finance literature tends to use

8 restatement frequency to infer the magnitude of financial misreporting (Burns and Kedia 2006; Efendi et al. 2007). However, As Srinivasan, Wahid, and Yu (2015) note, while lower level of restatements can represent an absence of errors, it can also indicate a lack of detection and disclosure. Given a large number of misstatements are concealed, it is important to understand how institutional factor contributes to firms voluntary restatements. Though this paper is not the first to differentiate voluntary and forced restatements (Pfarrer et al. 2008; Marciukaityte, Szewczyk, and Varma 2009), to our best knowledge it presents by far the largest and cleanest samples of misstated firms that made either voluntary, forced, or no restatements. Second, we add to the understanding of how legal environment affect firms voluntary disclosure. In the accounting literature, the question why firms voluntarily disclose bad news was asked in important work including Skinner (1994, 1997), Kothari, Shu, and Wysocki (2009), Rogers and Buskirk (2009), and Donelson et al. (2012), etc. The consensus is that firms actively consider litigation risk when making bad news disclosure such as earnings warnings (Field, Lowry and Shu 2005) and management forecast (Baginski, Hassell and Kimbrough 2002). Specifically on restatements, Srinivasan, Wahid, and Yu (2015) find U.S. listed firms headquartered in weak rule-of-law countries have low restatement frequency than firms from strong rule-of-law countries despite higher earnings management levels. Our paper, however, focuses on firms voluntaryrather than overall restatement frequency. Other than legal environments, Lin and Huang (2015) find both firm-level managerial incentives (CEO tenures) and governance characteristics (board independence) affect voluntary restatements. Finally, Pfarrer et al. (2008) show firms voluntary restatements are positively induced by the behavior of industry leaders and peer groups, but negative induced by formal sanctions.

9 Last but not the least, we enrich the law and accounting literature by proposing novel, court-based measure on firms legal environments. Prior work focuses on the laws on paper and implicitly assume that they are enforced with full strength. 6 Unlike these studies, we focus on the role of courts as law enforcers, and the resultant variations in litigation environment on headquartering firms. In economics literature, court and judge stringencies are used to study, for example, the effect of incarceration on individual s earnings prospect (Kling, 2006), patent protection on corporate innovation (Galasso and Schankerman, 2015), and bankruptcy rules on lending behaviors (Dobbie, Goldsmith-Pinkham, and Yang, 2016), etc. We extend this approach to examine the effect of court stringency on firms restatement policy. The rest of this paper proceeds as follows: Section 2 describes the institutional setup relevant to our analysis. Section 3 presents sample selection, data and descriptive statistics. Section 4 presents empirical results. Section 5 presents robustness tests. Section 6 concludes. 2. Restatements and Litigation Environment in the U.S The Restatement Process The observation of a restatement (voluntary or forced) is a joint outcome of two stages: First, firms made misstatements in financial reporting that involve accounting errors or irregularities. Second, upon discovery, the management make the decision on whether, when, and how to issue a restatement. This paper explicitly focuses on the second stage. As Palmrose, Richardson, and Scholz (2004) illustrate, the company, the SEC, an independent auditor or a 6 For example, many papers use staggered adoption of anti-takeover (Business Combination) laws by U.S. states as shocks to corporate governance (Garvey and Hanka, 1999; Bertrand and Mullainathan, 2003; Wald and Long, 2007; Atanassov, 2013).

10 combination thereof, can detect misreporting. 7 Upon discovery of misstatements, the management face statutory duty to take corrective restatements 8. However, the decision is often strategic. Some firms make preemptive restatement before revelation by external agencies, some firms make forced restatements after being caught up; others never make restatements. Conditional on issuing restatements, the mediums of report can differ (Files, Swanson, and Tse, 2009). Some restatements are reported in press release or series of press releases, some are in the Form 8-K filings with the SEC, and some by the filing of amended financials (10-Ks). The information provided in such disclosure, such as accounting issues involved and circumstances underlying the restatement, as well as specificity level, also vary (Palmrose, Richardson, and Scholz 2004) Securities Lawsuit as Deterrence Mechanism As public enforcement by the SEC is resource constrained, in the U.S., the private securities lawsuit pursuant to the SEC 10b-5 anti-fraud provision plays unique roles in compensating victims, and deterring frauds (See Habib et al for a review). The rule prohibits any fraud or deceit in connection with the purchase or sale of any security. The plaintiff under this rule are shareholders, and the defendants include the firm and any person involved in the fraudulent activity. Successful pleading of a securities lawsuit depends on the plaintiff s evidence on each of the legal elements: 7 For example, the company can identify misstatements through internal audits and other internal control procedures, such as period-end closing processes, policy reviews, and mechanisms that solicit and investigate complaints from employees. Occasionally, the SEC requests a restatement after reviewing company filings. Finally, when auditors discover that previously issued financial statements contain misrepresentations, GAAS requires that they advise the client to make appropriate disclosures, and to take the necessary steps to ensure this occurs (AICPA, 2002, Section AU 561). 8 The SEC has ruled that there is a duty to correct statements made in any filing if the statements either have become inaccurate by virtue of subsequent events or are later discovered to have been false or misleading from the outset, and the issuer knows or should know that persons are continuing to rely on all or any material portion of the statements (Sec. Act. Rel. 6084, 17 SEC Dock. 1048, 1054 (1979)). The FASB (2005) ASC Topic 250, Accounting Changes and Error Corrections, states, Any error in the financial statements of a prior period discovered after the financial statements are issued shall be reported as an error correction, by restating the prior-period financial statements. See also Accounting Principles Board Opinion 20; Statement of Financial Accounting Standards (SFAS) No. 16; and SFAS No. 154 (issued in May, 2005), among others.

11 (1) Materiality; (2) Misrepresentation or omission; (3) Scienter (defendant s fraudulent intent or recklessness); (4) Reliance; (5) Causation; and (6) Damages. It is worthwhile to illustrate below the typical stages of a securities class action lawsuit. In the first stage, a plaintiff files a lawsuit and asks to be the lead plaintiff. A 60-day clock for any individual or entity to file paperwork with the court asking to be the lead plaintiff is triggered when the first securities lawsuit is announced. After the deadline, the court reviews all pleadings and appoints the lead plaintiff and lead counsel. In the second stage, plaintiffs counsel files their amended consolidated complaint, and the defendants then have a deadline to file their motion to dismiss. A motion to dismiss is essentially an argument by the defendants that, even if all facts alleged in the complaint were true, they are insufficient to give rise to liability under SEC Rule 10b-5. The court then decides, based on both plaintiff s complaints and defendant s motions, whether to uphold plaintiff s Rule 10b-5 claim. If yes, the court enters an order denying defendant s motion to dismiss, which then gives class plaintiffs the right to obtain discovery from the defendant. Passing the motion to dismiss is the pivotal stage in securities lawsuits, for the costs of litigation increases substantially if the plaintiffs claim is not dismissed by court. Because almost all cases end up either dismissed or settled, prior literature uses whether the plaintiff passes the motion to dismiss as proxy for win (Choi, 2007; Choi, Nelson, and Pritchard, 2009; Dyck, Morse, and Zingales, 2010). If plaintiff survives the motion to dismiss then it enters the third stage of discovery. Discovery typically involves requests for document production, admissions, and depositions of officers, employees, experts and third parties. Once completed, plaintiff must seek class certification. If granted by court, the case officially becomes a securities class action. At this point, defendants can face great liability if the case goes to trial and the often likely outcome is a settlement.

12 In the final stage, the plaintiff and defendant s attorney often negotiate a settlement. The settlement must seek court s preliminary and final approval. Once approved, the claims administrator takes over to receive the settlement fund, sends out court-approved notices to the investor class, receives and processes the claims and distributes the settlement funds. The whole process takes a typical 3-4 years to complete Court Jurisdictions and Splits in Pleading Standards Under the Securities and Exchange Act, federal courts are given exclusive jurisdiction to hear securities lawsuits. 9 There are 94 district courts (5 outside the main territory), 13 circuit (appellate) courts, and the Supreme Court throughout the country. Each district court has geographical jurisdiction over a number of counties. 10 All federal judges receive appointment by the President and have lifetime tenure. Each district court has at least one judge, some busy courts such as Southern District of New York and Central District of California each has 28 judges. The assignment of cases to federal judges is on a rotational or, more often, random basis (Bird, 1975; Galasso and Schankerman, 2015). 11 Appeals against district court rulings go to its corresponding circuit court. There are twelve appellate courts dividing the country into different circuits 12. The 9 th circuit court in California, for example, has 29 appellate judges, overseeing 13 district courts. Circuit courts in the U.S. are influential lawmakers for their ability to set legal 9 See Section 27 of the 1934 Securities Exchange Act. 10 For geographical jurisdiction of federal district courts, see PACER: 11 Though there might be a need to assign more specialized and complex cases to more experienced judge, to implement a program that would attempt to assign cases according to the relative abilities of the judges in a district is understandably unpopular (Bird, 1975, pp. 483). Moreover, many courts see a danger in fostering judge specialization, because if certain judges in a district become experts to whom cases in particular areas of the law always would be assigned, it deprives other judges of the opportunity, provided by random selection, to gain expertise in that legal area. See Bird (1975). 12 The thirteenth court of appeals is the United States Court of Appeals for the Federal Circuit, which has nationwide jurisdiction over certain appeals based on their subject matter.

13 precedent with minimal supervision by the Supreme Court. 13 Figure 2 visualizes the geographical jurisdiction of federal district courts and their corresponding circuit courts. Colors and numbers indicate their average dismissal rate and standard deviation over our sample period. Attorneys, commentators and scholars have long recognized the divergent pleading standards among courts in securities lawsuits. The split centers on the pleading standard of Scienter, a core legal element to plead a 10b-5 claim. The element of Scienter requires plaintiffs to state with particularity facts giving rise to a strong inference that the defendant acted with the required state of mind. 14 It is well known that hard evidence of Scienter is difficult to obtain prior to discovery, and in practice whether plaintiff s evidence can satisfy Scienter depends largely on the pleading standard of the relevant court. For example, drawn from representative Circuit Court decisions legal scholars divide federal courts into three groups: The 1st, 4 th, 6 th, and 9 th circuits adopted a preponderance standard which is pro-defendant firms; the 2 nd, 8 th, 10 th and 11 th circuits adopted an equal inference standard positioned in the middle; the 3 rd and 7 th circuits adopted a reasonable person standard which is pro-plaintiffs (Choi and Pritchard 2012). 3. Sample and Data 3.1. Data and Sample Distribution We purchased the Securities Class Action Services (SCAS) database from RiskMetrics Institutional Shareholder Services (ISS) to identify all securities lawsuits filed in federal courts This is particularly the case for securities lawsuits: On average, securities cases make up less than 1% of Supreme Court s docket, or about 1.5 cases per year, making circuit courts the de facto final arbiter (Pritchard, 2011). 14 See Exchange Act 21D(b)(2), 15 U.S.C. 78u-4(b)(2). 15 Another (free) database popular for securities lawsuit study is the Stanford Law School Securities Class Action Clearinghouse (SCAC) database. However, as Karpoff et al. (2017) observe, the filing date on the SCAC database postdates the time at which investors first learn of the purported misconduct that triggers the litigation by an average of 150 calendar days. Since the identification of our sample relies crucially on the key event dates, we purchase the commercial database whose primary purpose is to assist institutional investors that have a claim in securities lawsuits, and supplement any missing variables using SCAC.

14 The SCAS database is used in accounting and finance studies such as Cheng et al. (2010) and Donelson et al. (2012). It offers detailed portfolio views of securities lawsuits including plaintiffs, defendants, court, allegations, class periods, claim deadline dates, claims administrator details and pertinent related data since To make up any missing values in the SCAS, we hand collect additional data from the Stanford Law School Securities Class Action Clearinghouse (SCAC) database. To merge with Compustat records, we hand collect the GVKEY from Compustat to each lawsuit case. The two data cleaning procedures leave us with 3,363 unique cases with valid GVKEY and non-missing data on federal filing date, dismissal date or settlement date (if the case is closed), class-start date, class-end date and allegations. For the construction of our court dismissal rate, we utilize 6,976 unique cases with valid case name, federal filing date and case status information from both SCAS and SCAC. The restatement data come from Audit Analytics (AA). For restatement identification, we exclude firms labelled in AA as Res Clerical Errors since we are interested in accounting irregularities. All financial statement variables are from Compustat, and the stock trading data come from CRSP. Our objective is to identify a group of firms that made accounting mistakes, some made voluntary restatements and others kept strategic silence. In practice, whether or not a firm has misreported earnings can only be identified through evidences ex post such as firms own restatements, the SEC sanctions, or court trial outcomes. As the SEC is resource constrained, and most securities litigations end up with settlements rather than trial, this paper employs the following methods to identify misreporting firms. Sample of Defendant Firms with alleged both US GAAP and Rule 10b-5 violations

15 Our first sample of misstated firms contains those sued by shareholders in securities litigation (hereafter Sample of Defendant Firms or SDF ). Some of the defendant firms made restatement before the class-end date, which is the date when the corrective disclosure that triggers the lawsuit was revealed to the market (Kellogg, 1984; Griffin, Grundfest, and Perino, 2004; Gande and Lewis, 2009). We identify these firms as voluntary restating firms. For the control sample, we further screen the remaining defendant firms, and require the causes of action to include both US GAAP violations and Rule 10b-5 violations. Appendix B.1 summarizes our screening process. For voluntary restating firms, we start with 3,363 lawsuits whose GVKEY, federal filing date, class periods and dismissal date (settlement date) are identifiable, and 11,377 restatement records merged from Audit Analytics (AA) restatement database and Compustat annual financial database. By matching the lawsuit class periods and restating periods, we obtain 789 non-duplicated defendant firms with non-error-based restatement. We then identify voluntary restating firms to be those that make restatement before class-end date, and obtain 275 observations. For control firms, we merge 3,363 lawsuits with Compustat annual financial database and obtain 3,175 non-duplicated defendant firms. Excluding those with voluntary restatement, we obtain 2,921 observations without voluntary restatement. We then require our defendant firms to be alleged of both GAAP violations and Rule 10b-5 violations, leaving 928 observations. Eliminating observations without valid variables in our tests and requiring fiscal-end date to be between December 31, 2000 to December 31, 2013, we obtain a final sample comprising 393 defendant firms from 2001 to 2013, with 111 voluntary restating firms and 282 control firms without voluntary restatement.

16 Sample of Restating Firms: Defendant firms with restatements The second sample of misstated firms contains defendant firms with (voluntary or forced) restatements (hereafter Sample of Restating Firms or SRF ). Our voluntary restating firms are the same 111 with the SDF sample. For control firms, we require our defendant firms to have made corresponding restatement on or after the class-end date (i.e. forced restatements). Intuitively, forced restating firms constitute the best counterfactual group, for they indicate these firms ought to, but did not make voluntary restatement. Our final SRF comprises 300 observations from 2001 to 2012, with 111 voluntary restating firms and 189 control firms. Appendix B.2 summarizes the screening process for SRF. Some of the defendant firms also receive the SEC sanctions. To ensure that our 111 voluntary restating firms are indeed voluntary restaters, we further compare their restatement filing date with the SEC enforcement date (if any), which we obtain from the SEC s Accounting and Auditing Enforcement Releases. We find 8 related SEC enforcements, but none is before the restatement date. Therefore, it is safe to conclude that all 111 observations in our SDF and SRF are all voluntary restatements Explanatory Variables Our key explanatory variable is court dismissal rate, defined by the number of securities cases dismissed within five years prior to a firm s fiscal year end in the federal district court where the firm is headquartered, divided by total such cases filed in the same period and court: dismissal i,t 4 t = no_dismissal i,t 4 t no_filings i,t 4 t (1) where, no_dismissal i,t 4 t is the number of cases dismissed within the five years prior to the end of fiscal year t of firm i handled by the district court where firm i is headquartered; and

17 no_filings i,t 4 t is the number of cases filed within the five years prior to the end of fiscal year t of firm i handled by the district court where firm i is headquartering. Two notes on this explanatory variable are in order: First, note it may take several years for some case to reach any sort of resolution, while other cases may be dismissed much faster. Therefore, cases dismissed within five years may not exactly correspond to cases filed during the same period. However, to account for the fact that different district courts have very different number of firms and lawsuit filings, we believe this scaling method is reasonable. In the robustness tests, we provide results using alternative estimation of court dismissal rate taking into account the lag between case filing and dismissal dates. Second, it is possible for a few less-busy district courts to have no case filed in the five-year period but are some cases dismissed in the same period. In such cases we assign value 1 to the court dismissal rate. Our key assumption, that securities lawsuits are typically heard by the district court that the defendant firm is headquartered, requires validation. Two statutory provisions: 28 U.S.C. 1404(a) and 1406(a) provide legal basis for this claim. Section 1404(a) protects parties and witnesses from an undue expenditure of time and money. Because of the nature of claims in securities lawsuits, substantially all witnesses and sources of proof are likely to be located at the firm s headquarters. Section 1406(a) provides the transfer of a case brought in an improper forum. Plaintiffs that file suits at inappropriate courts are either dismissed based on the doctrine of forum non conveniences, or transferred to the district court of the defendant firm s headquarter. Cox, Thomas, and Bai (2009) report their interview with plaintiffs counsels who consistently reflect that it is impractical for them to engage in forum shopping due to the strong likelihood that their choice of a venue other than the defendant firm s principal place of business will be immediately followed by a successful

18 defendant s motion to relocate the suit. Hence, rather than engaging in in futile act, they file suit initially in the defendant company s home district court. We proceed to verify this assumption using actual lawsuit and court data in our sample. Within the 393 observations in our Sample of Defendant Firms, 343 (or nearly 90%) of the cases are heard only heard by the defendant firms headquartering district court, providing validity to our assumptions. Table 1, Panel A displays the distribution of our two samples by district courts, and Panel B reports their yearly distribution. The top three busiest district courts are California (Northern), California (Central), and New York (Southern). Our control variables follow the literature on litigation risk and restatement. We first include court filing rate, defined as the total number of securities cases filed within five years prior to a firm s fiscal year end in the federal district court where the firm is headquartered, divided by total number of Compustat firms in the same period and court. Our firm-level controls include the natural logarithm of total assets, leverage ratio, and book-to-market ratio. Following the work on restatements (Files, Swanson, and Tse, 2009; Srinivasan, Wahid, and Yu, 2015) we take ROA, sales growth and last year stock return as control variables for firm performance. We further control the stock trading activities by including previous-year stock return volatility, market risk factor loading (beta), stock turnover, and stock return skewness (Kim and Skinner, 2012). Finally, to account for the strength of governance and monitoring system we include whether the firm is audited by a Big 4 auditing firm (Srinivasan, Wahid, and Yu, 2015). All variables are defined in Appendix A, and winsorized at 1% level, except for restating dummy, court dismissal rate and court filing rate. Table 2, Panel A and B summarizes the descriptive statistics of variables in our SDF and SRF, respectively, and compares characters of restated and non-restated firms in each sample.

19 For SDF (Panel A), the average court dismissal rate is 35.9%, and the average court filing rate is 15.7%. Mean log total assets (firm size) is 7.42, leverage 23.5% of total assets, and bookto-market ratio at The average ROA is -3.0% of total assets, and sales growth rate at 17.7%. Average daily return volatility is 3.6%, skewness 0.157, and annual turnover at 2, % of the firms are audited by the Big 4 auditing firms. Comparing voluntary restating firms with control firms in SDF, we find that voluntary restating firms have significantly higher court dismissal rate and lower court filing rate, and other characteristics are almost similar. This lends us confidence that our control firms are a good match for voluntary restating firms. For SRF (Panel B), the average dismissal rate is 36.7%, and the average court filing rate is 16.0%. Mean log total assets (firm size) is 7.22, leverage 25.3% of total assets, and book-to-market ratio at The average ROA is -3.3% of total assets, and sales growth rate at 22.4%. Average daily return volatility is 3.6%, skewness 0.133, and annual turnover at 2, % of the firms are audited by the Big 4. Compared with firms in SDF, firms in SRF have lower accounting returns but higher growth. Comparing voluntary restating firms with forced firms in SRF, we find that voluntary restating firms have higher court dismissal rate, lower court filing rate, and marginally higher skewness. Other characteristics are almost similar. Univariate analysis of the two samples reveals that court dismissal rate and court filing rate significantly distinguish voluntary restating firms with control firms. A higher court dismissal rate and a lower court filing rate are associated with a higher propensity to voluntarily restate, indicating that firms are more likely to make voluntary restatement when the court is more lenient and when the risk to be sued is lower, i.e. a lenient legal environment. Section 4 presents comprehensive results in regression analysis.

20 4. Empirical Results 4.1. Baseline Model To test the predictive power of court dismissal rate on misstating firms propensity to make voluntary restatement, we propose the following probit model: Restating i,t+1 (2) = β 0 + β 1 dismissal i,t 4 t + β 2 filed i,t 4 t + β 3 lev i,t + β 4 lnat i,t + β 5 return i,t + β 6 ROA i,t + β 7 salesgrth i,t + β 8 std i,t + β 9 turnover i,t + β 10 skewness i,t + β 11 Big4 i,t + β 12 btm i,t + β 13 beta i,t + ε i,t where, Restating t+1 is an indicator variable that takes 1 if the firm makes voluntary restatement before class-end date 16 ; dismissal i,t 4 t is the court dismissal rate for the headquartering firm; filed i,t 4 t is the court filing rate for the headquartering firm;lev t is the book leverage at the end of fiscal year t; lnat t is the natural logarithm of total assets at the end of fiscal year t; return t is the annual total return over fiscal year t; ROA t is the return on total assets in the fiscal year t; salesgrth t is sales growth from fiscal year t-1 to t; std t is the daily return volatility over fiscal year t; turnover t is the turnover over fiscal year t; skewness t is the return skewness over fiscal year t; Big4 t is an indicator variable that takes 1 if the firm has a Big-4 auditor firm as its auditor; btm t is the book-to-market ratio at the end of fiscal year t; and beta t is the market risk factor loading from CAPM model, estimated from the monthly stock returns of the 5-year period before the most recent fiscal year end. We control for the state fixed effects because some states have 16 Note that for the counterfactual firms in Sample of Defendant Firms and Sample of Restating Firms, we take their class-end date as the hypothetical restating date and the latest fiscal year up to the class-end date as fiscal year t in equation 2, since the restating firms and counterfactual firms are matched by lawsuit and class-end year in these two samples.

21 more than one district court and we need to disentangle the effect of court stringency from the unobservable state level economic, social and political effects. We also control for industry fixed effects and year fixed effects. Table 3 reports the impact of court dismissal rate on the likelihood of misstating firms issuing voluntary restatement using SDF (Column 1 and 2) and SRF (Column 3 and 4). Panel A presents the probit regression results, and Panel B exhibits the change in the likelihood of voluntary restatement with one-standard deviation increase (decrease) in court dismissal rate according to the probit regression results. Industry fixed effects and year fixed effects are included in all of the four regressions. State fixed effects are included in the regressions in Column 1 and 3. The industry and year fixed effects filter out time-varying and industry-level shocks that may affect the restatement decision, and the state fixed effect control for state-wide differences in business and regulatory environment. We find court dismissal rate exhibits a significantly positive impact on the likelihood of misreporting firms issuing voluntary restatements (p=0.0019, , and for the four regressions respectively). Remarkably, court dismissal rate is the only variable that has both statistical significance and consistent economic magnitude across all the four regressions. Specifically, Panel B shows that one-standard deviation increase in court dismissal rate (lowered court stringency) leads to 18.56%, 11.48%, 25.36% and 24.90% increase in voluntary restatement propensity for the four regressions in Column 1, 2, 3 and 4, compared to the average restating rate of 28.2% and 37.0% for the two samples, respectively. One standard deviation in dismissal rate amounts to the difference in the rates between Illinois (Northern) and California (Northern) district courts. Hypothetically, if a firm moves from Illinois (Northern) to California (Northern), ceteris

22 paribus, its voluntary restatement propensity would increase by 18.56% according to SDF, which increases over 65% of its voluntary restatement rate. ROA, sales growth, beta, turnover, return volatility, and Book-to-Market all exhibit insignificant effect on voluntary restatement. Surprisingly, court filing rate has no significant effect on voluntary restatement, probably because the litigation incidence is determined by other factors and controlling for firm characteristics and fixed effects could absorb its statistical significance. Also, having a Big-4 auditor has little impact on voluntary restatement, probably because auditors are concerned about their own legal and reputational penalties when their audited firm made accounting mistakes (Seetharaman, Gul, and Lynn, 2002; Hope and Langli, 2010). Overall, our baseline model analysis supports the defensive disclosure hypothesis The Availability Heuristic of Court Dismissal Rates To test whether court dismissal rates indeed affect managers estimate of their litigation environment, we employ the famous availability heuristic theory in social psychology (see Sherman and Corty 1984; and Strack 1985, for reviews). One of the most widely shared assumption in decision-making holds that people estimate the probability of an event by the ease with which instances or associations come to mind (Tversky and Kahneman 1973). It follows, managers have tendencies to overestimate the likelihood of events with greater availability in memory, which can be influenced by how recent the memories are, or how unusual or emotionally charged they may be (Schwarz et al. 1991). The availability heuristic enables cross-sectional tests to validate our key explanatory variable. If court dismissal rates indeed provide headquartering firms with an impression of stringency in

23 litigation environment, we expect to find (1) recent court dismissal rates to have larger impact than distant court dismissal rates; and (2) court dismissal rate in less busy district courts yield larger emotional charge, therefore larger impact, than dismissal rate in busier district courts. Table 4 and 5 present evidence supporting these hypotheses. Table 4 compares the impact of recent court dismissal rate with that of distant court dismissal rate on voluntary restatement. For recent dismissal rate, we use the 2-year court dismissal rate calculated from the most recent 2 years in the firm s headquartering district court. For distant dismissal rate, we use the 3-year court dismissal rate calculated from year 3 to year 5 in the firm s headquartering district court. Column 1 and 4 shows that impact of recent dismissal rate on firm s voluntary restating likelihood remains significantly positive (p= and for Column 1 and 4, respectively), whereas the impact of distant dismissal rate is less significant (p= and for Column 2 and 5, respectively). More clearly, when we regress firm s voluntary restating likelihood on both the recent and distant dismissal rate, the impact of distant dismissal rate is insignificant for both samples, but the impact of recent dismissal rate remains very significant (p= and for Column 3 and 6, respectively). For economic magnitude, the impact of recent dismissal rate is also larger than that of distant dismissal rate for both samples. Table 5 partitions our sample firms into those headquartered in the top 5 active courts and the rest. From the raw data of dismissal information, the top 5 active courts are California (Central), California (Northern), Florida (Southern), Massachusetts and New York (Southern), issuing 48% of the total dismissal decisions ever made since Column 1 and 2 presents the cross-sectional tests with the interaction term between Court Dismissal Rate and Top Active Courts Dummy. Column 3 to 6 display the subsample regressions, with Column 3 and 4 presenting the subsample without top 5 active courts and Column 5 and 6 displaying the subsample with only top 5 active

24 courts. Due to the small size of sample in the subsample tests, we only include year-fixed effects in Column 3 to 6. We find that while court dismissal rate exhibits significantly positive impact on the voluntary restatements in the subsample without top 5 active courts, the impact of court dismissal rate is insignificant for the subsample with only top 5 active courts. However, the coefficients for the interaction term between Court Dismissal Rate and Top Active Courts Dummy for both samples are insignificant, which might be explained by the insignificant but large coefficient for the court dismissal rate for the subsample with only top 5 active courts in Column 5 and 6. Taken together, Table 4 and 5 present evidence consistent with managers availability heuristics. Our tests suggest that (1) our court dismissal rate variable has information content of court stringency; (2) the positive sensitivity of voluntary restatement to court dismissal rate found in Table 3 is unlikely to be an artifact of measurement error or other randomness. Randomness cannot explain why recent dismissal rates have larger impact than remote ones, or why dismissal rates in less busy courts (which are easier to estimate) have larger impact than those in busier courts (which are harder to estimate) Shock in Court Stringency: Examining the Effect of Tellabs v. Makor To provide evidence that firms voluntary restatements are caused by court stringency, we exploit the 2007 case of Tellabs, Inc v. Makor Issues & Rights, Ltd. The case was originally dismissed by the district court of Northern Illinois, reversed by the 7 th circuit court upon appeal 17, further appealed to the Supreme Court which granted certiorati 18, and finally judge Posner of the 17 See 437 F.3d 588, 602 (7 th Cir. 2006) 18 See Tellabs, Inc. v. Makor Issues & Rights, Ltd. 551 U.S. 308 (2007).

25 7 th circuit court rendered final ruling following Supreme Court s clarified pleading standard. 19 We choose Tellabs because it is one of the Supreme Court s first efforts to clarify the key legal element in 10b-5 lawsuits: the strong inference standard for pleading Scienter, a mental state embracing intent to deceive, manipulate, or defraud. 20 Prior to Tellabs, there is longstanding confusion in the federal courts as to what is required of scienter allegations in order to defeat a motion to dismiss under the Private Securities Litigation Reform Act (PSLRA) of Different courts of appeals followed their own approach, and monitor by the Supreme Court is close to nonexistent (Westerland et al., 2010). For example, the 1st, 4th, 6th, and 9th circuit courts adopted a preponderance standard that is most favorable to defendant. It requires the inference that the defendants had the requisite Scienter (fraudulent intent or recklessness) to be the most plausible when compared with competing inference of No Scienter. The 2nd, 8th, 10th and 11th circuit and DC District Court adopted an equal inference standard that requires at least a tie of competing inference of Scienter and No Scienter. Lastly, the 3rd and 7th circuits adopted a reasonable person standard that is most favorable to plaintiffs. It only requires the court to look at the plausibility of the plaintiff s allegations, without requiring the assessment of competing inferences (Choi and Pritchard, 2012). Importantly, the Supreme Court s ruling on Tellabs in 2007 clarifies what is required for the plaintiff to plead Scienter. It held that plaintiffs shall survive a motion to dismiss only if a reasonable person would deem the inference of [culpable state of mind] cogent and at least as compelling as any opposing inference one could draw from the facts alleged. 22 This stance of the Supreme Court thus mimics the middle, equal inference standard, which is more stringent than 19 See Makor Issues & Rights, Ltd. v. Tellabs, Inc., F.3d, No , 2008 WL (7th Cir. Jan. 17, 2008). 20 See Ernst & Ernst v. Hochfelder, 425 U.S. 185, 193 (1976) See Tellabs, Inc. v. Makor Issues & Rights, Ltd. 551 U.S. 308 (2007), at 324.

26 the preponderance standard adopted by the 1st, 4th, 6th, and 9th circuit. To the extent that lower courts make decisions anticipating upper court s tendency and risk of reversal (Gulati, Choi and Posner, 2012; Choi, Gulati, and Posner, 2016), the effect of Tellabs is that it homogenizes what is required for plaintiffs to establish scienter across all circuits. Specifically, we hypothesize that the Tellabs decision exogenously increase the pleading standard of federal courts that previously adopted a pro-defendant standard of scienter relative to other federal courts. We follow this conjecture to design a quasi-difference-in-differences test. Using Tellabs as a shock on court pleading standard, the first difference is firms voluntary restatement likelihood between the post-event period and pre-event period. The second difference is that under pre-event lenient Circuits relative to that of district courts under pre-event non-lenient circuits. Note this diff-in-diff design in our analysis is different from the original diff-in-diff test, which requires the control group to be unaffected by the treatment. Here our control group comprising of misstated firms headquartered in non-lenient courts are also affected, but not in the direction of the treatment group. To the extent that the Supreme Court s decision is unpredictable, which affects the pleading standards of its Circuits in different directions, we believe the relevant changes in court dismissal rate between treatment group and control group are also exogenous, thus our research design is still valid. Following Choi and Pritchard (2011), we categorize district courts under the 1st, 4th, 6th, and 9th circuit as pre-event lenient circuit (treatment observations), and district courts under other circuits as control observations. We choose the 6-year window (3-year pre- and 3-year post- Tellabs), taking into account the fact that it takes time for managers of misstated firms to learn about their accounting mistakes and deliberate on restatement decision in response to altered pleading standard. As the Tellabs case spans 2006 through 2007, these two years are excluded

27 from our event window. Thus, the pre-event period for the 6-year window is from January 2003 to December 2005; and the post-event period for the 6-year window is from January 2008 to December To see whether Tellabs decision has homogenizing effect on federal courts, we first check the court dismissal rate for the district courts under the lenient (the preponderance standard) circuits and that under other circuits pre- and post-tellabs case. The results are shown in Panel A of Table 6. Consistent with Choi and Pritchard (2012), we find the court dismissal rates for the district courts under the treatment circuits decreases right after the Tellabs case while those under the control circuits increase after Tellabs. Panel B of Table 6 presents the results of the diff-in-diff analysis, i.e., the change in voluntary restatement likelihood after the Tellabs case under different court stringency. Column 1 and 2 use SDF and Column 3 and 4 use SRF. To eliminate potential bias of sample truncation when including control variables, we only control for industry fixed effects in regressions for Column 1 and 3, while we still perform regressions with full controls in Column 2 and 4. Remarkably, after Tellabs, the voluntary restatement probability of the lenient circuit firms significantly decreased relative to the control group. The coefficient estimate of the interaction term between pre-event lenient circuit dummy and post-event dummy is significantly negative (p=0.0484, p=0.0301, p= and in Column 1 to 4, respectively). Taken together, our result provides preliminary causal evidence that firms adjust their voluntary restatement policy in response to (exogenous) changes in court stringency.

28 4.4. Court Stringency and Legal / Market Outcomes of Voluntary Restatements So far, we have shown that misreporting firms are more likely to make voluntary restatements when they face a more lenient court. Whether this strategy works, in the sense that it results in better legal and market outcomes, is another question. Our working hypothesis is the following: If a strategy of voluntary restatement in lenient court jurisdictions is indeed effective, than we should find the legal and market outcomes of voluntary restaters in lenient courts to be better than in non-lenient courts. For a start, we cannot hope to compare the likelihood of litigation between voluntary restaters in lenient and stringent courts, because our definition of voluntary restatement requires firms to be litigated. We can, however, test: (1) whether lawsuits against voluntary restaters are easier to be dismissed in more lenient courts; (2) whether voluntary restaters pay less settlement amount in more lenient courts; and (3) the market reactions upon key event dates for voluntary restaters in lenient and stringent courts. Specifically, we conduct the following regression model in this section: Legal/MarketO i = γ 0 + γ 1 VR i + γ 2 LC i + γ 3 LC i VR i + Controls + ε i, (3) where, LegalO i is the lawsuit outcome measures, which include (1) indicator of dismissal, (2) settlement amount scaled by the defendant firm s total assets, and (3) the unconditional legal cost, which equals to the settlement amount scaled by the defendant firm s total assets when the motionto-dismiss is denied and 0 otherwise. MarketO i including 3-day (-1, +1) cumulative abnormal return (CAR) around class-end date and case-filing date. VR i is a dummy variable that takes 1 if the defendant firm in case i makes voluntary restatement and 0 otherwise. LC i is a dummy variable that takes 1 if the court dismissal rate of the headquartering court falls into the top quarter and 0 if the court dismissal rate of the headquartering court falls into the bottom quarter. To control for the

29 severity of the case we include the stock return during the class period and the natural logarithm of the length of the class period. Merging the daily stock trading data from CRSP and the filing dates and class-end dates from SCAS reduces our SDF firms to 149 observations and SRF firms to 116 observations. Table 7 Panel A present the result on legal outcomes. Column 1 and 2 show the probability of being dismissed, Column 3 and 4 show the settlement amount, and Column 5 and 6 show the unconditional legal cost. Consistent with our prediction, the interaction term between Voluntary Restating Dummy and Lenient Court Dummy is positive in dismissal, negative in settlement amount, and negative in unconditional cost amount, suggesting voluntary restating firms in lenient courts are more likely to win the litigation, pay less settlement amount, and incur less litigation cost than those in stringent courts. However, none of these interaction effects is statistically significant. Panel B of Table 7 presents the result on market outcome. We focus on the share price reaction around two key dates for SDF and SRF. The first is the class end date which is the date when the truth that corrected the stock price was allegedly revealed to the market (Kellogg, 1984; Griffin, Grundfest, and Perino, 2004; Gande and Lewis, 2009). The second event date is the case filing date, which is the date when the first plaintiff filed the lawsuit in a district court. We examine both event dates because the class end date was the first time that market knows about the misreporting, and studies that only focus on filing date returns tend to underestimate the true economic costs associated with securities lawsuits (Gande and Lewis, 2009). Column 1-4 of show the 3-day (-1, +1) cumulative abnormal return (CAR) around class-end date and case-filing date on SDF, and Column 5-8 on SRF using regression model in (3). We find insignificant coefficients for the interaction term between Voluntary Restating Dummy and Lenient Court Dummy (all p-

30 values>0.1), indicating that the benefit on the market reactions around class-end date and casefiling date of making voluntary restatement under lenient courts is not different from that under stringent courts. There is weak evidence that voluntary restatements are beneficial for misstated firms, indicated by the marginally significant positive coefficient for the Voluntary Restatement Dummy in Column 5 (p-value=0.0648) Discussion Results in Table 7 suggest making voluntary restatement in lenient courts can gain some advantage for the firm, but such advantage is not statistically significant between lenient and stringent courts. This result, however, shall not be taken as evidence defying our baseline results for two reasons: First, note our empirical design only allows us to use litigated firms. It is possible that a strategy of making voluntary restatement in lenient courts prevents shareholder litigation more in lenient than stringent courts. For example, using a simultaneous equations methodology, Field, Lowry, and Shu (2005) find that voluntary disclosure of bad news does not trigger, and even deters certain types of litigation. If that is the case, such strategy might be successful. Second, as this paper argues, court dismissal rates are taken as a heuristic, or mental short cut for court stringency. It is possible that managers responsiveness to court dismissal rates reflect their cognitive biases, such as the availability heuristics that we demonstrate in earlier test. It is therefore not surprising that we do not find a voluntary restatement strategy to work significantly better in lenient than stringent courts. However, it is important to note that none of these interpretations affects the validity of the fact we establish: misstated firms are more likely to make voluntary restatement in lenient than stringent courts.

31 5. Robustness Check 5.1. Alternative Measure of Court Dismissal Rate One shortcoming in the construction of our key explanatory variable, court dismissal rate, is that cases dismissed within five years may not exactly correspond to cases filed during the same period. However, to account for the fact that different district courts have different number of headquartering firms and lawsuit filings, we believe this scaling method is reasonable. To address this issue empirically, this subsection tests our baseline hypothesis using alternative court dismissal rate that takes into account the gap between case filing date and case dismissal date. dismissal i,t 5.5 t 1.5 = no_dismissal i,t 4 t no_filings i,t 5.5 t 1.5 (5) where, no_dismissal i,t 4 t is the number of cases dismissed within the five years prior to the end of fiscal year t of firm i handled by the district court where firm i is headquartered; and no_filings i,t 5.5 t 1.5 is the number of cases filed within the five years ended 18 months prior to the end of fiscal year t of firm i handled by the district court where firm i is headquartered. The period of 18 months is the average gap between the filing date and dismissal date for our dismissed cases from 1996 to 2013 in the SCAS database. Panel A of Table 8 compares the alternative court dismissal rate with our original court dismissal rate in all Compustat firm-years between 2000 and Our alternative court dismissal rate is insignificantly different from our original court dismissal rate (p=0.2538), confirming that our original court dismissal rate well captures court stringency despite the gap between filing period and dismissal period. We re-run the baseline regression analysis in Equation 2 with this alternative court dismissal rate. Panel B of Table 8 reports the results using SDF (Column 1 and 2) and SRF (Column 3 and

32 4). We find court dismissal rate still exhibits a significantly positive impact on the propensity of misreporting firms issuing voluntary restatement (p=0.0013, , and for the four regressions, respectively). This evidence supports that the gap between case filing date and case dismissal date does not contaminate our results Material Weakness Sample and External Validity This paper s research design relies on lawsuits to identify voluntary restating firms, which has the advantage of internal validity. However, in practice only a proportion of misstated firms are sued in securities lawsuits. To test whether our baseline findings that firms in lenient courts are more likely to make voluntary restatement is generalizable to larger sample of misstated firms, this section employs an alternative sample of firms that receive the material weakness opinion from their auditors. Our material weakness sample is constructed based on the internal control reports on the material weakness pursuant to SOX 404. After the SOX, there is auditor responsibility to identify Material Weakness (MW) in the internal control of the firm following Section 404, which is approved on June 5, 2003 and mandatorily enforced after April 15, Studies show firms who receive MW opinion have high restatement propensity and are likely to continue having misstatements in the following two years after receiving MW opinions (e.g. Myllymäki, 2013). Therefore, firms that receive MW opinion could be good candidate for us to construct alternative sample of misstated firms not based on ex post lawsuits. Appendix B.3 presents the selection process of our material weakness sample. First, we restrict our sample to 149,223 SOX 404 disclosure records from Audit Analytics SOX 404 database with opinion fiscal year in the period of 2003 to Then, we merge the SOX 404

33 records with Compustat firm-years by CIK and ensure that the Compustat datadate is within the 3-year window (-1,+1) taking SOX 404 opinion fixcal year as year 0 (Myllymäki, 2013), leaving 23,625 observations with at least 1 item of material weakness reported. By combining the 23,625 material weakness records with restatement database from Audit Analytics, we obtain 24,084 observations with or without restatement records 23. We select the firms with non-accounting-error restatement and restating date within 1-year period after restating period as our voluntary restating firms, and the remaining as control firms. Finally, by matching the sample with SCAS, CRSP and Compustat variables in the later tests, we obtain a final sample comprising 6,436 observations from January 1, 2003 to December 31, 2013, with 1,591 voluntary restating firm-years and 4,845 control observations. Table 9, Panel A displays the distribution of material weakness sample by district courts, Panel B reports their yearly distribution, and Panel C reports the summary statistics of major variables in the material weakness sample. Panel D reports empirical results using our material weakness sample. Consistent with our baseline result, court dismissal rate exhibits a significantly positive impact on the likelihood of material weakness firms issuing voluntary restatements (p= and two regressions respectively). Column (3) displays the results of diff-in-diff analysis in the Tellabs case. After Tellabs, the likelihood of making voluntary restatements of firms in the pre-event lenient circuits decreased significantly compared with that of firms in the pre-event equal inference circuits, which is indicated by the significantly negative coefficient estimate of the interaction term between pre-event lenient circuit and post-event dummy. Column (4) of Panel D presents result of an interesting placebo test on error-based restatements. Unlike accounting irregularities which are often related to fraud, accounting errors are mainly due 23 The number of observations (24,084) in matched result exceeds 23,625 because some MW firms made multiple restatements on different periods of a fiscal year.

34 to accidental omissions which are unlikely to support a 10b-5 lawsuit. This key difference allows us to use propensity of voluntary restatement on accounting errors as placebo. Intuitively, if firms restatement policy is indeed affected by their home court stringency in 10b-5 lawsuits, which, by legal criteria, is only relevant to fraud, then we should not expect to find court dismissal rate to have significant effect on restatement propensity for accounting errors. As predicted, Column (4) of Panel D shows that the court dismissal rate exhibits an insignificantly positive impact on the likelihood of material weaknss firm restating accounting errors (p=0.2882). This evidence, albeit indirect, lends some support to our argument that firms actively consider their home court stringency when making irregularity-based restatement. 6. Conclusion This paper provides the first evidence that district court stringency affects misreporting firms propensity to admit their accounting mistakes through restatements. We find strong and robust evidence that more stringent legal environment makes misreporting firms unlikely to make voluntary restatements. This strategy, however, appears not rewarded by the market or the court. This result sheds light on the long-debated question on the effect of institutional environment on voluntary disclosure of bad news. Whilst stronger legal environment may deter firms misreporting, it also deters misreporting firms tendency to admit their mistake. Future studies on the impact of legal environment on financial reporting quality will need to account for this complication in their conclusions.

35 Reference Ajinkya, B., Bhojraj, S., Sengupta, P., The association between outside directors, institutional investors and the properties of management earnings forecasts. Journal of Accounting Research 43, Akerlof, G.A., The market for lemons : Quality uncertainty and the market mechanism. The Quarterly Journal of Economics 84, Amoah, N.Y., Tang, A.P., Resolution of restatement-induced lawsuits after the Private Securities Litigation Reform Act. Research in Accounting Regulation 25, Atanassov, J., Do hostile takeovers stifle innovation? Evidence from antitakeover legislation and corporate patenting. Journal of Finance 68(3): Badertscher, B., J. Phillips, M. Pincus, and S. Rego Earnings management strategies and the trade-off between tax benefits and detection risk: To conform or not to conform? The Accounting Review 84 (1): Baginski, S.P., Hassell, J.M., Kimbrough, M.D., The effect of legal environment on voluntary disclosure: Evidence from management earnings forecasts issued in US and Canadian markets. The Accounting Review 77, Ball, R., Infrastructure requirements for an economically efficient system of public financial reporting and disclosure. Brookings-Wharton papers on financial services 2001, Beatty, A., Weber, J., Yu, J.J., Conservatism and debt. Journal of Accounting & Economics 45, Bertrand, M., and Mullainathan, S., Enjoying the quiet life? Corporate governance and managerial preferences. Quarterly Journal of Economics 111(5), Bhide, A. V., and H. H. Stevenson., Why be honest is honesty does not pay? Harvard Business Review 68(5),

36 Bhojraj, S., Libby, R., Yang, H., Analyzing guidance at the firm level: The association between guidance frequency, guidance properties, and market reactions. Unpublished Paper, Available a t/ ssrn. com/sol3/papers. cfm. Billings, Brooke, M., Cedergren, M.C., Strategic silence, insider selling and litigation risk. Journal of Accounting & Economics 59, Bird, S. W The assignment of cases to federal district court judges. Stanford Law Review 27(2), Brown, S.V., Tucker, J.W., Large sample evidence on firms year over year MD&A modifications. Journal of Accounting Research 49, Brown, S., Hillegeist, S.A., Lo, K., Management forecasts and litigation risk. Sauder School of Business Working Paper. Burks, J.J., Sarbanes-Oxley and the effect of restatements on CEO and CFO compensation and turnover. THE University of Iowa. Burns, N., Kedia, S., The impact of performance-based compensation on misreporting. Journal of Financial Economics 79, Cao, Z., Narayanamoorthy, G.S., The effect of litigation risk on management earnings forecasts. Contemporary Accounting Research 28, Cao, Z., Narayanamoorthy, G.S., Accounting and litigation risk: evidence from Directors and Officers insurance pricing. Review of Accounting Studies 19, Cazier, R.A., Christensen, T.E., Merkley, K.J., Treu, J.S., Litigation Risk and Non-GAAP Reporting. Working paper, available at: Chalmers, K., Naiker, V., Navissi, F Earnings quality and Rule 10b-5 securities class action lawsuits. Journal of Accounting and Public Policy 31, Cheng, C.S., Huang, H.H., Li, Y., Lobo, G., Institutional monitoring through shareholder litigation. Journal of Financial Economics 95,

37 Choi, S.J., Do the merits matter less after the Private Securities Litigation Reform Act? Journal of Law, Economics, and Organization 23, Choi, S.J., Gulati, M., and Posner, E.A The Role of Competence in Promotions from the Lower Federal Courts, 44 Journal of Legal Studies Choi, S.J., Pritchard, A., The Supreme Court s impact on securities class actions: An empirical assessment of Tellabs. Journal of Law, Economics, and Organization 28, 4: Choi, S.J., Pritchard, A., Securities Regulation: Cases and Analysis, Foundation Press (4 th edition) Choi, S.J., Nelson, K.K., Pritchard, A.C., The screening effect of the private securities litigation reform act. Journal of Empirical Legal Studies 6, Collins, D., Reitenga, A.L., Sanchez-Cuevas, J.M., The managerial consequences of earnings restatements. Memphis, TN. Collins, D., Masli, A., Reitenga, A. and Sanchez, J., Earning Restatements, the Sarbanes- Oxley Act, and the Disciplining of Chief Financial Officers, Journal of Accounting, Auditing & Finance 24, Cox, J.D., Thomas, R.S., Bai, L., Do difference in pleading cause forum shopping in securities class action lawsuits? Wisconsin Law Review 421: Dechow, P.M., Sloan, R.G., Sweeney, A.P., Causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research 13, Dechow, P.M., Ge, W., Larson, C.R., Sloan, R.G., Predicting material accounting misstatements. Contemporary Accounting Research 28, Desai, H., Hogan, C.E., Wilkins, M.S., The reputational penalty for aggressive accounting: Earnings restatements and management turnover. The Accounting Review 81, Djankov, S., Hart, O., McLiesh, C., and Shleifer, A., 2008, Debt enforcement around the world. Journal of Political Economy 116,

38 Dobbie, W., Goldsmith-Pinkham, P., Yang, C.S., 2016 forthcoming. Consumer bankruptcy and financial health. Review of Economics and Statistics. Donelson, D.C., McInnis, J.M., Mergenthaler, R.D., Yu, Y., The timeliness of bad earnings news and litigation risk. The Accounting Review 87, Dyck, A., Morse, A., Zingales, L., Who blows the whistle on corporate fraud? The Journal of Finance 65, Efendi, J., Srivastava, A., Swanson, E.P., Why do corporate managers misstate financial statements? The role of option compensation and other factors. Journal of Financial Economics 85, Feroz, E.H., Park, K.J., Pastena, V., The financial and market effects of the SEC's accounting and auditing enforcement releases. Journal of Accounting Research 29, Field, L., Lowry, M., Shu, S., Does disclosure deter or trigger litigation? Journal of Accounting & Economics 39, Files, R., Swanson, E.P., Tse, S., Stealth disclosure of accounting restatements. The Accounting Review 84, Francis, J., Philbrick, D., Schipper, K., Shareholder litigation and corporate disclosures. Journal of Accounting Research, Galasso, A., and Schankerman, M., Patents and cumulative innovation: causal evidence from the courts. The Quarterly Journal of Economics 130, Gande, A., and Lewis, C.M Shareholder-initiated class action lawsuits: Shareholder wealth effects and industry spillovers. Journal of Financial and Quantitative Analysis 44, Garvey, G., and Hanka, G., Capital structure and corporate control: the effect of antitakeover statutes on firm leverage. Journal of Finance 54, Graham, J., Harvey, C., and Rajgopal, S., The economic implications of corporate financial reporting. Journal of Accounting and Economics 40, 3 73.

39 Griffin, P.A., Grundfest, J.A., and Perino, M.A., Stock price response to news of securities fraud litigation: an analysis of sequential and conditional information. Abacus 40, Gulati, M., Choi, S.J. and Posner, E.A., What Do Federal District Judges Want?: An Analysis of Publications, Citations, and Reversals. Journal of Law, Economics & Organization 28(3), Hope, O. K., and Langli, J. C., Auditor independence in a private firm and low litigation risk setting. The Accounting Review 85, Hribar, P., Jenkins, N.T., The effect of accounting restatements on earnings revisions and the estimated cost of capital. Review of Accounting Studies 9, Hribar, P., Kravet, T., Wilson, R., A new measure of accounting quality. Review of Accounting Studies 19, Jayaraman, N., Mulford, C., Wedge, L., Accounting fraud and management turnover. Working paper, Georgia Institute of Technology. Jayaraman, S., Milbourn, T., Does equity-based CEO compensation really increase litigation risk. Unpublished Working Paper, Washington University. Johnson, M.F., Nelson, K.K., Pritchard, A.C., Do the merits matter more? The impact of the Private Securities Litigation Reform Act. Journal of Law, Economics, and Organization 23, Johnson, M.F., Kasznik, R., Nelson, K.K., The impact of securities litigation reform on the disclosure of Forward Looking information by high technology firms. Journal of Accounting Research 39, Karpoff, J., Koester, A., Lee, S., and Martin, G.S., Proxies and Databases in Financial Misconduct Research. The Accounting Review, forthcoming. Kellogg, R. L., Accounting activities, security prices, and class action lawsuits. Journal of Accounting and Economics 6,

40 Kim, I., Skinner, D.J., Measuring securities litigation risk. Journal of Accounting & Economics 53, Kinney, W.R., McDaniel, L.S., Characteristics of firms correcting previously reported quarterly earnings. Journal of Accounting & Economics 11, Kling, J.R., Incarceration length, employment, and earnings. The American Economic Review 96, Kothari, S.P., Shu, S. and Wysocki, P.D., Do managers withhold bad news?. Journal of Accounting Research 47, La Porta, R., Lopez-de Silanes, F., Shleifer, A., and Vishny, R. W., 1997, Legal Determinants of External Finance. The Journal of Finance 52, La Porta, R., Lopez-de Silanes, F., Shleifer, A., and Vishny, R. W. 1998, Law and Finance. Journal of Political Economy 106, Laux, V., Stocken, P.C., Managerial reporting, overoptimism, and litigation risk. Journal of Accounting & Economics 53, Lowry, M., Discussion of Shareholder litigation and changes in disclosure behavior. Journal of Accounting & Economics 47, Matsumoto, D.A., Management s incentives to avoid negative earnings surprises. The Accounting Review 77, Nagar, V., The role of manager s human capital in discretionary disclosure. Journal of Accounting Research 37, Palmrose, Z.V., Scholz, S., The Circumstances and Legal Consequences of Non GAAP Reporting: Evidence from Restatements. Contemporary Accounting Research 21, Palmrose, Z.V., Richardson, V.J., Scholz, S., Determinants of market reactions to restatement announcements. Journal of Accounting & Economics 37,

41 Pritchard, A.C., Securities Law in the Roberts Court: Agenda or Indifference. Journal of Corporate Law 37, Rogers, J.L., Buskirk, A.V., Shareholder litigation and changes in disclosure behavior. Journal of Accounting & Economics 47, Rogers, J.L., Stocken, P.C., Credibility of management forecasts. The Accounting Review 80, Schrand, C.M., Zechman, S.L.C., Executive overconfidence and the slippery slope to financial misreporting. Journal of Accounting & Economics 53, Schwarz, N., Bless, H.; Strack, F., Klumpp, G.; Rittenauer-Schatka, H., Simons, A., Ease of Retrieval as Information: Another Look at the Availability Heuristic. Journal of Personality and Social Psychology 61 (2): Seetharaman, A., Gul, F. A., Lynn, S. G., Litigation risk and audit fees: Evidence from UK firms cross-listed on US markets. Journal of Accounting and Economics 33, Skinner, D.J., Why firms voluntarily disclose bad news. Journal of Accounting Research 32, Skinner, D.J., Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics 23, Srinivasan, S., Wahid, A.S., Yu G., Admitting mistakes: Home country effect on the reliability of restatement reporting. The Accounting Review 90, Wald, J., and Long, M., The effect of state laws on capital structure. Journal of Financial Economics 83, Westerland, C., Segal, J.A., Epstein, L., Cameron, C.M., Comparato, S., Strategic defiance and compliance in the US courts of appeals. American Journal of Political Science 54, Wu, M., Earnings restatements: A capital market perspective. Available at SSRN

42 Yermack, D., Good Timing: CEO Stock Option Awards and Company News Announcements. Journal of Finance 52,

43 Appendix A. Variable Definition and Construction =1 if Voluntary Restating = indicator variable, equals to 1 if the firm restated its financial statements before class-end date in Sample 1 and 2 or within 1 year after restating period in Sample 3 and is not labelled in Audit Analytics as Res Clerical Errors, 0 otherwise Court Dismissal Rate = court dismissal rate is the number of dismissed cases in the Federal District Court that the firm belongs to in the 5 years before the most recent fiscal year end, scaled by the number of cases filed during the same period in the Court Court Filing Rate = court filing rate is the number of filed cases in the Federal District Court that the firm belongs to in the 5 years before the most recent fiscal year end, scaled by the number of Compustat firms during the same period in the Court Recent Court Dismissal Rate = recent court dismissal Rate is a 2-year court dismissal rate calculated from the most recent 2 years dismissal number in the Federal District Court that the firm belongs to scaled by the filing number in the corresponding period Distant Court Dismissal Rate = recent court dismissal Rate is a 3-year court dismissal rate calculated from the dismissal number in the Federal District Court that the firm belongs to in the period 3 years from the most recent fiscal year end to 5 yeasr from the most recent fiscal year end scaled by the filing number in the corresponding period. Log Total Assets = natural Logarithm of total assets (log(at)) Leverage = leverage, total debt scaled by total assets ((dltt+dlc)/at) ROA = return on assets, net income scaled by total assets (ni/at) Sales Growth = the different between sales in the most recent fiscal year and pervious year divided by the sales in pervious year Last Year Stock Return = compounded gross return over the most recent fiscal year Beta = beta from CAPM model, estimated from the monthly stock returns of the 5-year period before the most recent fiscal year end Return Volatility = return volatility, standard deviation of the daily return within the most recent fiscal year

44 Turnover (in 1000s) = 1-(1-TURN)n, where turn is average daily trading volume divided by the number of shares outstanding and n is the number of trading days in the most recent fiscal year Skewness = the third moment of the return distribution over the most recent fiscal year Book-to-Market = the book value of equity (CEQ) plus book value of deferred taxes (TXDB) divided by market value (PRC * SHROUT/1000), measured at the most recent fiscal year end. =1 if Auditor is Big 4 = indicator variable, equals to 1 if the firm is audited by big 4, 0 otherwise Pre-Event Lenient Circuit Dummy = indicator variable, equals to 1 if the firm is in the jurisdiction of district courts under the 1 st, 4 th, 6 th, and 9 th circuit following Choi and Pritchard (2011), 0 otherwise Lenient Court Dummy = indicator variable, equal to 1 if the court dismissal rate of the observation is in the top quarter of court dismissal rate for all compustat firms, 0 otherwise Filing Date CAR (-1,+1) = 3-day [t-1, t+1] cumulative abnormal return around the Federal case filing date, calculated by the cumulative return of the defendant firm s stock over the event window minus the cumulative return of the CRSP valueweighted return including dividends over the event window Revealing Date CAR (-1,+1) = 3-day [t-1, t+1] cumulative abnormal return around the accounting misconduct revealing date, calculated by the cumulative return of the defendant firm s stock over the event window minus the cumulative return of the CRSP value-weighted return including dividends over the event window =1 if Dismissed = indicator variable in Sample 1 and 2, equal to 1 if the case is dismissed, 0 otherwise Settlement Amount = the total settlement amount scaled by total assets Unconditional Legal Cost Amount = the total settlement amount scaled by total assets if the case is settled, 0 otherwise Class-period Return = The cumulative stock return over the class period Log Class Length = The natural logarithm of the number of days over the class period =1 if Sued = indicator variable in Sample 3, equals to 1 if the firm is filed with class action lawsuits as defendant after making voluntary restatement and the lawsuit s class period overlaps with the restating period, 0 otherwise

45

46 Appendix B.1 Selection of Defendant Firms Sample This table presents the selection process of our defendant firms sample, where each observation is a defendant firm in securities class action lawsuits (SCA). Voluntary restating firms are those with restatement before class-end date, and control firms are those without voluntary restatement but alleged with both GAAP violation and Rule 10b-5 violation. Sample Selection Procedures No. of Obs. Voluntary Restating Firms Begin with securities class action lawsuit defendant firms whose GVKEY, federal filing date, class periods and dismissal date (settlement date) are identifiable; 3,363 Merge with 11,377 restatement records from Audit Analytics; 789 Eliminate 514 observations with restatement on or after class-end date; 275 Eliminate observations without valid variables in our tests and require fiscal-end date to be between December 31, 2000 to December 31, Control Firms without Voluntary Restatement Begin with securities class action lawsuit defendant firms whose GVKEY, federal filing date, class periods and dismissal date (settlement date) are identifiable; 3,363 Merge with Compustat annual financial database; 3,170 Exclude 249 observations with restatement before class-end date; 2,921 Require defendant firms to be alleged of both GAAP violations and Rule 10b-5 violations; 928 Eliminate observations without valid variables in our tests and require fiscal-end date to be between December 31, 2000 to December 31, Total Sample of Defendant Firms 393 Voluntary Restating Firms 111 Control Firms without Voluntary Restatement 282

47 Appendix B.2 Selection of Restating Firms Sample This table presents the selection process of our restating firms sample, where each observation is an SCA defendant firm with restatement. Voluntary restating firms are defined and selected using the same method in Appendix B.1. For control firms, we require our defendant firms to have made corresponding restatement on or after the class-end date. The final sample comprises 300 observations from 2001 to 2012, with 111 voluntary restating firms and 189 control firms. Sample Selection Procedures No. of Obs. Begin with securities class action lawsuit defendant firms whose GVKEY, federal 3,363 filing date, class periods and dismissal date (settlement date) are identifiable; Merge with 11,377 restatement records from Audit Analytics; 789 Take 275 observations with restatement before class-end date as voluntary restating firms and 514 observations with restatement on or after class-end date as control firms; Eliminate observations without valid variables in our tests and require fiscal-end 300 date to be between December 31, 2000 and December 31, Total Sample of Restating Firms 300 Voluntary Restating Firms 111 Control Firms without Voluntary Restatement 189

48 Appendix B.3 Selection of Material Weakness Firms Sample This table presents the selection process of our material weakness firms sample, where each observation is a firm-year with material weakness (MW) from auditor s opinion. We define voluntary restating observations as those that made non-error-based restatement within 1 year after restating period, and the remaining as control observations. The final sample comprises 6,436 firm-year observations from January 1, 2003 to December 31, 2013, with 1,591 voluntary restating observations and 4,845 control observations. No. of Sample Selection Procedures Obs. Begin with SOX 404 disclosure records from Audit Analytics SOX 404 database with 149,223 opinion fiscal year 2004 to 2013 Merge with Compustat firm-years with three criteria: 1) CIK is matched; 2) at least 1 material weakness item is reported; and (3) the Compustat datadate is within the 3-23,625 year window (-1,+1) taking SOX 404 opinion fiscal year as year 0 Combine with restatement records from Audit Analytics restatement database 24,084 Eliminate observations without valid variables in our tests 6,436 Total Sample 6,436 Voluntary Restating Firms 1,591 Control Firms 4,845

49 Figure 1 Trends in Voluntary Restatement for Securities Class Action Defendants This figure presents the trends in voluntary restatement among securities class action (SCA) defendants. We assign the value series of lawsuits to the left vertical axis while the right axis indicates the proportion of voluntary restatements. Voluntary restatements are defined as SCA lawsuit defendants that make nonaccounting-error restatement before class-end date (misstatement revealing date). Involuntary restatements are defined as SCA lawsuit defendants that make non-accounting-error restatement on or after class-end date. Other SCA lawsuits are the SCA lawsuits excluding voluntary restatements and involuntary restatements % % # of Voluntary Restatements # of Involuntary Restatements # of Other SCA Defendants % of Voluntary Restatements 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0%

50 Figure 2 Federal District Court Dismissal Rate on 10b-5 securities lawsuits ( ) This figure displays the variation of court dismissal rate among district courts and volatility of court dismissal rate within each district court. According to the five-year court dismissal rate calculated in Appendix A, we take average of each court s dismissal rate in the period between December 31, 2000 and December 31, 2013 and calculate the standard deviation of court dismissal rate for each court in this period. Then, we rank all the courts according to their average court dismissal rate into three tertiles and mark them on the map with three different colors. The numbers in the map represent the standard deviation of court dismissal rate for each court and indicate the volatility of dismissal rate within each district court.