Summary and Analysis

THE ACCOUNTING REVIEW American Accounting Association Vol. 95, No. 3 DOI: 10.2308/accr-52589 May 2020 pp. 251–278

Does Litigation Deter or Encourage Real Earnings Management?

Sterling Huang Singapore Management University

Sugata Roychowdhury Boston College

Ewa Sletten The Ohio State University

ABSTRACT: In this paper, we rely on an exogenous shock to examine the impact of litigation risk on real earnings management (REM). We conduct difference-in-differences tests centered on an unanticipated court ruling that

reduced litigation risk for firms headquartered in the Ninth Circuit. REM increases significantly following the ruling for Ninth Circuit firms relative to other firms, consistent with litigation risk deterring REM. Additional analyses reveal that

REM rises more following the ruling when firms issue more optimistic disclosures. The evidence is consistent with

litigation deterring REM by constraining managers’ ability to issue optimistic and misleading disclosures that can

conceal the myopic and opportunistic motives underlying REM. We further document that an increase in REM in

response to a decline in litigation risk is more pronounced when managers have higher incentives to manipulate

earnings and governance mechanisms are weaker.

Keywords: real earnings management; earnings management; deterrence; litigation; corporate governance.

I. INTRODUCTION

R eal earnings management (REM) refers to managers’ attempts to overstate earnings by altering underlying real

transactions in a way that is often detrimental for long-term value (Graham, Harvey, and Rajgopal 2005;

Roychowdhury 2006 ). In anonymous surveys, executives admit to a preference for this method of managing earnings

relative to manipulating accruals (Bruns and Merchant 1990; Graham et al. 2005). This is cause for concern, because existing

evidence indicates that REM is more opaque to external stakeholders and more negatively associated with future firm

performance and stock returns than accruals management (Cohen and Zarowin 2010; Kothari, Mizik, and Roychowdhury

2016). Not surprisingly, the literature has been interested in factors that constrain REM, with most of the focus centering on

external and internal corporate governance (Bushee 1998; Zang 2012; Chen, Cheng, Lo, and Wang 2015; Cheng, Lee, and

Shevlin 2016 ). In the case of voluntary disclosures and financial reporting, litigation risk has been shown to be a powerful

disciplining force, in addition to governance (Skinner 1994; Hopkins 2018). In this context, the goal of our paper is to examine

whether the threat of litigation deters managers from engaging in REM.

REM involves the overstatement of earnings via real actions that depart from normal operational and investment processes.

Examples include overproduction, aggressive acceleration of sales, and reduction of discretionary expenses (Graham et al.

We thank Brad A. Badertscher (editor), Thomas Bourveau, Justin Hopkins, Kenneth Merkley, Lakshmanan Shivakumar (discussant), Siew Hong Teoh, Rencheng Wang, anonymous referees, and seminar participants at Boston College, The Chinese University of Hong Kong, The Hong Kong Polytechnic University, The Hong Kong University of Science and Technology, Northwestern University, Buffalo State College, SUNY, University of Massachusetts Lowell, University of California, Los Angeles, University of California, Irvine, University at Albany, SUNY, HEC Montreal, The Ohio State University, University of Illinois at Chicago, INSEAD, University of Pittsburgh, and the 2018 ISB Accounting Research Conference for valuable comments and suggestions. We are grateful to Siew Hong Teoh for providing earnings press release data, and Khrystyna Bochkay for providing MD&A tone data.

Editor’s note: Accepted by Brad A. Badertscher.

Submitted: July 2018 Accepted: October 2019

Published Online: October 2019

251

2005; Roychowdhury 2006; Zang 2012). The influence of litigation risk on REM is more ambiguous than that on disclosure

and financial reporting practices, as discussed below.

REM is not directly susceptible to class action lawsuits, which typically allege that managers issued misleading disclosures

or financial reports. The ‘‘business judgment’’ principle allows for the possibility that managers exercised their best judgment in choosing their actions, even though there were negative consequences ex post, providing legal cover for REM. It is important to consider, however, that REM hardly exists in a vacuum. A significant number of class action lawsuits accuse managers of

issuing misleading disclosures whose origins are rooted in REM. This is because managers bear the risk that they have to

explain to the investing community and financial analysts the rationale for short-term actions undertaken to boost earnings, but

detrimental to long-term value. REM can, thus, require managers to misrepresent business circumstances or the true intent of

their actions (Lo, Ramos, and Rogo 2017). Such statements, when revealed later to be misleading and deceptive, can become

subject to lawsuits. For example, over 2009–2010, analysts and auditors suspected Green Mountain Coffee of overproducing

inventory over successive periods to under-report their cost of goods sold (Green Mountain Coffee Class Action 2012). Green

Mountain Coffee managers repeatedly represented the sustained inventory increases as a response to ‘‘booming’’ business and surging demand. When Green Mountain eventually started missing sales and earnings targets, shareholders sued the company,

alleging that managers had misrepresented business conditions and misled shareholders about their inventory policies. 1

Because of this link between REM and misleading disclosures, higher threat of litigation can deter managers from engaging in

REM.

Litigation may also encourage REM. Managers intending to overstate earnings tend to shift from manipulating accruals to

real actions when accounting practices are under intense regulatory or court scrutiny (Cohen and Zarowin 2010; Zang 2012).

Consistent with this notion, Cohen, Dey, and Lys (2008) document that firms switched from accruals management to REM

after the Sarbanes-Oxley Act (SOX), which strengthened regulatory oversight over financial reporting. Since higher litigation

risk is known to constrain accruals manipulation (Hopkins 2018), it may induce managers to engage in more REM. Overall, the

net effect of litigation risk on REM is not obvious ex ante. To circumvent endogeneity in the relation between litigation risk and earnings management, we study changes in REM

following an exogenous shock to the stringency of securities class action litigation standards in the U.S. Ninth Circuit Court

District. On July 2, 1999, the U.S. Ninth Circuit Court of Appeals issued a surprising ruling (Re: Silicon Graphics Inc. Securities Litigation) requiring plaintiffs to prove that defendants acted with ‘‘deliberate recklessness.’’ The requirement significantly increased the hurdle for successful litigation against corporations headquartered in this circuit and reduced their

litigation risk (Pritchard and Sale 2005). Since the shock affected only firms located in states belonging to the Ninth Circuit, we

are able to compare their post-ruling changes in REM to those of firms located in states belonging to other circuits in difference-

in-differences tests.

Using a sample of firm-years spanning four years before and four years after the 1999 ruling that reduced litigation risk for

Ninth Circuit firms, we find significant post-ruling increases in REM for these firms relative to firms located in other circuits.

Our tests include firm and year fixed effects. The results are incremental to any increase in accruals management following the

ruling, and are robust to a battery of tests controlling for systematic differences between Ninth Circuit firms and other firms. In

parallel trends analyses, we find no trending differences preceding the ruling. Further, we document that REM also increases

following exogenous shocks that reduce investors’ ability to launch derivative lawsuits. The collective evidence is strongly

indicative of litigation risk acting as a deterrent to REM.

To substantiate the argument that the deterrence effect of litigation on REM arises due to its link with misleading optimistic

disclosure, we next focus on three different manifestations of optimism: (1) optimistic tone in the management discussion and

analysis (MD&A) section of 10-Ks, (2) optimistic tone in the earnings press releases, and (3) optimistic management forecasts.

We find that the post-ruling increase in REM is more pronounced in Ninth Circuit firms, with abnormally positive tone in the

MD&As and in the earnings press releases. In addition, REM increases more for Ninth Circuit firms that issue optimistic

forecasts. Overall, these results provide evidence on the link between litigation risk, REM, and misleading disclosures.

Our next set of tests provides evidence that litigation risk is a more salient deterrent to REM in those firms in which

managers have heightened incentives to manipulate earnings, and face weaker governance constraints on their ability to do so.

Consistent with the role of incentives, REM increased more following the ruling for Ninth Circuit firms meeting/beating

analysts’ forecasts, and for those in which managers engaged in greater opportunistic insider sales of stock. Consistent with the

positive role of governance, post-ruling REM increases were less pronounced for Ninth Circuit firms with lower managerial

entrenchment, higher institutional ownership, and higher ownership by institutions more likely to monitor the firm. One

possibility is that managers are truly optimistic at the time of their operational and investment decisions, and these decisions

1 In Section II, the subsection ‘‘Class Action Lawsuits Involving REM’’ discusses in greater detail litigation cases in which misleading disclosures issued to justify opportunistic overproduction of inventory, cost-cutting, and channel-stuffing led to litigation.

252 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

manifest as REM in our empirical measurement. Our cross-sectional findings are important in this context. They highlight that,

at the very least, litigation risk deters managerial actions that are not just motivated by optimism, but are also more likely to be

perceived as opportunistic.

In the light of previous evidence suggesting that managers make more aggressive accrual choices following the Ninth

Circuit ruling (Hopkins 2018), we examine how post-ruling increases in earnings management via real activities vary with

those via accruals. We find that REM increases more for Ninth Circuit firms that report lower discretionary accruals. Our results imply that increases in REM and accruals manipulation following the ruling were negatively correlated in the cross-section of

Ninth Circuit firms. Thus, the increase in REM cannot be attributed simply to the same firms increasing all types of earnings

management.

Research on factors that constrain REM has focused on various aspects of the firm’s governance: the appropriate design of

the executive compensation package (Chen et al. 2015), the checks and balances within a firm, that is, internal corporate

governance (Cheng et al. 2016), and the monitoring role of long-term institutional investors (Bushee 1998; Roychowdhury

2006). Our results suggest that litigation is an important force because it provides recourse to shareholders when governance,

incentive mechanisms, and other sources of monitoring fail. The ex post settling-up opportunity that litigation provides also makes it an effective ex ante deterrent to REM.

By considering both earnings management and disclosure choices, our paper provides a significantly expanded view of

how litigation risk affects managerial attempts to mislead investors. Hopkins (2018) and Chen, Li, Lo, and Xu (2018) provide

evidence on litigation risk deterring managers from engaging in aggressive accrual choices. Accrual manipulations, in and of

themselves, are misrepresentations and can lead to financial restatements and Securities and Exchange Commission (SEC)

enforcement actions, which can directly become the subject of lawsuits (Dechow, Sloan, and Sweeney 1996 ). The ability of

litigation risk to constrain REM is less obvious, since REM per se does not necessarily constitute a misrepresentation. Our findings are important in this context, as they suggest that opportunistic real actions can attract scrutiny and require managers to

misrepresent the purpose or the circumstances of these actions, which, in turn, can lead to litigation. Thus, higher litigation risk

can discipline managers’ real actions by tightening the constraints on managers’ ability to issue misleading disclosures.

Finally, Bourveau, Lou, and Wang (2018) document that a decline in the threat of derivative lawsuits is associated with an

increase in the frequency of both optimistic and pessimistic management forecasts, implying an improvement in voluntary

disclosure quality. Using the same setting, Lin, Liu, and Manso (2017) find that innovation increases when litigation risk

subsides. Thus, a decline in litigation risk can improve the quality of disclosures, as well as operational and investment

decisions. Our results strike a cautionary note on these benefits by pointing to the costs associated with a reduced threat of

lawsuits. In the presence of short-term goals and managerial opportunism, reduced litigation risk can increase myopic real

actions accompanied by misleading disclosures to conceal the true intent of those actions.

II. SETTING AND HYPOTHESES

Real Earnings Management (REM) Methods

The literature points to multiple ways in which managers can overstate earnings via their real actions. These actions

include overproduction, channel-stuffing, and aggressive reduction of discretionary expenses (Bushee 1998; Graham et al.

2005; Roychowdhury 2006; Stubben 2010; Zang 2012).

Overproduction refers to the practice of producing quantities significantly higher than expected sales and normal levels of

target inventory. Higher than normal production levels can conceivably result from anticipated increases in demand. However,

overproduction (higher than normal production without corresponding anticipated increases in future demand) can also be

driven by a managerial focus on inflating short-term profitability at the expense of long-term value, in at least two ways. First,

when the overproduced units are not sold and remain in inventory, fixed costs are spread out over all produced units because of

absorption costing, and cost of goods sold is understated, inflating current-period earnings. Larger inventories resulting from

overproduction impose holding costs on firms and can also lead to future losses because of increased obsolescence risk.

Second, managers can aggressively ship the overproduced inventories to intermediate dealers, even when retail demand is

lacking, in order to book sales. This practice, commonly referred to as ‘‘channel-stuffing,’’ is often achieved by incentivizing

dealers to accept higher inventories with especially lenient sales terms, such as price discounts. Concerns with channel-stuffing

include a reduction in margins, supply chain inefficiencies, and inventory backlogs. 2

2 Roychowdhury (2006), Cohen et al. (2008), and Cohen and Zarowin (2010) refer to the acceleration of sales via price discounts and lenient credit terms that typically characterize channel-stuffing as a manipulation of earnings through real actions. Stubben (2010) points to managers overstating revenues using multiple possible methods that can include real actions such as channel-stuffing.

Does Litigation Deter or Encourage Real Earnings Management? 253

The Accounting Review Volume 95, Number 3, 2020

Overproduction and channel-stuffing can lead to unusual patterns either in the firm’s own inventory or in the inventory

held by downstream dealers that are likely to come under scrutiny from investors and financial analysts. Consequently,

managers have incentives to characterize these actions as a response to increased demand, even though the increases in

production and shipping of inventory may be disproportionately high relative to that in demand.

Another type of REM involves an aggressive (and opportunistic) reduction of discretionary expenses, such as maintenance,

advertising, or research and development (R&D) expenditures. Managers can portray these reductions as having a persistent

positive influence on earnings, when, in fact, they are unsustainable and/or detrimental to a company’s long-term

competitiveness (Cohen and Zarowin 2010; Kothari et al. 2016; Vorst 2016 ).

The Link between REM and Misleading Disclosure

As discussed above, REM involves decisions on pricing, production, and discretionary expenditures that involve firms’

operations and investments and can have implications for value across multiple periods. Investors and analysts are, therefore,

keenly interested in, and often raise questions about, these operational and investment choices, for example, in conference calls

(Tasker 1998).

In engaging in REM, managers bear the risk that they are asked to justify actions that boost short-term earnings, but are

detrimental for long-term value. To successfully mislead, managers would then be forced to misrepresent their true intentions

and/or issue misleadingly optimistic statements that justify their real actions. For example, managers may be tempted to justify

excessive shipments to dealers as motivated by surging demand, or to characterize an aggressive reduction of discretionary

expenses as an elimination of inefficiencies. If these claims and declarations are later suspected or revealed to be false or

motivated by a desire to mislead, then they can attract lawsuits. Lawsuits that allege misrepresentation may additionally rely on

evidence of REM to argue scienter and intent on the part of managers. The lawsuit cases discussed next describe circumstances

in which firms engaging in REM attracted lawsuits because of associated misleading and optimistic statements that managers

made to justify their actions.

Class Action Lawsuits Involving REM

Class action lawsuits that explicitly refer to managers’ real actions share some common features. They typically allege

intent to mislead, managerial actions that were detrimental for firm and shareholder value, and associated misrepresentations

and misleading disclosures. We discuss a few examples below.

Over 2009–2010, analysts and auditors suspected Green Mountain Coffee of overproducing inventory to underreport their

cost of goods sold, due to a sustained increase in inventory levels over successive periods (Green Mountain Coffee Class

Action 2012). Green Mountain Coffee managers repeatedly represented the inventory increases as a response to ‘‘booming’’ business and surging demand it was ‘‘straining to meet’’ (Green Mountain Coffee Class Action 2012). Eventually, in an effort to convince investors that their production was indeed driven by high expected demand, company officials allegedly started

destroying and hiding excess inventory that had expired. Shareholders ultimately formed a class action in September 2010

accusing managers of misrepresenting demand and actively concealing their overproduced inventory. Plaintiffs relied heavily

on evidence of rising inventory levels resulting from overproduction to support the claim that managers had intended to

mislead. The case was further strengthened by the evidence that managers benefited personally by selling their shares while

misleading investors and attributing the increases in production to surging demand. The case was initially dismissed, but the

dismissal was later reversed by the Appeals Court. The case was relaunched in 2015, with Green Mountain finally settling in

2018 for $36.5 million (Green Mountain Coffee Class Action 2012, 2015; Green Mountain Coffee Settlement Agreement

2018).

In a well-publicized channel-stuffing case, Coca Cola Co. paid $137.5 million to settle a lawsuit accusing it of overstating

demand while using price discounts to ship excessive inventory of concentrate to their bottlers, where it stagnated because retail

demand was lacking (Coca Cola Co. Settlement Agreement 2008). Other high-profile security class actions involving channel-

stuffing include the lawsuit against Bristol-Myers Squibb Company in 2002 and that against General Motors in 2012 (Bristol-

Myers Squibb Class Action 2002; General Motors Co. Class Action 2012; Coca Cola Co. Class Action 2001).

The securities class action lawsuit against Hospira Inc. in 2011 was focused on an aggressive reduction of discretionary

expenses. Hospira had initiated a significant cost-cutting initiative in March 2009 that they called ‘‘Project Fuel.’’ Managers claimed the cost cuts would ‘‘increase shareholder value by eliminating underperforming and duplicative units and reducing its global workforce.’’ Shareholders alleged that managers misrepresented their true intentions, because the initiative reduced remediation costs and resulted in higher margins in the short-term while compromising crucial quality control procedures

(Hospira Inc. Class Action 2011; Raymond 2014). Specifically, the lawsuit accused the management of ‘‘gutting quality control efforts through cost cutting aimed at boosting short-term profitability.’’ The lawsuit alleged that by the time the Food and Drug Administration (FDA) detected the quality control deficiencies, product quality had suffered, sales had declined, and the

254 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

delayed remediation had become significantly more costly. Hospira settled the case in 2014 for $60 million (Hospira Inc.

Settlement Agreement 2014).

Excessive cost-cutting to boost short-term profitability at the expense of long-term competitiveness was also a key

component of the class action lawsuit against AT&T (AT&T Corporation Class Action 2001). Shareholders claimed that

upon becoming CEO at AT&T in 1997, C. Michael Armstrong was keen to demonstrate improved profitability, and pursued

extensive cost-cutting measures that eliminated ‘‘billions of dollars of overhead’’ and involved the termination of thousands of employees, including experienced service representatives and sales personnel. This aggressive expense-reduction

initially yielded a sharp increase in earnings and a surge in AT&T’s stock price. However, AT&T’s management did not

reveal that their intense cost-cutting resulted in a significantly depleted ability to grow revenues in their most important

business: long-distance calls by multinational companies. Furthermore, the elimination of service personnel negatively

impacted AT&T’s ability to retain customers, leading to the departure of several high-profile clients, such as BP Amoco,

Pepsi, and Chase Manhattan. When profitability and revenues ultimately declined immediately following AT&T’s public

issue of tracking stock, shareholders sued the company. They alleged that managers at AT&T were aware that the positive

effects of their cost cuts were short-lived and deliberately concealed imminent revenue declines. Further, the lawsuit

alleged that AT&T managers intentionally misattributed their poor profitability to ‘‘higher-than-anticipated wireless and cable upgrade costs,’’ when, in fact, they were aware that it was driven by revenue declines resulting from their earlier aggressive cost-cutting programs (AT&T Corporation Class Action 2001). The lawsuit lasted four years and was eventually

settled for $100 million.

Class Action Lawsuits and the Ninth Circuit Court Ruling

The 1995 Private Securities Litigation Reform Act (PSLRA) requires plaintiffs in securities class action lawsuits to

provide proof of scienter (i.e., intent or knowledge of wrong-doing). But the exact interpretation of the Act’s pleading

standards is left to the individual U.S. circuit courts. The Ninth Circuit Court was considered one of the most plaintiff-

friendly circuit courts, with a high volume of securities class action litigation (Gibney 2001; Pritchard and Sale 2005). 3

It,

thus, came as a surprise that on July 2, 1999, the Ninth Circuit Court of Appeals issued a ruling (Re: Silicon Graphics Inc. Securities Litigation, 183 F. 3d 970) that resulted in a much stricter interpretation of pleading standards than in any other circuit courts (Johnson, Nelson, and Pritchard 1999; Grundfest and Pritchard 2002).

4 According to the ruling, to form a class,

the plaintiffs in the Ninth Circuit are required to show that the defendants acted with ‘‘deliberate recklessness,’’ rather than mere ‘‘recklessness,’’ which is sufficient in other circuits.

5 The Ninth Circuit ruling introduced a particularly high burden of

proof given that evidence of intent is usually obtained only in discovery, after a class has been formed. Empirical evidence

indicates that the ruling significantly reduced the incidence of class action lawsuits. Crane and Koch (2018) document that

relative to the first half of 1999, lawsuit filings in the second half decreased in the Ninth Circuit by 43 percent, compared to

an increase of 14 percent in other circuits. 6

Pritchard and Sale (2005) point out that the exceptionally strict pleading

standards in the Ninth Circuit led to a higher rate of case dismissals.

Studying the effect of litigation risk on earnings management is challenging because of endogeneity. If firms more likely to

engage in earnings management are more susceptible to litigation, then this would manifest as a positive association between

the two, which, however, would not necessarily imply that litigation risk encourages earnings management (Skinner 1997;

Hopkins 2018). The Ninth Circuit ruling allows us to circumvent this endogeneity. It facilitates a difference-in-differences

framework in which we compare post-ruling changes in REM in firms headquartered in the Ninth Circuit to similar changes in

firms headquartered in other circuits. We use the firm’s headquarters to determine the most likely location of litigation, because

most lawsuits are ultimately litigated where the firm is headquartered. 7

3 The Ninth Circuit includes the following states: Alaska, Arizona, California, Hawaii, Idaho, Montana, Nevada, Oregon, and Washington.

4 The ruling was very likely driven by the ideologies of the ruling judges. The Ninth Circuit randomly selects three judges to adjudicate such cases (Hopkins 2018). In the case of Silicon Graphics, two conservative judges were selected, and they voted against the plaintiffs. The liberal-leaning third judge dissented, voting in favor of the plaintiffs.

5 Re: Silicon Graphics Inc. Securities Litigation, 183 F. 3d 970 (9th Cir. 1999) involved an allegation that managers engaged in insider trading after issuing misleading statements to hype the stock price. The case was dismissed on the premise that stock sales, coupled with negative internal performance news, are insufficient to prove that managers’ actions were deliberate.

6 Crane and Koch (2018) find a similar magnitude of post-ruling changes in lawsuit filings between the Ninth Circuit and other circuits over longer horizons, as well (from the beginning of 1998 to the end of 2000, and over their sample period: 1996–2014). Houston, Lin, Liu and Wei (2015b) also confirm decline in the number of class action lawsuits filed in the Ninth Circuit relative to other jurisdictions following the ruling.

7 Hopkins (2018) finds that at least 84 percent of lawsuits are ultimately litigated in the district of the firm’s headquarters. Filing lawsuits in districts other than the firm’s headquarters is costly and time-consuming for the plaintiffs. Moreover, even if shareholders were to file lawsuits against the firm in various locations, legal panels consolidate these lawsuits and relocate them to the district with the easiest access to documents and witnesses, i.e., a firm’s headquarters (Cox, Thomas, and Bai 2009; Hopkins 2018; Houston, Lin, and Xie 2015a).

Does Litigation Deter or Encourage Real Earnings Management? 255

The Accounting Review Volume 95, Number 3, 2020

Hypotheses

Recent evidence points to the power of the Ninth Circuit setting to detect changes in managers’ disclosure and reporting

behavior in response to reduced litigation risk. Houston et al. (2015b) and Cazier, Christensen, Merkley, and Treu (2017) use

the Ninth Circuit ruling to provide evidence on the disciplining effect of litigation on management forecasts and non-GAAP

(generally accepted accounting principles) reporting. 8

Hopkins (2018) finds that Ninth Circuit firms exhibit increased

restatement frequency and higher revenue management following the ruling, consistent with litigation risk deterring managers

from misstating financial statements. Hopkins (2018) attributes his evidence to misleading reporting choices. One caveat to his

interpretation is that the empirical proxy for revenue management cannot distinguish between higher revenues being achieved

via accruals or via real actions (see Stubben 2010).

The influence of changes in litigation risk on REM is ambiguous. Given the intrinsic information asymmetry about

real operations between a firm’s managers and its external stakeholders, REM is more opaque and, thus, harder to

detect than financial reporting violations and accrual manipulations (Cohen and Zarowin 2010; Kothari et al. 2016 ).

Further, REM involves business decisions that managers could justify as sound given their interpretation of the

economic circumstances, invoking the so called ‘‘business judgment’’ principle. Proving intentional wrong-doing under

securities law may, thus, be significantly more difficult in the case of REM, negatively impacting shareholders’ ability

to form a class. Thus, it is possible that opportunistic and myopic real actions are ‘‘immune’’ to changes in litigation

risk.

However, as already discussed above, REM is likely to be accompanied by misleading disclosures and misrepresentations,

as managers attempt to conceal the myopic and opportunistic nature of REM. An increase in the threat of litigation raises the

expected costs to managers of issuing such misleading disclosures and misrepresentations. Reduced managerial ability to issue

optimistic and misleading disclosures, in turn, is likely to deter REM, since it erodes managers’ ability to disguise the true

intent underlying their actions.

The expected costs of litigation associated with REM can be significant and include not only monetary expenditures, but

also loss of reputation and capital market penalties, even in instances when the lawsuits are ultimately dismissed. For example,

settlement costs alone for Bristol-Myers-Squibb, Coca Cola, and Hospira Inc. amounted to $300 million, $137.5 million, and

$60 million, respectively. 9

The case against General Motors alleging overproduction and channel-stuffing of inventory was

ultimately dismissed, but dragged on for three years, absorbing managers’ attention, their time, and expensive legal resources.

GM’s stock price declined over 30 percent as the market learned about the inventory issues (General Motors Co. Class Action

2012). Thus, the uncertainty about the eventual outcome of lawsuits, the opportunity cost of managers’ time, and the loss of

market value can make litigation extremely costly. This has the potential to deter REM. If that is the case, following the decline

in expected litigation costs with the 1999 ruling, then REM would increase in firms headquartered in the Ninth Circuit relative

to other firms.

It is also conceivable that a more litigious environment that discourages opportunistic reporting choices can encourage

REM. Cohen et al. (2008) document that firms replaced accruals management with more real earnings management after

the Sarbanes-Oxley Act in 2002 (SOX), which strengthened regulatory oversight over financial reporting. Cohen and

Zarowin (2010) find evidence consistent with REM being higher when there is greater auditor scrutiny of accruals and

lower ability to manipulate accruals. Similarly, Badertscher (2011) provides evidence that managers engage in REM to

sustain overvaluation only when they have exhausted their ability to manage accruals. Thus, it is possible that higher

litigation risk encourages managers to manipulate earnings via real activities instead of accruals. In that case, the reduction

in litigation risk with the 1999 ruling would lead to declines in REM at the Ninth Circuit firms, as they begin relying more

on accruals management.

In summary, the net effect of litigation risk on real earnings management is an empirical question. Our subsequent analyses

are, thus, designed to test the following two-sided hypothesis:

H1 (null): Real earnings management (REM) in firms headquartered in the Ninth Circuit, relative to firms headquartered

in other circuits, did not change after the 1999 ruling.

8 Houston et al. (2015b) find that management forecasts become less frequent and less accurate, and Cazier et al. (2017) find that non-GAAP reporting increases following the Ninth Circuit ruling.

9 See Bristol-Myers Squibb Settlement Agreement (2004), Coca Cola Co. Settlement Agreement (2008), and Hospira Inc. Settlement Agreement (2014).

256 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

III. SAMPLE DESCRIPTION, RESEARCH DESIGN, AND VARIABLES

Sample

We begin the sample selection process by obtaining data on all publicly listed firms from the Compustat/CRSP database with

non-missing information on historical headquarters over the period 1995–2003. 10

Our tests compare the post-ruling period, four

years beginning the year after the ruling (2000–2003), to the pre-ruling period, four years preceding the Ninth Circuit court ruling

(1995–1998). The year of the ruling, 1999, is excluded from the analyses. We restrict our analyses to four years before and four

years after the ruling to limit concerns about the potential effect of confounding events over longer horizons. 11

The sample excludes firms in financial industries (SIC 6000–6999) and utilities (SIC 4000–4999), as well as penny stocks

(firms with stock price smaller than $1). We next require the availability of Compustat and CRSP data necessary to construct

our control variables. 12

We include only firms with at least one year of data in both the pre- and post-1999 periods. We

winsorize all continuous variables at the 1st and 99th percentiles.

Research Design and Variable Definitions

We classify a firm as treated (or a ‘‘Ninth Circuit’’ firm) if its headquarters are located in one of the states subject to the Ninth Circuit Court.

13 The remaining firms are classified as ‘‘controls.’’ Our multivariate tests employ a difference-in-

differences research design in which we compare the changes in REM following the 1999 ruling for treated firms to the

corresponding changes for control firms.

We estimate the following difference-in-differences regression specification:

REMit ¼ b0 þ biPostt 3 Ninth Circuiti þ c0Xit þ FirmFE þ YearFE þ eit ð1Þ

i indexes firms, and t time. The Post indicator takes the value of 1 from 2000 to 2003, and 0 from 1995 to 1998, while the Ninth Circuit indicator takes the value of 1 in all sample years if the firm’s headquarters are located in one of the Ninth Circuit states. We expect a significant and positive b, the coefficient on the interaction term Post 3 Ninth Circuit. Because we include firm (FirmFE) and year (YearFE) fixed effects, we do not include Ninth Circuit and Post indicators separately. These indicators are absorbed in the firm fixed effects and year fixed effects, respectively. We cluster standard errors by the state of

headquarters.

Our tests use three measures of REM: (1) abnormal discretionary expenses (AbnDisx), (2) abnormal production costs (AbnProd), and (3) an aggregate measure that combines AbnDisx and AbnProd, denoted RM (see, for example, Cohen and Zarowin 2010; Zang 2012).

14 Discretionary expenses are the sum of advertising expenses, R&D expenses, and selling, general,

and administrative (SG&A) expenses, while production costs refer to the sum of the cost of goods sold and the change in

inventory. The estimation of AbnDisx and AbnProd follows Kothari et al. (2016 ). The two metrics are described in detail in in Appendix A.

Two specific features of our empirical proxies are worth noting. First, AbnDisx is appropriately signed to yield a higher value when real earnings management is higher, that is, when abnormal discretionary expenses are abnormally lower. Second,

AbnProd is distinct from the revenue management proxy used in Stubben (2010) and Hopkins (2018), which captures overstatement of sales via both accruals and real actions. Unusually high AbnProd will capture the effects of both overproduction and price discounts (as discussed in Roychowdhury [2006]), but will not be affected positively by any

associated accrual manipulations, for example, under-provision for estimated returns on sales. 15

We add AbnDisx and AbnProd

10 To determine the relevant court circuit, we use the firm’s historical headquarters at the time of the ruling, obtained from 10-K filings from EDGAR online. The historical headquarters are not available from Compustat, which backfills the data and instead reports the current headquarters for all firm- years in the database.

11 Using a shorter alternative window of three years before and after 1999, that is, 1996–2002, does not materially influence our results. This shorter window would exclude any effect following the passage of the Sarbanes-Oxley Act in 2002.

12 Control variables we require include ROA, total assets, firm age, dividends, capital expenditures, leverage, cash, debt issues, equity issues, abnormal accruals, and institutional ownership (which is assumed to be zero when data are unavailable on Thomson 13F). These variables are described in greater detail in the next section.

13 Firms change their headquarters infrequently. During our sample period, 147 firms (4.3 percent of the sample) changed headquarters location, with 46 firms switching from non-Ninth Circuit states to Ninth Circuit states. The largest shift occurred within non-Ninth Circuit firms, with 96 firms changing states, but not moving out of or into the Ninth Circuit. Our results are robust to excluding firms that changed states of headquarters.

14 We do not use abnormal cash flow from operations (CFO) because different methods of real earnings management can have opposite effects on CFO (Roychowdhury 2006 ).

15 If a firm under-provides for sales returns, then net sales will be overstated. Since we use net sales as an explanatory variable in the model for production costs, under-provision for sales returns will tend to yield lower production costs relative to sales, not higher. This possibly weakens the power of the proxy AbnProd, but does not cause it to be higher as a result of accruals manipulation.

Does Litigation Deter or Encourage Real Earnings Management? 257

The Accounting Review Volume 95, Number 3, 2020

to construct the aggregate measure RM. In untabulated tests, we confirm that all three metrics of REM, AbnDisx, AbnProd, and RM, are associated negatively with future operating performance, measured as one-year-ahead ROA.

X in Equation (1) represents a vector of control variables. We control for firm characteristics such as size (Firm Size), performance (ROA), capital structure and various investment and financing decisions (Leverage, Cash, Dividends, CAPEX, Debt Issue, Equity Issue, Acquisition). Given prior evidence on the link between institutional ownership and REM, we include the percentage of shares outstanding held by institutional investors (IO) in our models. We assume zero institutional ownership when the firm is not included in Thomson Financial 13F files. To capture real earnings management independent of any

accruals management, we control for abnormal accruals (AbnAcc). Finally, we control for the number of acquisitions, and assume zero for this variable for firm-years with no acquisitions according to Securities Data Company (SDC). Since the Ninth

Circuit Court ruling represents an exogenous geographic shock to litigation risk, we include state-level variables to control for

variation in socioeconomic factors that can affect managers’ propensity to manipulate their real actions (Hilary and Hui 2009;

McGuire, Omer, and Sharp 2012; Heider and Ljungqvist 2015). These variables include GDP Growth, Unemployment, Income, Population, Gender, Ethnicity, Education, and Vote Democrat. All variables are defined in Appendix A.

IV. DESCRIPTIVE STATISTICS AND PRIMARY RESULTS

Descriptive Statistics

Table 1 reports the descriptive statistics for real earnings management measures, as well as the control variables for our

sample. In every test, we use the maximum number of observations for each dependent variable after requiring that all cross-

sectional partitioning variables be available, as well. For example, our main analysis of the ruling’s effect on abnormal

discretionary expenses of the Ninth Circuit firms relative to non-Ninth Circuit firms includes 3,387 distinct firms and 15,225

firm-year observations between 1995 and 2003. The number of observations is slightly lower for AbnProd, at 14,958, than it is for AbnDisx as a consequence of data availability. The intersection of observations with sufficient data to compute both measures and to construct our aggregate measure RM includes 14,215 firm-years. The exclusion of penny stocks and the data requirements for computing earnings management proxies lead to slightly larger and more profitable firms in our sample

relative to the Compustat population. The median firm in our sample has an ROA of 13.3 percent, total assets of $199.8 million,

and institutional ownership of 34.7 percent.

Table 2 provides some preliminary insights into the difference-in-differences in REM in a univariate setting. The table

reports pre- and post-ruling levels of our three REM measures, as well as changes in those measures separately for firms

headquartered in the Ninth Circuit (Column (1)) and those headquartered in other circuits (Column (2)). Columns (1) and (2)

also report p-values from t-tests of changes in REM from the pre- to the post-ruling period. As Column (1) demonstrates, Ninth

Circuit firms exhibit increases in all three metrics of REM (AbnDisx, AbnProd, and RM) following the ruling. These increases are statistically significant for the individual REM metrics AbnDisx and AbnProd, but not for the composite metric RM. Column (3) reports the results for our main difference-in-differences tests comparing changes in REM for Ninth Circuit firms relative to

those for non-Ninth Circuit firms. All three measures increase significantly at the 1 percent level. The univariate tests, thus,

yield preliminary evidence consistent with a decline in litigation risk resulting in a significant increase in REM.

Results of Multivariate Tests on Changes in Real Earnings Management

This section examines in multivariate tests whether the Ninth Circuit Court ruling led to any significant changes in REM

(H1). We estimate Equation (1) using our three measures of REM as dependent variables. Table 3 presents the results. The

coefficients on the interaction term Post � Ninth Circuit are positive and significant, with p-values less than or equal to 0.05 for all three measures of REM. The coefficients point to statistically significant increases in REM following the ruling for Ninth

Circuit firms relative to other firms. Moreover, these increases are economically significant. The increase in AbnDisx represents 10 percent of one standard deviation in AbnDisx, and that in AbnProd represents 8.9 percent of one standard deviation of AbnProd. The increase in the aggregate measure represents 10.1 percent of its standard deviation. The results indicate that a decline in litigation risk with the Ninth Circuit ruling led to significant increases in REM at Ninth Circuit firms relative to other

firms, consistent with litigation’s deterrence effect on REM. 16

The results in Table 3 also provide some support for the possibility that REM and accruals manipulation occur in

conjunction. The coefficient on AbnAcc is positive and significant at the 5 percent level for all three REM metrics, suggesting

16 Our results are not sensitive to removing observations with missing institutional ownership. Moreover, for a restricted sample with data available, we confirm that our results are robust to controlling for CEO tenure, where we include both its first and second power (to allow for potential nonlinearity, see Ali and Zhang [2015]).

258 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

complementarity between REM and accruals management. Examples of this complementarity include managers channel-

stuffing, but not adequately providing for estimated returns, and overproducing inventory, but not increasing their inventory

obsolescence reserves.

We next perform a number of robustness tests, beginning with a parallel trends analysis. This analysis investigates the

possibility that the post-ruling changes reflect trending differences between Ninth Circuit firms and other firms from prior to the

ruling, and do not result from the ruling itself. Table 4 reports the results. In these tests, 1995 (that is, year t�4) serves as the benchmark year. We find no trending differences between the Ninth Circuit and other circuit firms before the ruling. The

TABLE 2

Univariate Tests

(1) Ninth Circuit States

(2) Other Circuit States

(3) Diff-in-Diff

Before After p-value Before After p-value Diff-in-Diff p-value

AbnDisx �0.004 0.010 0.032 0.025 �0.016 0.000 0.027 0.000 AbnProd �0.006 0.014 0.002 �0.003 0.003 0.005 0.013 0.000 RM �0.001 0.007 0.368 0.020 �0.012 0.000 0.040 0.000

The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000– 4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. This table reports the univariate comparison between firms located in Ninth Circuit states and those in other circuit states. All variables are defined in Appendix A.

TABLE 1

Descriptive Statistics

n Mean Median Std P25 P75

AbnDisx 15,225 0.004 0.027 0.142 �0.030 0.065 AbnProd 14,958 0.001 0.014 0.137 �0.038 0.059 RM 14,215 0.004 0.031 0.197 �0.044 0.091 Ninth Circuit 15,225 0.169 0.000 0.375 0.000 0.000 ROA 15,225 0.099 0.133 0.286 0.060 0.205 Total Assets (Millions) 15,225 1,275.460 199.794 3,154.246 49.223 823.280 Dividend 15,225 0.007 0.000 0.017 0.000 0.008 CAPEX 15,225 0.062 0.042 0.066 0.022 0.076 Leverage 15,225 0.236 0.195 0.243 0.027 0.357 Cash 15,225 0.164 0.071 0.209 0.020 0.229 Debt Issue 15,225 0.012 0.000 0.095 �0.017 0.023 Equity Issue 15,225 0.038 0.000 1.101 �0.010 0.017 Acquisition 15,225 0.035 0.000 0.184 0.000 0.000 IO 15,225 0.370 0.347 0.281 0.099 0.610 AbnAcc 15,225 0.000 0.004 0.104 �0.039 0.045 GDP Growth 15,225 4.387 3.125 4.562 �0.059 7.850 Unemployment 15,225 0.029 0.029 0.004 0.027 0.033 Income 15,225 16.886 14.020 13.012 6.758 21.906 Population 15,225 2.933 2.970 0.821 2.378 3.633 Gender 15,225 0.504 0.506 0.008 0.499 0.510 Ethnicity 15,225 0.845 0.871 0.086 0.775 0.927 Education 15,225 0.217 0.219 0.036 0.194 0.237 Vote Democrat 15,225 0.444 0.444 0.069 0.414 0.478

The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000– 4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All variables are defined in Appendix A.

Does Litigation Deter or Encourage Real Earnings Management? 259

The Accounting Review Volume 95, Number 3, 2020

TABLE 3

Real Earnings Management (REM) Changes

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.016 0.013 0.023 (3.97)*** (2.03)** (3.66)***

ROA 0.158 �0.155 �0.001 (5.29)*** (�7.03)*** (�0.04)

Firm Size 0.013 0.009 0.022 (2.48)** (1.33) (2.45)**

Dividend 0.043 0.205 0.218 (0.38) (1.57) (1.08)

CAPEX 0.050 0.093 0.117 (1.34) (2.50)** (2.69)***

Leverage 0.056 �0.037 0.010 (2.85)*** (�2.65)** (0.35)

Cash 0.067 �0.014 0.020 (2.66)** (�0.76 ) (0.60)

Debt Issue �0.150 0.093 �0.032 (�6.75)*** (5.63)*** (�1.25)

Equity Issue �0.004 0.001 �0.002 (�1.16) (0.55) (�0.41)

Acquisition 0.005 0.012 0.010 (0.86) (1.25) (0.84)

IO �0.018 �0.016 �0.038 (�1.29) (�0.95) (�1.57)

AbnAcc 0.058 0.106 0.176 (2.11)** (3.14)*** (3.89)***

GDP Growth �0.000 0.001 0.001 (�0.09) (0.71) (0.67)

Unemployment �0.191 1.201 0.621 (�0.19) (0.70) (0.30)

Income 0.001 �0.001 0.000 (1.61) (�1.19) (0.24)

Population �0.002 �0.010 �0.013 (�0.10) (�0.50) (�0.77)

Gender �1.207 0.469 �1.803 (�1.18) (0.43) (�1.14)

Ethnicity �0.236 0.299 �0.128 (�1.91)* (2.54)** (�0.69)

Education 0.635 �0.505 0.605 (2.21)** (�1.60) (1.03)

Vote Democrat �0.053 �0.097 �0.049 (�0.72) (�1.20) (�0.46)

Observations 15,225 14,958 14,215

R 2

0.420 0.312 0.405

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 3 examines changes in REM following the Ninth Circuit Court ruling. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

260 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

coefficients on D(t ¼�1) � Ninth Circuit, D( ¼�2) � Ninth Circuit, and D(t ¼�3) � Ninth Circuit indicators are not significantly different from zero in any of the columns. The results of this analysis imply that there are no preexisting

differential trends in REM between the Ninth Circuit firms and other firms. The differences appear in the year after the ruling.

The coefficient on D(t¼1) � Ninth Circuit is significant and positive for all three REM metrics. The differences continue to be significantly different in years 2 and 3 after the ruling for AbnDisx, in year 4 for AbnProd, and in all three years for the composite metric RM. Additionally, Figure 1, Panel A provides a graph for the composite metric RM for both treatment firms and control firms. The data in the graph are consistent with the results we observe with the parallel trend regressions.

While real earnings management increases up to four years following the ruling, the increases are economically more

pronounced and statistically more significant in the first two post-event years. One reason for this may be a lack of power—

breaking the post-event period into separate years and analyzing separate components of REM potentially weakens the power

of our tests. Second, it is also possible that in later years following the ruling, alternative mechanisms step in to compensate for

the reduced discipline from litigation risk. For example, Crane and Koch (2018) document that institutional ownership in Ninth

Circuit firms rises following the ruling, which they interpret as a capital market response to the decline in governance by

litigation.

Our next set of robustness analyses aim to alleviate concerns that systematic differences between the Ninth Circuit firms

and non-Ninth Circuit firms contribute to our findings. First, we confirm that our results are robust to reestimating our

regressions after augmenting them with industry-year fixed effects. Next, we consider the specific concern that Ninth Circuit

firms are disproportionately represented by firms in high-technology industries, which experienced the crash of the internet

bubble in 2000 (i.e., in the post-ruling period). We conduct two separate tests in which we exclude from our sample (1) all firms

in high-technology industries, and (2) all firms in industries with negative cumulative abnormal returns over the years 2000–

2003. Despite a significant decline in the number of observations, we continue to find significant and positive coefficients on

TABLE 4

Real Earnings Management (REM) Changes: Parallel Trends Analysis

(1) AbnDisx

(2) AbnProd

(3) RM

D(t ¼ �3) � Ninth Circuit �0.009 0.016 0.018 (�0.89) (1.38) (1.60)

D(t ¼ �2) � Ninth Circuit 0.008 0.004 0.020 (0.68) (0.46) (1.19)

D(t ¼ �1) � Ninth Circuit 0.011 0.018 0.035 (0.83) (1.40) (1.20)

D(t ¼ 1) � Ninth Circuit 0.026 0.029 0.050 (2.53)** (2.54)** (2.50)**

D(t ¼ 2) � Ninth Circuit 0.045 0.014 0.057 (2.86)*** (1.24) (2.77)***

D(t ¼ 3) � Ninth Circuit 0.009 0.005 0.031 (0.92) (0.50) (1.77)*

D(t ¼ 4) � Ninth Circuit �0.000 0.030 0.036 (�0.03) (1.70)* (2.53)**

Controls Yes Yes Yes

Observations 15,225 14,958 14,215

R 2

0.427 0.313 0.407

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 4 examines changes in REM in year t relative to the year of the Ninth Circuit Court ruling. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

Does Litigation Deter or Encourage Real Earnings Management? 261

The Accounting Review Volume 95, Number 3, 2020

FIGURE 1 Parallel Trends: RM and AbnAcc

Panel A: Parallel Trend—RM

Panel B: Parallel Trend—AbnAcc

Figure 1 plots the changes in RM (AbnAcc) in year t relative to the year t�4. The x-axis represents event year, where t¼0 is the year of the Ninth Circuit ruling. We plot residual RM (AbnAcc) from the regression of RM (AbnAcc) on all control variables (other than Post � Ninth Circuit) in Equation (1). The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000-4999 and SIC 6000-6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample.

262 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

Post � Ninth Circuit in all columns. To control for geographic factors, we perform an adjacent-county analysis, in which we only include firms with headquarters located within 40 miles of the state borders, where one of the borders passes the Ninth

Circuit. 17

We find positive coefficients on Post � Ninth Circuit in all three columns, with p-values less than 0.05 (results untabulated). We also perform falsification tests in which we replace Ninth Circuit firms with those from other circuits, circuit

by circuit. We do not find consistent significant difference-in-differences increases in REM for any other circuit. The results of

these tests are untabulated, but available upon request.

Finally, we construct a propensity score matched sample of treatment and control firms. Specifically, we limit our sample

to the Ninth Circuit and non-Ninth Circuit firm-years that are propensity score matched in year 1998 (the year before the ruling)

on several firm characteristics. 18

Table 5, Panel A shows the mean values of firm characteristics for the Ninth Circuit and non-

Ninth Circuit firms in this propensity score matched sample, as well as p-values from t-tests of differences. The matching is

successful, yielding no significant differences in the overall propensity matching score. Among individual firm characteristics,

only Firm Size is different at the 10 percent level between the treatment and the propensity score matched control sample. Table 5, Panel B presents the various REM proxies for both the treated and propensity score matched control firms

separately for the pre-ruling period and the post-ruling period. No REM metric is significantly different between treated and

control firms in the pre-ruling period, but all REM metrics are significantly different for treatment firms following the ruling. 19

Table 5, Panel C, Columns (1)–(3) report the results from reestimating Equation (1) on the propensity score matched sample.

Despite a substantially lower number of observations, we find significant post-ruling increases in REM at the Ninth Circuit

firms relative to those at similar firms in other circuits.

Results on Changes in REM and Misleading Disclosure

We next turn our attention to examining potential complementarities between REM and misleading disclosures. Our tests

utilize two primary measures of optimistic/misleading disclosures: (1) abnormally optimistic tone of the MD&A section of the

10-Ks (OptTone_MDA), and (2) abnormally optimistic tone of earnings announcement press releases (OptTone_EPR). Tone is measured as the difference between the count of positive and negative words scaled by the total number of words, following the

Loughran and McDonald (2011) methodology. Abnormal tone is measured following Huang, Teoh, and Zhang (2014), as a

residual from a regression of tone on a variety of factors that capture the firm’s characteristics, performance, and volatility. 20

Thus, positive or negative abnormal tone captures unusual optimism or pessimism, respectively, on the part of the managers

providing the disclosures. OptTone_MDA and OptTone_EPR are set equal to 1 if abnormal tone in the MD&A or the earnings press release, respectively, are positive.

Rogers, Van Buskirk, and Zechman (2011) find evidence that optimistic tone attracts class action litigation. Huang et al.

(2014) report that unusually positive tone in earnings press releases is associated with lower future earnings and cash flows, as

well as negative future returns, and conclude that managers use positive tone to mislead shareholders. Prior evidence suggests

that significant variation exists in the content and tone of MD&As (Brown and Tucker 2011; Mayew, Sethuraman, and

Venkatachalam 2015; Muslu, Radhakrishnan, Subramanyam, and Lim 2015). Further, Rogers et al. (2011) report that in a

randomly selected sample of 20 lawsuits, 55 percent referenced optimistic statements occurring within the 10-Ks. Requiring the

availability of MD&As and the data for computing abnormal tone reduces the sample size significantly. Earnings press releases

are also a natural disclosure medium through which managers can channel their optimism. 21

Unfortunately, the availability of

earnings press releases is even more limited during our sample period (1995–2003). The SEC mandated the filing of earnings

releases via Form 8-K starting only in 2003. As a result, the texts of earnings releases before 2003 are available only for some

firms via PR Newswire or Business Wire starting from 1997 (Huang et al. 2014). In the tests relying on unusual optimism in

earnings press releases, the number of years in the pre-event period is, thus, restricted to two.

We expand our main specification to include each disclosure optimism measure (OptTone_MDA and OptTone_EPR), as well as its respective interaction terms with Post, Ninth Circuit, and Post � Ninth Circuit. We are particularly interested in the

17 This test exploits the fact that economic conditions and other unobserved confounding factors (e.g., social, political) centric to certain geographic locations are likely to be similar across state borders, while the effects of the Ninth Circuit ruling stop at a state’s border.

18 The characteristics include: industry membership, defined as two-digit SIC code, ROA, size, dividends, capital expenditures, leverage, cash holdings, net debt issuance, equity issues, number of acquisitions, institutional ownership, and abnormal accruals.

19 In untabulated robustness tests, we also match on REM metrics measured in 1998. After the matching, we find no statistically significant differences between control and treated firms on any dimensions, including pre-ruling REM. Our results are fully robust to this alternative matching procedure.

20 In arriving at abnormal tone, managerial tone is orthogonalized specifically with respect to the following factors: earnings, returns, size, book-to- market, volatility of returns and earnings, firm age, number of business and geographic segments, the incidence of losses, change in earnings, analyst forecast of earnings, and analyst forecast error.

21 While conference calls would be another form of communication worth examining to capture less scripted optimism in managerial responses to analysts’ questions, conference call data are only available starting from 2001.

Does Litigation Deter or Encourage Real Earnings Management? 263

The Accounting Review Volume 95, Number 3, 2020

TABLE 5

Real Earnings Management (REM) Changes and Propensity Score Matching

Panel A: Matching Quality

Control Treated p-value

AbnDisx 0.002 0.025 0.000 AbnProd �0.004 0.019 0.000 RM �0.005 0.041 0.000 PSM Score 0.199 0.199 0.987 ROA 0.084 0.097 0.203 Firm Size 4.949 5.078 0.058 Dividend 0.005 0.005 0.910 CAPEX 0.067 0.068 0.944 Leverage 0.208 0.213 0.574 Cash 0.222 0.214 0.348 Debt Issue 0.016 0.016 0.852 Equity Issue 0.030 0.063 0.518 Acquisition 0.045 0.043 0.798 IO 0.330 0.341 0.254 AbnAcc �0.002 �0.003 0.646

Panel B: Univariate Comparison

Control Treated p-value

Pre-Ruling Period

AbnDisx 0.015 0.012 0.669 AbnProd �0.008 �0.004 0.539 RM 0.004 0.005 0.926

Post-Ruling Period

AbnDisx �0.023 0.050 0.000 AbnProd 0.004 0.065 0.000 RM �0.022 0.111 0.000

Panel C: Multivariate Regression

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.030 0.025 0.056 (2.44)** (1.93)* (2.58)**

Controls Yes Yes Yes

Observations 3,276 3,276 3,276

R 2

0.728 0.684 0.704

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Panel A gauges matching quality of the dependent and independent variables for the propensity score matched treatment and control samples. Panel B provides univariate comparison of REM for both the pre- and the post-ruling periods. Panel C estimates Equation (1) using this propensity score matched sample. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

264 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

sign of the interaction term between each optimism measure and Post � Ninth Circuit. Table 6 presents results with abnormally optimistic tone of the MD&As (Panel A) and that of earnings press releases (Panel B). The coefficients on Post � Ninth Circuit � OptTone_MDA are significant and positive for AbnProd and RM. The coefficients on Post � Ninth Circuit � OptTone_EPR are significant and positive for all three measures (AbnDisx, AbnProd, and RM). These results indicate that following the decline in litigation risk with the ruling, REM increases are more pronounced in those Ninth Circuit firms whose managers use

abnormally optimistic tone in their MD&As and earnings press releases, potentially to justify their actions.

In further analyses, we test whether the post-ruling REM increases are more pronounced for Ninth Circuit firms that also

issued optimistic management forecasts. A firm-year is classified as having an optimistic forecast if the last forecast before the

fiscal period-end exceeds the consensus analyst expectations at the time. Hilary, Hsu, Segal, and Wang (2016) characterize

management forecasts that exceed analyst expectations as indicative of over-optimism. Table 6, Panel C presents the results

based on the indicator variable Optimistic Forecast. We find that the coefficient on Post � Ninth Circuit � Optimistic Forecast is positive in all three columns and statistically significant for AbnDisx and RM. The results are consistent with managers using

TABLE 6

Real Earnings Management (REM) and Optimistic/Misleading Disclosure

Panel A: Optimistic Tone in the MD&A

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.039 0.021 0.019 (1.62) (1.54) (0.99)

Post � Ninth Circuit � OptTone_MDA 0.019 0.054 0.038 (0.59) (3.06)*** (1.73)*

Post � OptTone_MDA �0.009 �0.014 �0.026 (�1.13) (�1.36) (�2.11)**

Ninth Circuit � OptTone_MDA �0.008 �0.039 �0.052 (�0.37) (�2.66)** (�2.23)**

OptTone_MDA 0.010 0.010 0.023 (1.57) (1.39) (2.50)**

Controls Yes Yes Yes

Observations 4,803 4,733 4,459

R 2

0.522 0.469 0.498

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

Panel B: Optimistic Tone in the Earnings Press Release

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit �0.021 �0.069 �0.049 (�0.66) (�3.02)*** (�0.92)

Post � Ninth Circuit � OptTone_EPR 0.120 0.131 0.189 (2.34)** (3.18)*** (2.42)**

Post � OptTone_EPR �0.010 0.009 �0.003 (�0.50) (0.28) (�0.08)

Ninth Circuit � OptTone_EPR �0.115 �0.110 �0.164 (�2.50)** (�2.85)*** (�2.24)**

OptTone_EPR 0.008 �0.016 �0.008 (0.46) (�0.51) (�0.21)

Controls Yes Yes Yes

Observations 3,954 3,901 3,717

R 2

0.707 0.621 0.700

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

(continued on next page)

Does Litigation Deter or Encourage Real Earnings Management? 265

The Accounting Review Volume 95, Number 3, 2020

more optimistic forecasts of future earnings to justify their opportunistic REM in the current period. Alternatively, it is also

possible that managers issue more optimistic earnings forecasts following the decline in litigation risk (Bourveau et al. 2018),

and then engage in increased REM to meet/beat those forecasts, consistent with Kasznik (1999). 22

TABLE 6 (continued)

Panel C: Optimistic Forecast

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.013 0.013 0.020 (3.28)*** (1.75)* (3.10)***

Post � Ninth Circuit � Optimistic Forecast 0.065 0.035 0.099 (1.71)* (1.64) (3.24)***

Post � Optimistic Forecast 0.008 0.008 0.012 (1.08) (0.86 ) (0.96 )

Ninth Circuit � Optimistic Forecast 0.011 �0.017 �0.014 (0.30) (�0.78) (�0.67)

Optimistic Forecast �0.011 �0.001 �0.009 (�1.91)* (�0.14) (�0.93)

Controls Yes Yes Yes

Observations 15,225 14,958 14,215

R 2

0.426 0.312 0.407

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

Panel D: Optimistic Disclosure

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.007 0.003 0.007 (1.54) (0.37) (0.97)

Post � Ninth Circuit � Optimistic Disclosure 0.036 0.032 0.061 (4.63)*** (2.97)*** (4.86)***

Post � Optimistic Disclosure �0.000 �0.004 �0.011 (�0.09) (�0.60) (�1.54)

Ninth Circuit � Optimistic Disclosure �0.027 �0.009 �0.044 (�4.46)*** (�1.09) (�3.94)***

Optimistic Disclosure 0.001 �0.006 0.002 (0.29) (�1.03) (0.38)

Controls Yes Yes Yes

Observations 15,225 14,958 14,215

R 2

0.426 0.312 0.407

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 6 examines cross-sectional variation in changes in REM following the Ninth Circuit Court ruling with optimistic disclosures. Panels A and B examine the abnormal optimistic tone of the MD&A section of the 10-K and of the earnings press release, respectively. Panel C focuses on optimistic management forecasts. Panel D focuses on a composite measure of optimistic disclosure. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

22 In additional untabulated tests, we find that post-ruling increases in REM at the Ninth Circuit firms, relative to non-Ninth Circuit firms, are significantly more pronounced for firms that issue an optimistic forecast for the first time (i.e., initiate an optimistic forecast).

266 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

Table 6, Panel D uses a composite indicator measure of disclosure optimism (Optimistic Disclosure) that is set equal to 1 if abnormal tone in the MD&A or the earnings press release was positive, or the manager issued an optimistic forecast prior to the

fiscal year-end. 23

Using this composite optimism metric, we observe that all three REM measures exhibit a significantly more

pronounced post-ruling increase for those firms that also issue optimistic disclosures. The results suggest that in the post-ruling

period, when managers manipulate real activities to report higher current earnings, they are more likely to support and

legitimize their actions with optimistic projections and explanations for their actions and/or reported performance.

Results on the Role of Incentives to Engage in Real Earnings Management

If the post-ruling increases in REM are symptomatic of heightened managerial myopia and opportunism, then the REM

increases should be more pronounced when managers have greater short-term incentives to report stronger financial

performance. Specifically, we examine the role of incentives to meet/beat earnings targets, as well as incentives arising from

insider selling.

To study whether increases in REM enable managers to meet/beat earnings targets, we augment our main model to include

variables that capture firms narrowly meeting or beating earnings expectations (Narrow Beat), beating these expectations comfortably (Large Beat), and narrowly missing expectations (Narrow Miss). We calculate the market expectation based on the consensus analyst forecast for firms with analyst following, and last year’s earnings per share (EPS) for firms that are not

followed by analysts. Narrow Beat is an indicator variable representing cases in which the actual EPS met or beat market expectations by less than two cents. Large Beat refers to firm-years in which the actual EPS exceeded the expectation by two cents or more. Narrow Miss is defined as actual EPS falling below the expectation by less than two cents. We then interact these variables with Post, Ninth Circuit, and Post � Ninth Circuit. Our interest is in the coefficients on Post � Ninth Circuit � Narrow Beat, Post � Ninth Circuit � Large Beat, and Post � Ninth Circuit � Narrow Miss. Firm-years in which actual earnings are below the expectation by a large margin are captured by Post � Ninth Circuit.

Table 7, Panel A reports the results from these tests. We find significant positive coefficients on Post � Ninth Circuit � Narrow Beat in all three columns, indicating that the increased use of REM after the ruling helps firms to meet/beat earnings targets by narrow margins. The coefficients on Post � Ninth Circuit � Large Beat also indicate significant post-ruling increases in REM for the firms that beat expectations by wider margins, even though the coefficients are economically of smaller magnitude

than for the Narrow Beat group. Since REM has to occur before fiscal year-end, managers cannot always calibrate accurately the extent to which they need to alter real actions to just meet/beat outstanding expectations, potentially resulting in beating

expectations by a larger margin. Interestingly, the coefficients on Post � Ninth Circuit � Narrow Miss are negative and significant in two of the three columns, indicating that fewer firms engage in REM and miss expectations by a narrow margin. We note that

the coefficient on Post � Ninth Circuit is not significantly different from zero; thus, there were no post-ruling changes in REM for Ninth Circuit firms missing analyst forecasts by a large margin. Together, our results suggest that reduced litigation risk

‘‘emboldened’’ managers: they were willing to engage in more REM and err on the side of comfortably beating expectations rather than risk missing these expectations narrowly. However, in the absence of short-term incentives, firms did not increase their REM.

We next examine whether post-ruling increases in REM at Ninth Circuit firms vary with insider selling. In order to distinguish

between routine and opportunistic trades by executives, we employ the methodology from Cohen, Malloy, and Pomorski (2012). We

then use the opportunistic insider trades to compute Insider Net Sell, which is defined as the percentage of shares outstanding sold net of the percentage of shares purchased by executives when their sales exceed purchases, and 0 otherwise.

24 We expand our main

model to include the interactions of this variable with Post, Ninth Circuit, and Post � Ninth Circuit. Consistent with our expectations, we find in Table 7, Panel B that the coefficient on Post � Ninth Circuit � Insider Net Sell is positive and significant, with p-values less than 0.05 in all three columns. The result indicates that the increases in REM are particularly pronounced when managers have

incentives to report strong performance to execute their opportunistic sales at more beneficial prices.

In untabulated additional analyses, we limit the sample to those firm-years in which managers had incentives to mislead

(i.e., they met or beat earnings targets or were net sellers of stock). Within this sample, we find that post-ruling increases in

REM at Ninth Circuit firms are significantly more pronounced if managers also engaged in optimistic disclosures. These results

confirm a significant relation between REM and optimistic disclosures in those firms in which managers are likely to have acted

opportunistically. 25

23 Since our measure is meant to capture any form of optimism, be it in management discussion and analysis (MD&As), earnings releases, or earnings forecasts, we rely on the maximum possible data points available in the dataset, that is, the union of the datasets used in the tests with each individual proxy for optimism.

24 Executives include officers of the firm, such as CEOs, CFOs, Senior VPs, and VPs.

25 Specifically, we repeat our tests reported in Table 6, Panel D on the subsample of firm-years in which managers had incentives to act opportunistically (i.e., they met or beat earnings expectations or were net sellers of stock); we continue to find a significant positive coefficient on Post � Ninth Circuit � Optimistic Disclosure for all three measures of REM.

Does Litigation Deter or Encourage Real Earnings Management? 267

The Accounting Review Volume 95, Number 3, 2020

It is possible in principle that REM represents, in part, managers’ business judgment at the time that it was undertaken. Our

evidence on meeting/beating earnings targets and insider selling is salient in this context, as it suggests that litigation risk

constrains REM, especially when it is likely to be viewed as opportunistic by shareholders and other external stakeholders.

Results on the Role of Governance Constraints on REM

In this section, we investigate cross-sectional variation in how litigation’s deterrence effect on REM varies cross-

sectionally with corporate governance. We expect litigation risk to be a more valuable deterrent when governance is weaker and

less effective in restraining managers from engaging in REM.

We focus on two aspects of governance: managerial entrenchment and institutional ownership. Greater managerial

entrenchment implies weaker governance, as both the board of directors and the market for corporate control find it difficult to

discipline entrenched managers (Bebchuk, Cohen, and Ferrell 2009). Thus, even if shareholders were to expect increased REM

at Ninth Circuit firms at the time of ruling, their ability to constrain REM is lower when managers are more entrenched. We,

thus, expect more pronounced post-ruling increases in REM at Ninth Circuit firms with higher managerial entrenchment.

Turning to our second measure of governance, institutional owners are sophisticated investors who are likely to monitor

managers’ real choices more closely and discourage actions detrimental for long-term value. They are also known to influence

TABLE 7

Real Earnings Management (REM) and Short-Term Incentives

Panel A: REM and Meet Beat

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.005 0.002 0.005 (1.03) (0.25) (0.54)

Post � Ninth Circuit � Narrow Beat 0.024 0.028 0.094 (1.85)* (1.75)* (3.24)***

Post � Ninth Circuit � Large Beat 0.019 0.020 0.028 (3.80)*** (2.29)** (2.89)***

Post � Ninth Circuit � Narrow Miss �0.033 0.011 �0.042 (�2.38)** (0.26) (�1.92)*

Post � Narrow Beat �0.007 �0.014 �0.024 (�0.91) (�1.38) (�1.78)*

Ninth Circuit � Narrow Beat 0.018 0.002 �0.012 (1.84)* (0.14) (�0.67)

Post � Large Beat �0.002 0.004 0.004 (�0.37) (0.86) (0.50)

Ninth Circuit � Large Beat 0.013 �0.007 �0.004 (1.64) (�1.36) (�0.36 )

Post � Narrow Miss �0.002 0.001 �0.007 (�0.14) (0.08) (�0.47)

Ninth Circuit � Narrow Miss 0.010 0.018 0.011 (0.77) (1.55) (0.73)

Narrow Beat 0.011 �0.002 0.008 (2.11)** (�0.27) (1.19)

Large Beat �0.001 �0.019 �0.018 (�0.41) (�4.09)*** (�2.85)***

Narrow Miss 0.005 �0.014 �0.005 (0.99) (�1.73)* (�0.49)

Controls Yes Yes Yes

Observations 15,225 14,958 14,215

R 2

0.427 0.315 0.409

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

(continued on next page)

268 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

managerial actions via either voice or the threat of exit and do not rely solely on litigation (Parrino, Sias, and Starks 2003;

Edmans 2009; Edmans and Manso 2011). Thus, institutional investors are likely to discourage REM because of its negative

consequences for firm value and future performance. At firms with low institutional ownership, the lack of this external

monitoring makes it easier for managers to inflate current earnings myopically via REM (Bushee 1998; Roychowdhury 2006 ).

We, thus, expect the decline in litigation risk after the court ruling to lead to greater increases in REM at Ninth Circuit firms

with lower institutional ownership.

As in Bebchuk et al. (2009), we use the entrenchment index (E-Index) to capture managerial entrenchment. E-Index is

based on six provisions, four of which constitute limitations on shareholders’ voting power, and the remaining two are

measures against hostile takeovers. To conduct this test, we augment Equation (1) with the following terms: E-Index, Post � E- Index, Ninth Circuit � E-Index, and Post � Ninth Circuit � E-Index. In this specification reported in Table 8, Panel A, the coefficient on Post � Ninth Circuit � E-Index is significant and positive in all three columns, implying that post-ruling increases in REM at the Ninth Circuit firms are stronger when managers are more entrenched.

To test how changes in REM depend on institutional ownership, we expand Equation (1) to supplement the main effect

of institutional owners (IO) with the following interaction terms: Post � IO, Ninth Circuit � IO, and Post � Ninth Circuit � IO. Table 8, Panel B reports the results of this test in Columns (1) through (3). The coefficient on Post � Ninth Circuit � IO is negative and significant with AbnProd and RM as dependent variables, providing some support for our prediction that the

post-ruling increases in REM at the Ninth Circuit firms are stronger when institutional ownership is lower. We next explore

the role of institutional owners in more depth by focusing on ownership of those institutions that are known to monitor and

exert positive influence on the firm’s governance: dedicated investors and quasi-indexers (Bushee 1998; Appel, Gormley,

and Keim 2016 ). We repeat our tests replacing IO with IO_Monitor, which is computed as the percentage of shares owned

by dedicated investors and quasi-indexers divided by the percentage of shares owned by all institutional investors. The

results are reported in Columns (4)–(6 ) of Table 8, Panel B. In all three columns, the coefficient on Post � Ninth Circuit � IO_Monitor is negative and statistically significant, with p-values less than 0.05. These results indicate that higher ownership

by institutions known to exert positive influence on the firm’s governance constrained post-ruling increases in REM at the

Ninth Circuit firms.

TABLE 7 (continued)

Panel B: REM and Insider Selling

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.011 0.012 0.016 (2.77)*** (1.65) (2.33)**

Post � Ninth Circuit � Insider Net Sell 0.041 0.020 0.070 (3.38)*** (2.10)** (5.17)***

Post � Insider Net Sell �0.001 0.010 �0.006 (�0.19) (1.79)* (�0.53)

Ninth Circuit � Insider Net Sell �0.010 0.002 �0.014 (�1.43) (0.19) (�1.77)*

Insider Net Sell �0.010 �0.001 �0.006 (�3.60)*** (�0.14) (�1.39)

Controls Yes Yes Yes

Observations 15,225 14,958 14,215

R 2

0.427 0.313 0.408

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 7 examines cross-sectional variation in changes in REM following the Ninth Circuit Court ruling with short-term incentives to overstate earnings. Panel A studies the incentives to meet or beat earnings target, while Panel B studies the incentives arising from insider selling. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

Does Litigation Deter or Encourage Real Earnings Management? 269

The Accounting Review Volume 95, Number 3, 2020

TABLE 8

Real Earnings Management (REM) Changes and Corporate Governance

Panel A: Managerial Entrenchment

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit �0.010 �0.031 �0.041 (�1.17) (�1.23) (�1.53)

Post � Ninth Circuit � E-Index 0.009 0.012 0.020 (3.20)*** (2.08)** (2.32)**

Post � E-Index 0.000 �0.003 �0.001 (0.31) (�1.40) (�0.61)

Ninth Circuit � E-Index �0.009 0.007 0.005 (�2.59)** (0.96) (0.56 )

E-Index 0.003 0.005 0.007 (0.87) (1.89)* (1.38)

Controls Yes Yes Yes

Observations 4,701 4,631 4,538

R 2

0.485 0.308 0.433

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

Panel B: Institutional Ownership

InstOwn ¼ IO InstOwn ¼ IO_Monitor

(1) AbnDisx

(2) AbnProd

(3) RM

(4) AbnDisx

(5) AbnProd

(6) RM

Post � Ninth Circuit 0.021 0.038 0.054 0.063 0.066 0.102 (2.02)** (2.95)*** (3.73)*** (4.04)*** (2.16)** (5.24)***

Post � Ninth Circuit � InstOwn �0.004 �0.066 �0.074 �0.066 �0.076 �0.110 (�0.13) (�2.68)*** (�1.83)* (�3.04)*** (�2.13)** (�4.19)***

Post � InstOwn 0.026 0.005 0.024 0.006 �0.010 �0.010 (2.22)** (0.49) (1.50) (0.54) (�0.87) (�0.64)

Ninth Circuit � InstOwn �0.043 0.032 �0.003 0.003 0.020 �0.015 (�1.06) (0.60) (�0.03) (0.19) (0.92) (�0.65)

InstOwn �0.027 �0.017 �0.044 0.036 0.004 0.052 (�2.01)** (�1.39) (�2.33)** (2.88)*** (0.39) (3.30)***

Controls Yes Yes Yes Yes Yes Yes

Observations 15,225 14,958 14,215 15,225 14,958 14,215

R 2

0.421 0.312 0.405 0.428 0.313 0.408

Firm FE Yes Yes Yes Yes Yes Yes

Year FE Yes Yes Yes Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 8 examines cross-sectional variation in changes in REM following the Ninth Circuit Court ruling with corporate governance attributes. Panel A studies the role of managerial entrenchment, while Panel B studies the role of institutional ownership. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

270 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

V. ADDITIONAL ANALYSES

Results on Changes in REM and Accrual-Based Earnings Management

Although we control for accruals management in all our tests, in this section, we test for an interacted effect: does REM

increase more for those Ninth Circuit firms that report high discretionary accruals following the ruling? As a preliminary

analysis, we test whether accruals management in Ninth Circuit firms exhibits an increase relative to that of control firms

following the ruling. Figure 1, Panel B plots abnormal accruals in event time. As the figure reveals, abnormal accruals exhibits

a significant increase in year 2 following the ruling, but the difference in abnormal accruals between treatment and control firms

is not significant in years 3 and 4.

Turning to our main analysis in this section, we augment our main model to include not only AbnAcc, but also its interaction terms with Post, Ninth Circuit, and Post � Ninth Circuit. Table 9 reports the results from these tests. The coefficients on Post � Ninth Circuit � AbnAcc are significantly negative in all three columns, implying that REM increases more for Ninth Circuit firms with lower discretionary accruals. Together with the results in Hopkins (2018), those in Table 9 suggest that REM and accruals both increased following the ruling, but the increases were negatively correlated in the cross-section of Ninth

Circuit firms. In further untabulated tests, we find that increases in REM were particularly pronounced for Ninth Circuit firms

that had limited ability to manipulate accruals, as measured by high net operating assets (Barton and Simko 2002; Zang

2012). 26

Results on the Importance of Firm-Specific Litigation Risk

Prior literature documents that litigation risk can vary across firms based on characteristics such as industry membership,

return volatility, and growth (Francis, Philbrick, and Schipper 1994; Rogers and Stocken 2005; Kim and Skinner 2012). We

expect that increases in REM accompanying the Ninth Circuit Court ruling should be the most pronounced for firms most

constrained by litigation risk in the year before the ruling.

We modify Equation (1) to include a firm-specific litigation risk measure (LitRisk) introduced by Kim and Skinner (2012), along with its interaction terms with the following variables: Post, Ninth Circuit, and Post � Ninth Circuit. Table 10 presents the results. The coefficient on Post � Ninth Circuit � LitRisk is positive and statistically significant in the first two columns, providing some support for the notion that greater post-ruling REM increases in Ninth Circuit firms occur at those firms that

were more constrained by litigation risk before the ruling. 27

Other Untabulated Analyses

REM Escalation

In untabulated analyses, we examine whether the decline in litigation risk following the 1999 ruling motivated managers to

escalate the level of REM from year to year. In the presence of litigation risk, we expect managers to limit any escalation of

REM escalation from one period to the next, out of concern that this escalation can cause them to issue misleading statements

that, upon being uncovered, attract class action lawsuits. Thus, following the decline in litigation risk with the 1999 ruling, we

expect Ninth Circuit firms to exhibit greater REM escalation relative to other firms. To measure REM escalation, we use an

indicator variable equal to 1 if a firm moves from tercile 1 of an REM proxy last year to tercile 2 or 3 in the current year for that

proxy; or moves from tercile 2 of last year to tercile 3 in the current year; or belonged to tercile 3 last year and remains in tercile

3 in the current year. Untabulated results confirm that Ninth Circuit firms exhibit greater year-to-year escalation in REM

following the decline in litigation risk with the 1999 ruling.

Derivative Lawsuits and Real Earnings Management

To increase the generalizability of our results, we next examine shocks to the ability of shareholders to file derivative

lawsuits. Unlike the Ninth Circuit Court ruling, shocks to the filing of derivative lawsuits were staggered over time, which

addresses potential concerns that the deterrence effect of litigation on REM is limited to one specific time period or event. 28

26 Firms with high net operating assets (NOA) do not show significant increases in abnormal accruals following the ruling. The evidence is consistent with high-NOA firms having reduced flexibility to increase, or even sustain, aggressive accrual choices following the ruling and resorting to REM instead.

27 Our results are stronger and statistically significant in all three columns when we replace continuous LitRisk with a binary variable, HighLitRisk, which takes the value of 1 for firms with above-median LitRisk, suggesting that the effect of firm-specific litigation risk is not perfectly linear.

28 Derivative lawsuits allow shareholders to sue managers and/or directors on behalf of the corporation (rather than on behalf of shareholders) for a breach of fiduciary duty. Plaintiffs almost always prefer securities class action litigation, but derivative lawsuits are often filed, in addition, whenever there are legal grounds for initiating both (Erickson 2010; Appel 2016 ).

Does Litigation Deter or Encourage Real Earnings Management? 271

The Accounting Review Volume 95, Number 3, 2020

Staggered adoption of universal demand (UD) laws in 23 states between 1989 and 2005 made the filing of derivative lawsuits

significantly more difficult by requiring that shareholders seek the board’s approval before initiating litigation (Appel 2016 ).

The passage of UD laws reduced the incidence of litigation and contributed to poorer governance and higher cost of capital for

affected firms (Appel 2016; Houston et al. 2015a). We create a sample of firm-years encompassing years 1985–2009, which

ranges from four years before the first state adoption of UD laws to four years after the last state adoption. Untabulated results

from staggered difference-in-differences tests centered on UD law adoption reveal significant increases in REM for firms when

their respective states of incorporation adopt UD laws. Consistent with our main results based on securities class action

litigation, we conclude that litigation serves a deterrence role with respect to real earnings management and that REM increases

when that role is reduced.

VI. CONCLUSION

Shareholders can pursue various courses of action to prevent managers from violating their fiduciary duty and engaging in

actions that transfer wealth away from shareholders. These actions include exercising governance through vote or exit (Parrino

et al. 2003; Edmans 2009; Edmans and Manso 2011; Appel, Gormley, and Keim 2016 ). In certain circumstances, shareholders

are also able to file lawsuits against the firm’s management and/or directors. Because litigation has serious reputation and career

implications for management and directors (Srinivasan 2005; Brochet and Srinivasan 2014), it also serves a deterrence role.

While prior literature has investigated the disciplining effect of litigation on management forecasts, as well as GAAP violations

(Skinner 1994; Houston et al. 2015b; Hopkins 2018), no prior research has explored whether this effect of litigation extends to

real earnings management (REM). Establishing whether it does is important because the impact of REM on future performance

is known to be more negative than that of accruals and, at the same time, REM is more opaque and difficult to litigate against

because of the business judgment principle.

In this paper, we provide robust evidence that litigation deters REM. Our results further indicate that litigation’s deterrence

effect arises because of REM’s link with excessively optimistic disclosures. Managers that manipulate real actions bear the risk

that they have to misrepresent the true intent and circumstances underlying those actions, and these misrepresentations can

TABLE 9

Real Earnings Management (REM) and Accrual Earnings Management

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.015 0.014 0.023 (3.97)*** (1.80)* (3.33)***

Post � Ninth Circuit � AbnAcc �0.175 �0.258 �0.374 (�1.71)* (�1.96 )* (�2.48)**

Post � AbnAcc �0.069 �0.008 �0.058 (�0.83) (�0.14) (�0.63)

Ninth Circuit � AbnAcc 0.116 0.148 0.200 (1.50) (1.57) (1.49)

AbnAcc 0.089 0.108 0.203 (1.66) (1.71)* (2.90)***

Controls Yes Yes Yes

Observations 15,225 14,958 14,215

R 2

0.428 0.313 0.408

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 9 examines cross-sectional variation in changes in REM following the Ninth Circuit Court ruling with accrual-based earnings management (AbnAcc). The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

272 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

become subject to lawsuits. Further, we find that litigation’s role in deterring REM is more salient when managerial incentives

to overstate earnings are stronger and corporate governance is weaker.

We caution that our findings need not imply that an increase in litigation risk is necessarily the most efficient way to

constrain managerial opportunism under all circumstances. For example, Lin et al. (2017) find that the threat of litigation can

constrain risk-taking and innovation, which suggests that increased litigation risk imposes costs. Our results should be

interpreted as indicating a partial benefit of litigation risk. When managers face short-term incentives to myopically inflate

earnings, the prospect of ex post settling up via litigation effectively deters real earnings management. A second caveat to bear

in mind is that the effects of a decline in litigation risk with the 1999 ruling may not necessarily have been permanent. As Crane

and Koch (2018) note, alternative governance mechanisms, such as higher institutional ownership, possibly stepped in to

discipline the expected increase in managerial opportunism and myopia in Ninth Circuit firms after the 1999 ruling. However, if

any of these governance changes were unconditionally optimal, then they could have taken place even when litigation risk was

high, that is, prior to the ruling. An increase in REM, even one that is the strongest in the year immediately following the ruling,

implies that litigation risk constrains myopic real actions that would be costlier to discipline via alternative governance

mechanisms.

REFERENCES

Ali, A., and W. Zhang. 2015. CEO tenure and earnings management. Journal of Accounting and Economics 59 (1): 60–79. https://doi.org/

10.1016/j.jacceco.2014.11.004

Appel, I. 2016. Governance by litigation. Working paper, Boston College. https://dx.doi.org/10.2139/ssrn.2532278

Appel, I., T. Gormley, and D. Keim. 2016. Passive investors, not passive owners. Journal of Financial Economics 121 (1): 111–141.

https://doi.org/10.1016/j.jfineco.2016.03.003

AT&T Corporation Class Action. 2001. In Re AT&T Corporation Securities Litigation, No. 3:00-CV-05364.

TABLE 10

Real Earnings Management (REM) Changes and Litigation Risk

(1) AbnDisx

(2) AbnProd

(3) RM

Post � Ninth Circuit 0.009 0.009 0.024 (1.51) (0.95) (2.15)**

Post � Ninth Circuit � LitRisk 0.030 0.038 0.044 (1.72)* (1.68)* (1.12)

Post � LitRisk �0.019 0.012 0.002 (�1.60) (0.80) (0.10)

Ninth Circuit � LitRisk �0.020 0.038 0.011 (�0.93) (1.64) (0.32)

LitRisk 0.060 �0.019 0.036 (4.97)*** (�1.40) (2.28)**

Controls Yes Yes Yes

Observations 13,244 12,939 12,421

R 2

0.449 0.306 0.406

Firm FE Yes Yes Yes

Year FE Yes Yes Yes

*, **, *** Denote two-tailed significance levels at 10 percent, 5 percent, and 1 percent, respectively. Table 10 examines cross-sectional variation in changes in REM following the Ninth Circuit Court ruling with firm-specific litigation risk measured in the year before the ruling. The sample period is from 1995 to 2003, excluding 1999, the year of the Ninth Circuit ruling. We exclude the financial and utility industries (SIC 4000–4999 and SIC 6000–6999). We exclude penny stocks (share price less than $1) and firms with missing state of headquarters location. Only firms with at least one year of data in both the pre- and post-event periods are included in the sample. All regressions contain the same set of control variables as in Table 3. t-statistics are presented beneath the coefficients within parentheses. Standard errors are corrected for heteroscedasticity and are clustered at state of location level. All variables are defined in Appendix A.

Does Litigation Deter or Encourage Real Earnings Management? 273

The Accounting Review Volume 95, Number 3, 2020

Badertscher, B. 2011. Overvaluation and the choice of alternative earnings management mechanisms. The Accounting Review 86 (5): 1491–1518. https://doi.org/10.2308/accr-10092

Barton, J., and P. Simko. 2002. The balance sheet as an earnings management constraint. The Accounting Review 77 (s-1): 1–27. https:// doi.org/10.2308/accr.2002.77.s-1.1

Bebchuk, L., A. Cohen, and A. Ferrell. 2009. What matters in corporate governance? Review of Financial Studies 22 (2): 783–827. https:// doi.org/10.1093/rfs/hhn099

Bochkay, K., and C. B. Levine. 2017. Using MD&A to improve earnings forecasts. Journal of Accounting, Auditing and Finance 32: 1– 25.

Bourveau, T., Y. Lou, and R. Wang. 2018. Shareholder litigation and corporate disclosure: Evidence from derivative lawsuits. Journal of Accounting Research 56 (3): 797–842. https://doi.org/10.1111/1475-679X.12191

Bristol-Myers Squibb Class Action. 2002. In Re Bristol-Myers Squibb Company, No. 02-CV-2251. Bristol-Myers Squibb Settlement Agreement. 2004. In Re Bristol-Myers Squibb Company, No. 02-CV-2251. Brochet, F., and S. Srinivasan. 2014. Accountability of independent directors: Evidence from firms subject to securities litigation. Journal

of Financial Economics 111 (2): 430–449. https://doi.org/10.1016/j.jfineco.2013.10.013 Brown, S. V., and J. W. Tucker. 2011. Large-sample evidence on firms’ year-over-year MD&A modifications. Journal of Accounting

Research 49 (2): 309–346. https://doi.org/10.1111/j.1475-679X.2010.00396.x Bruns, W., and K. Merchant. 1990. The dangerous morality of managing earnings. Management Accounting 72: 22–25. Bushee, B. 1998. The influence of institutional investors on myopic R&D investment behavior. The Accounting Review 73: 305–333. Cazier, R. A., T. E. Christensen, K. J. Merkley, and J. S. Treu. 2017. Litigation risk and non-GAAP reporting. Available at: https://papers.

ssrn.com/sol3/papers.cfm?abstract_id¼2928260 Chen, F., Q. Li, A. Lo, and L. Xu. 2018. Universal demand laws and the monitoring device role of accounting conservatism. Available at:

https://papers.ssrn.com/Sol3/papers.cfm?abstract_id¼2963162 Chen, X., Q. Cheng, A. Lo, and X. Wang. 2015. CEO contractual protection and managerial short-termism. The Accounting Review 90

(5): 1871–1906. https://doi.org/10.2308/accr-51033.

Cheng, Q., J. Lee, and T. Shevlin. 2016. Internal governance and real earnings management. The Accounting Review 91 (4): 1051–1085. https://doi.org/10.2308/accr-51275

Coca Cola Co. Class Action. 2001. Carpenters Health & Welfare Fund v. Coca-Cola Co., No. 00-CV-2838. Coca Cola Co. Settlement Agreement. 2008. Carpenters Health & Welfare Fund v. Coca-Cola Co., No. 00-CV-2838. Cohen, D., and P. Zarowin. 2010. Accrual-based and real earnings management activities around seasoned equity offerings. Journal of

Accounting and Economics 50 (1): 2–19. https://doi.org/10.1016/j.jacceco.2010.01.002 Cohen, D., A. 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. https://doi.org/10.2308/accr.2008.83.3.757 Cohen, L., C. Malloy, and L. Pomorski. 2012. Decoding inside information. Journal of Finance 67 (3): 1009–1043. https://doi.org/10.

1111/j.1540-6261.2012.01740.x

Cox, J., R. Thomas, and L. Bai. 2009. Do differences in pleading standards cause forum shopping in securities class actions?: Doctrinal

and empirical analyses. Wisconsin Law Review 2009: 421–453. Crane, A., and A. Koch. 2018. Shareholder litigation and ownership structure: Evidence from a natural experiment. Management Science

64 (1): 5–23. https://doi.org/10.1287/mnsc.2016.2561

Dechow, P. M., A. G. Sloan, and A. Sweeney. 1996. Causes and consequences of earnings manipulation: An analysis of firms subject to

enforcement actions by the SEC. Contemporary Accounting Research 13 (1): 1–36. https://doi.org/10.1111/j.1911-3846.1996. tb00489.x.

Edmans, A. 2009. Blockholder trading, market efficiency, and managerial myopia. Journal of Finance 64 (6 ): 2481–2513. https://doi.org/ 10.1111/j.1540-6261.2009.01508.x

Edmans, A., and G. Manso. 2011. Governance through trading and intervention: A theory of multiple blockholders. Review of Financial Studies 24 (7): 2395–2428. https://doi.org/10.1093/rfs/hhq145

Erickson, J. 2010. Corporate governance in the courtroom: An empirical analysis. William & Mary Law Review 51: 1749–1832. Francis, J., D. Philbrick, and K. Schipper. 1994. Shareholder litigation and corporate disclosures. Journal of Accounting Research 32 (2):

137–164. https://doi.org/10.2307/2491279

General Motors Co. Class Action. 2012. Scott et al. v. Whitacre et al., No. 14-CV-05124. Gibney, B. C. 2001. The end of the unbearable lightness of pleading: Scienter after Silicon Graphic. UCLA Law Review 48: 973–1015. Graham, J. R., C. R. Harvey, and S. Rajgopal. 2005. The economic implications of corporate financial reporting. Journal of Accounting

and Economics 40 (1/3): 3–73. https://doi.org/10.1016/j.jacceco.2005.01.002 Green Mountain Coffee Class Action. 2012. Louisiana Municipal Police Employees’ Retirement System v. Green Mountain Coffee

Roasters, Inc. et al., No. 11-CV-00289. Green Mountain Coffee Class Action. 2015. Employees’ Retirement System v. Green Mountain Coffee Roasters Inc. et al., No. 14-CV-

00199.

Green Mountain Coffee Settlement Agreement. 2018. Louisiana Municipal Police Employees’ Retirement System v. Green Mountain Coffee Roasters, Inc. et al., No. 11-CV-00289.

274 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

Grundfest, J. A., and A. C. Pritchard. 2002. Statutes with multiple personality disorders: The value of ambiguity in statutory design and

interpretation. Stanford Law Review 54 (4): 627–736. Heider, F., and A. Ljungqvist. 2015. As certain as debt and taxes: Estimating the tax sensitivity of leverage from state tax changes.

Journal of Financial Economics 118 (3): 684–712. https://doi.org/10.1016/j.jfineco.2015.01.004 Hilary, G., and K. W. Hui. 2009. Does religion matter in corporate decision making in America? Journal of Financial Economics 93 (3):

455–473. https://doi.org/10.1016/j.jfineco.2008.10.001

Hilary, G., C. Hsu, B. Segal, and R. Wang. 2016. The bright side of managerial over-optimism. Journal of Accounting and Economics 62 (1): 46–64. https://doi.org/10.1016/j.jacceco.2016.04.001

Hopkins, J. 2018. Do securities class actions deter misreporting? Contemporary Accounting Research 35 (4): 2030–2057. https://doi.org/ 10.1111/1911-3846.12367

Hospira Inc. Class Action. 2011. City of Sterling Heights General Employees’ Retirement System et al. v. Hospira, Inc. et al., No. 11-CV- 08332-AJS.

Hospira Inc. Settlement Agreement. 2014. City of Sterling Heights General Employees’ Retirement System et al. v. Hospira, Inc. et al., No. 11-CV-08332-AJS.

Houston, J., C. Lin, and L. Xie. 2015a. Shareholder protection and the cost of capital. Available at: https://papers.ssrn.com/sol3/papers. cfm?abstract_id¼2661292

Houston, J., C. Lin, S. Liu, and L. Wei. 2015b. Shareholder litigation and management earnings forecasts. Available at: https://www. researchgate.net/publication/315491187_Shareholder_Litigation_and_Management_Earnings_Forecasts

Huang, X., S. H. Teoh, and Y. Zhang. 2014. Tone management. The Accounting Review 89 (3): 1083–1113. https://doi.org/10.2308/accr- 50684

Johnson, M., K. Nelson, and A. Pritchard. 1999. In Re Silicon Graphics Inc.: Shareholder wealth effects resulting from the interpretation of the Private Securities Litigation Reform Act’s pleading standard. S. California Law Review 73: 773–810.

Kasznik, R. 1999. On the association between voluntary disclosure and earnings management. Journal of Accounting Research 37 (1): 57–81. https://doi.org/10.2307/2491396

Kim, I., and D. J. Skinner. 2012. Measuring securities litigation risk. Journal of Accounting and Economics 53 (1/2): 290–310. https://doi. org/10.1016/j.jacceco.2011.09.005

Kothari, S. P., N. Mizik, and S. Roychowdhury. 2016. Managing for the moment: The role of earnings management via real activities

versus accruals in SEO valuation. The Accounting Review 91 (2): 559–586. https://doi.org/10.2308/accr-51153 Lin, C., S. Liu, and G. Manso. 2017. Shareholder litigation and corporate innovation. Available at: http://faculty.haas.berkeley.edu/

manso/litigation.pdf

Lo, K., F. Ramos, and R. Rogo. 2017. Earnings management and annual report readability. Journal of Accounting and Economics 63 (1): 1–25. https://doi.org/10.1016/j.jacceco.2016.09.002

Loughran, T., and B. McDonald. 2011. When is a liability not a liability? Textual analysis, dictionaries, and 10-Ks. Journal of Finance 66 (1): 35–65. https://doi.org/10.1111/j.1540-6261.2010.01625.x

Mayew, W. J., M. Sethuraman, and M. Venkatachalam. 2015. MD&A disclosure and the firm’s ability to continue as a going concern.

The Accounting Review 90 (4): 1621–1651. https://doi.org/10.2308/accr-50983 McGuire, S., T. Omer, and N. Sharp. 2012. The impact of religion on financial reporting irregularities. The Accounting Review 87 (2):

645–673. https://doi.org/10.2308/accr-10206

Muslu, V., S. Radhakrishnan, K. R. Subramanyam, and D. Lim. 2015. Forward-looking MD&A disclosures and the information

environment. Management Science 61 (5): 931–948. https://doi.org/10.1287/mnsc.2014.1921 Parrino, R., R. W. Sias, and L. T. Starks. 2003. Voting with their feet: Institutional ownership changes around forced CEO turnovers.

Journal of Financial Economics 68 (1): 3–46. https://doi.org/10.1016/S0304-405X(02)00247-7 Pritchard, A. C., and H. A. Sale. 2005. What counts as fraud? An empirical study of motions to dismiss under the Private Securities

Litigation Reform Act. Journal of Empirical Legal Studies 2 (1): 125–149. https://doi.org/10.1111/j.1740-1461.2005.00033.x Raymond, N. 2014. Hospira agrees to $60 mln settlement over quality control problems. Reuters (March 27). Available at: https://www.

reuters.com/article/hospira-lawsuit-idUSL1N0MO2BS20140327

Rogers, J., and P. Stocken. 2005. Credibility of management forecasts. The Accounting Review 80 (4): 1233–1260. https://doi.org/10. 2308/accr.2005.80.4.1233

Rogers, J., A. Van Buskirk, and S. Zechman. 2011. Disclosure tone and shareholder litigation. The Accounting Review 86 (6 ): 2155– 2183. https://doi.org/10.2308/accr-10137

Roychowdhury, S. 2006. Earnings management through real activities manipulation. Journal of Accounting and Economics 42 (3): 335– 370. https://doi.org/10.1016/j.jacceco.2006.01.002

Skinner, D. J. 1994. Why firms voluntarily disclose bad news. Journal of Accounting Research 32 (1): 38–60. https://doi.org/10.2307/ 2491386

Skinner, D. J. 1997. Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics 23 (3): 249–282. https://doi. org/10.1016/S0165-4101(97)00010-4

Srinivasan, S. 2005. Consequences of financial reporting failure for outside directors: Evidence from accounting restatements and audit

committee members. Journal of Accounting Research 43 (2): 291–334. https://doi.org/10.1111/j.1475-679x.2005.00172.x

Does Litigation Deter or Encourage Real Earnings Management? 275

The Accounting Review Volume 95, Number 3, 2020

Stubben, S. 2010. Discretionary revenues as a measure of earnings management. The Accounting Review 85 (2): 695–717. https://doi.org/ 10.2308/accr.2010.85.2.695

Tasker, S. 1998. Bridging the information gap: Quarterly conference calls as a medium for voluntary disclosure. Review of Accounting Studies 3 (1/2): 137–167. https://doi.org/10.1023/A:1009684502135

Vorst, P. 2016. Real earnings management and long-term operating performance: The role of reversals in discretionary investment cuts.

The Accounting Review 91 (4): 1219–1256. https://doi.org/10.2308/accr-51281 Zang, A. Y. 2012. Evidence on the trade-off between real activities manipulation and accrual-based earnings management. The

Accounting Review 87 (2): 675–703. https://doi.org/10.2308/accr-10196

276 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

APPENDIX A

Variable Definitions

Variables Definition

Ninth Circuit Indicator equal to 1 if firms are located in Ninth Circuit states. Source: Compustat Post Indicator equal to 1 for years 2000–2003, and 0 for years 1995–1998. Source: Compustat ROA Operating income over lagged total assets. Source: Compustat Total Assets Total assets (in millions). Source: Compustat Firm Size Log of total assets. Source: Compustat Dividend Dividend over total assets. Source: Compustat CAPEX Capital expenditures over total assets. Source: Compustat Leverage Short-term debt plus long-term debt, divided by total assets. Source: Compustat Cash Cash over total assets. Source: Compustat Debt Issue Long-term debt issuance minus long-term debt reduction, divided by total assets. Source: Compustat Equity Issue Sale of common stock divided by shareholder equity. Source: Compustat Acquisition Number of acquisitions. Source: SDC IO Percent of shares outstanding held by institutional investors. Source: Thomson Reuters IO_Monitor Percentage of shares owned by dedicated investors and quasi-indexers divided by the percentage of

shares owned by all institutional investors.

GDP Growth State GDP growth rate. Source: Census Unemployment County unemployment rate. Source: Census Income County income per capita. Source: Census Population Log of county population. Source: Census Gender Percentage of female population at county level. Source: Census Ethnicity Percentage of white population at county level. Source: Census Education Percentage of people with at least tertiary education at county level. Source: Census Vote Democrat Percentage of people voting for the Democrat candidate in the last presidential election. Source: Census AbnProd Following Kothari et al. (2016), we estimate abnormal production cost as residuals from the following

first-order autoregressive model incorporating fixed effects:

Prodi,t ¼ a0 þ a1Prodi,t�1 þ a21/Assetsi,t�1 þ a3Salesi,t þ a4DSalesi,t þ a5DSalesi,t�1 þ ei,t, where Prodi,t is the sum of COGS and change in inventory during the year scaled by lagged total assets; Prodi,t�1 is its lagged value; Assetsi,t�1 is lagged total assets; Salesi,t is sales during the year scaled by lagged total assets; DSalesi,t is the sale growth scaled by lagged total assets; and DSalesi,t�1 is its lagged value.

To control for year-specific and firm-specific effects that induce model misspecification, we adapt the

model proposed by Kothari et al. (2016) for estimating abnormal production cost. As described in

Kothari et al. (2016): ‘‘First, every firm’s production cost is differenced from the cross-sectional mean

for that year. Second, for every firm, the annual deviation of production cost from the cross-sectional

mean is differenced from the corresponding deviation in the previous year. The explanatory variables

in the model are also differenced twice in the same manner. The model is estimated every year.’’ The

firm-year residual minus the mean value of the residual across all years for the corresponding firm

yields abnormal production costs. Source: Compustat

AbnDisx Following Kothari et al. (2016), we estimate abnormal discretionary expenditures as residuals from the following first-order autoregressive model incorporating fixed-effects:

DiscExpi,t ¼ a0 þ a1DiscExpi,t�1 þ a21/Assetsi,t�1 þ a3 Salesi,t þ ei,t, where DiscExpi,t is the sum of advertising expenses, R&D expenses, and SG&A expenses, all scaled by lagged total assets; DiscExpi,t�1 is its lagged value; Assetsi,t�1 is lagged total assets; and Salesi,t is sales during the year scaled by lagged total assets.

Please see notes on AbnProd for a description of how the model is estimated. Following Kothari et al. (2016), ‘‘the firm-year residual minus the mean value of the residual across all years for the

corresponding firm, times �1, yields abnormal discretionary expense or AbnDisx for that firm-year. Higher values of AbnDisx represent greater cuts in discretionary expenses and more earnings management.’’ Source: Compustat

RM AbnDisx þ AbnProd (continued on next page)

Does Litigation Deter or Encourage Real Earnings Management? 277

The Accounting Review Volume 95, Number 3, 2020

APPENDIX A (continued)

Variables Definition

AbnAcc Following Kothari et al. (2016), we estimate abnormal discretionary accruals with a fixed effect first- order autoregressive model:

TAi,t ¼ a0 þ a1TAi,t�1 þ a21/Assetsi,t�1 þ a3DRevi,t þ a4PPEi,t þ a5CFOi,t�1 þ a6CFOi,t þ a7CFOi,t �1 þ ei,t, where TAi,t is total accruals of firm i in year t, scaled by average total assets; TAi,t�1 is its lagged value; DRevi,t is the annual change in revenues scaled by average total assets; PPEi,t is property, plant, and equipment scaled by average total assets; and CFO is operating cash flow, scaled by average total assets.

Please see notes on AbnProd for a description of how the model is estimated. Source: Compustat D(t ¼ �1) An indicator equal to 1 for year 1998, and 0 otherwise. Source: Compustat D(t ¼ �2) An indicator equal to 1 for year 1997, and 0 otherwise. Source: Compustat D(t ¼ �3) An indicator equal to 1 for year 1996, and 0 otherwise. Source: Compustat D(t ¼ 1) An indicator equal to 1 for year 2000, and 0 otherwise. Source: Compustat D(t ¼ 2) An indicator equal to 1 for year 2001, and 0 otherwise. Source: Compustat D(t ¼ 3) An indicator equal to 1 for year 2002, and 0 otherwise. Source: Compustat D(t ¼ 4) An indicator equal to 1 for year 2003, and 0 otherwise. Source: Compustat OptTone_MDA An indicator equal to 1 if abnormal tone of the Management Discussion and Analysis Section (MD&A)

of the 10-Ks is positive, and 0 otherwise. Tone is measured as the difference between the count of

positive and negative words scaled by the total number of words, following the Loughran and

McDonald (2011) methodology. Abnormal tone is measured following Huang et al. (2014), as a

residual from a regression of tone on earnings, returns, size, book-to-market, volatility of returns and

earnings, firm age, number of business and geographic segments, the incidence of losses, change in

earnings, analyst forecast of earnings and analyst forecast error. Source: MD&A Tone data from

Bochkay and Levine (2017), generously provided by Khrystyna Bochkay at the University of Miami,

Compustat, CRSP, I/B/E/S

OptTone_EPR An indicator that takes the value of 1 if abnormal tone of earnings announcement press releases is positive, and 0 otherwise. Abnormal tone is measured following Huang et al. (2014), following the

same procedure as in OptTone_MDA. Source: Earnings press release tone data from Huang et al. (2014), generously provided by Siew Hong Teoh at the University of California, Irvine, Compustat,

CRSP, I/B/E/S

Optimistic Forecast An indicator equal to 1 if the last annual management forecast before fiscal year-end beats median analyst consensus computed over the past 90 days. Source: First Call

Optimistic Disclosure An indicator equal to 1 if abnormal tone in the MD&A or the earnings press release was positive, or the manager issued an optimistic forecast prior to the fiscal year-end, and 0 otherwise.

Narrow Beat An indicator equal to 1 if actual EPS meets or beats earnings target by less than two cents, and 0 otherwise. We define earnings target using analyst consensus forecasts within first 90 days since the

last earnings announcement. For firms without analyst coverage, we define earnings target using

earnings from last year. Source: I/B/E/S

Large Beat An indicator equal to 1 if actual EPS meets or beats earnings target by at least two cents, and 0 otherwise. We define earnings target using analyst consensus forecasts within the first 90 days since

the last earnings announcement. For firms without analyst coverage, we define earnings target using

earnings from last year. Source: I/B/E/S

Narrow Miss An indicator equal to 1 if actual EPS misses earnings target by less than two cents, and 0 otherwise. We define earnings target using analyst consensus forecasts within the first 90 days since the last

earnings announcement. For firms without analyst coverage, we define earnings target using earnings

from last year. Source: I/B/E/S

Insider Net Sell Insider sales minus insider purchases if insider sales are greater than insider purchases, and 0 otherwise. Insider sales ( purchases) are computed as the percentage of shares sold ( purchased) by executives.

Insider executives exclude directors and large owners (that is, those with . 10 percent

blockholdings), and include officers of the firm such as CEOs, CFOs, COOs, Senior VPs, and VPs.

We follow Cohen et al. (2012) and exclude routine transactions in calculating insider sales and

purchases. Source: Thomson Reuters

E-Index E-Index is based on six provisions, four of which constitute limitations on shareholders’ voting power, and the remaining two are measures against hostile takeovers. Source: IRRC

LitRisk LitRisk is the estimated litigation probability from Kim and Skinner (2012) model. Source: Compustat

278 Huang, Roychowdhury, and Sletten

The Accounting Review Volume 95, Number 3, 2020

Copyright of Accounting Review is the property of American Accounting Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.