Earnings Management and Corporate Investment Decisions

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Earnings Management and Corporate Investment Decisions Brandon Julio and Youngsuk Yook 2016-086 Please cite this paper as: Julio, Brandon, and Youngsuk Yook (2016). “Earnings Management and Corporate Investment Decisions,” Finance and Economics Discussion Series 2016-086. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2016.086. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Earnings Management and Corporate Investment Decisions BRANDON JULIO∗ University of Oregon YOUNGSUK YOOK† Federal Reserve Board of Governors October 2016 ABSTRACT We investigate the relationship between earnings management and the efficiency of corporate investment decisions. Using discretionary accruals to measure intertemporal transfers ofearnings, weshowthatearnings management exhibits aconcave relationship withtheinvestmentsensitivitytoinvestmentopportunities asmeasuredbyTobin’sQ.We findthattheassociation isconcentrated amonghighQfirms. Theeffectispresentamong wellgoverned firms,suggesting thatbettergoverned firmsmanageaccruals strategically. The concave relationship suggests that the marginal impact of earnings management on investmentefficiencydecreaseswiththeamountofearningsmanagement. Usingcasesof misreporting,wedocumentthatexcessiveearningsmanagementdoesnotimproveinvestmentefficiency. Takentogether, theseresultssupporttheviewthatamoderateamountof earnings management helps improve corporate investment decisions while an excessive amountundoes thebenefitofearnings management. Keywords: CorporateInvestmentDecisions;EarningsManagement. ∗DepartmentofFinance,LundquistCollegeofBusiness, UniversityofOregon,Eugene,OR97403;e-mail: bjulio@uoregon.edu;phone:+15413464449. †Federal Reserve Board of Governors, 20th Street and Constitution Avenue NW, Washington, DC 20551; e-mail: youngsuk.yook@frb.gov; phone: +1 202 475 6324. The views expressed in this article are those of the authorsand not necessarily of the Federal Reserve System. Part of the paper was written when Julio was at London Business School and Yook was at Sungkyunkwan University. We thank Luke Stein and seminar participants at Asian Financial Association International Conference and Financial Management Association AnnualMeeting.

1. Introduction This study empirically investigates how the use of intertemporal transfers of earnings affects afirm’s investmentpolicy. Managers’discretion overaccruals, defined as the difference between earnings and cash flows, allows for such transfer without violating the guidelines of Generally Accepted Accounting Principles (GAAP). We argue that earnings management, though often associated with poor corporate governance or fraudulent behavior, can be used by managers to signalgood earnings prospects to investors. In perfect capital markets, accruals management is irrelevant since all information is observable and verifiable. However, in a world with market frictions, accruals management can serve as a tool to help overcome information asymmetry between the firm and outsiders, improving access to external financing and internal asset allocation decisions. Managing accruals to obtain external financing, while sometimesviewedasopportunistic,canfacilitatebetterinvestmentdecisionstotheextentthat these funds are used to finance value-enhancing projects. Discretion overaccruals may allow internal funds to be allocated for valuable investment projects rather than for real earnings management: In the absence of managerial discretion over accruals, managers may resort to value-destructive real earnings management by delaying or foregoing investment, improving short-term profit at the expense of long-term firm value. According to Graham et al.’s (2005) survey of over 400 executives, managers candidly admit that they would take real economic actionssuchasdelayingmaintenanceoradvertisingexpenditure,andwouldevengiveuppositiveNPVprojects,tomeetshort-termearningsbenchmarks. Inthisstudy,weexplorewhether strategic earnings management can improve investment decisions. Specifically, we examine whether the ability to transfer earnings between periods allows managers to better align the firm’sinvestmentdecisionswithitsinvestmentopportunities. The 2001-2002 accounting scandals and the subsequent regulatory response have highlighted the opportunistic aspect of accruals management, which are typically in violation of GAAPguidelines. Alargebodyofliteraturehasexaminedthecausesandeffectsoffraudulent reporting1. In particular, some studies have stressed the association between aggressiveearn- 1Forexample,seeBenish,1999;BurnsandKedia,2006;Burns,Kedia,andLipson,2010;Efendi,Srivastava, andSwanson,2007;PlumleeandYohn,2010;Wang,Winton,andYu,2010;Wilson,2008. 1

ings management and financial policies including investment decisions. Kedia and Philippon (2009), for example, document that poorly performing firms overinvest and overstate their financial statements to mimic their better performing peers. McNichols and Stubben (2008) document that firms misreporting earnings overinvest during the misreporting period. However,theprioraccountingliteraturealsodemonstratesthatmanagerialdiscretionoveraccruals can enhance earnings’ informativeness. Managers can use accruals to signal private information about the firm. Discretionary accruals, a discretionary portion of total accruals, help managersproduceareliableandmoretimelymeasureoffirmperformancethanusingnondiscretionary accruals alone (Dechow, 1994; Dechow, Kothari, and Watts, 1998; Subramanyam, 1996). Thesignalisquitecredibledespitemanagerialdiscretionoveraccruals becauseaccruals management does not allow for permanent changes in earnings but only for a shift over time. We test ourprediction by examiningtheassociationbetween theabsolutevalueof discretionary accruals and investment efficiency. Discretionary accruals, estimated using a crosssectional version of the modified Jones model and expressed as percentage of lagged assets, have been used widely to proxy for accounting-based earnings management.2 We evaluate investment efficiency as the sensitivity of investment expenditures to investment opportunities as measured by Tobin’s Q. We augment the standard investment specification to allow for interactions between Q and the absolutevalue of discretionary accruals. For our analysis, we consider firms in the Compustat universebetween 1989 and 2012 excluding financial and utilityindustries. ControllingforTobin’sQandcashflows,wefindthataccrualsmanagement has a concave relationship with the sensitivity of investment to Tobin’s Q. That is, the additional usage of accruals improves investment decisions to a certain point, beyond which the investment-Q sensitivity deteriorates. Furthermore, we document that this pattern is mainly driven by high-Q firms. Despite having more investment opportunities, high-Q firms in our sample manifest a lower investment-Q sensitivitythan low-Q firms. The result highlightsthe importanceofstrategicaccruals managementbyshowingthatthebenefit ofaccrualsmanagementis greaterforfirms withmoreinvestmentneeds. 2For instance, see Bergstresser and Philippon, 2006; Healy and Wahlen, 1999; Teoh, Welch, and Wong, 1998a;Teoh,Welch,andWong,1998b;Yu,2008. 2

Theconcaverelationshipdocumentedabovesuggestseffectsoninvestmentdecisionsvary with the amount of earnings management. A modest amount of earnings management is associated with better investment responses to changing investment opportunities. However, marginal benefit diminishes with earnings management, suggesting that excessive earnings management hurts investmentefficiency. We further check a case of excessiveearnings management by examining firms misreporting financial statements only to restate in later dates. Accruals management tends to be modest in nature. First, accruals are managed within the boundary of GAAP. Second, accruals management requires that the sum of a firm’s income over all years equal the sum of its cash flows, meaning that managers must at some point in time reverse any excessive accruals made in the past. On the other hand, misreporting is often in violation of GAAP and sometimes results in SEC investigations or lawsuits, incurring large economic costs. Karpoff, Lee and Martin (2008) examine the firms targeted by SEC enforcement actions for financial misrepresentations and find that the size of lost sales and highercontracting and financing costs resulting from the earnings manipulationoutweighthe amount inflated by manipulation. Firms that restate their accounting statements in later dates facetighterloancontracttermsincludinghigherspreads,shortermaturities,higherlikelihood of being secured, and more covenant restrictions (Graham, Li, and Qiu, 2008). Given the relatively large expected costs, value-maximizing managers are not likely to rely on such aggressive earnings management. Consistent with this interpretation, we find misreporting that leadstorestatementinfuturedatesdoesnotimprovetheinvestment-Qsensitivity. Theresults reinforce the concave relation documented earlier: The cost of earnings management starts outweighingitsbenefit beyondacertain level. Oneconcern ininterpretingtheresultsisthatunobservedfactors maydriveaccruals management and the investment-Q association simultaneously, generating a spurious correlation. We address this concern by employing a difference-in-differences (DID) methodology. This approachiswellsuitedforattemptingtodisentanglecausalityinaquasi-experimentalsetting. We compare changes in investment efficiency for a sample of firms subject to an exogenous reductionin earnings managementtochanges in investmentefficiency forthosethat were not affected by the event. Specifically, we examine the effects of earnings management by using the passage of the Sarbanes-Oxley Act (SOX). Undoubtedly, SOX affected the way firms 3

managetheirearnings. SOXwasintendedtocurveearningsmanagementand,indeed,various studies document significant declines in the accruals management practice in the post-SOX periods. We select control groups in two different ways. First, we use the pre-SOX years as the control period and examine within firm variation in investment efficiency around the enactmentofSOX,wherefirmsactastheirowncontrols. Second,forourDIDestimation,we selectfirmsintheUnitedKingdomand CanadaasourcontrolgroupsinceSOX influencesall firms in the U.S. (our treatment group). The UK and Canada are considered to have similar accounting guidelines and practices. In addition, investment expenditures for firms in these countries follow similar time series patterns as those of the US firms. Our DID regressions showalargereductionininvestmentefficiencyforUSfirmsaroundthepassageofSOX.This provides support for the hypothesis that the decline in the use of earnings management after SOX reduced investmentsensitivityto investmentopportunitiesforUSfirms. Finally, we examine how well-governed firms view accruals management that can potentiallyimproveinvestmentresponsetoinvestmentopportunities. Arecentliteraturehasfocused on the opportunistic aspect of earnings management and has linked earnings management to poor corporate governance (e.g., Klein, 2002; Agrawal and Chadha, 2005; Cornett, Marcus, and Tehranian, 2008). However, the size of discretionary accruals alone does not address the strategic aspect of accruals management that can help secure internal or external funds necessary for valuable investment projects. A corporate governance mechanism should be designed to deter earnings management intended to manipulate earnings, but should not discourageaccrualsmanagementthatcanimproveresourceallocationtofinancevaluableinvestment projects. We test this hypothesis by examining the association between accruals management and investment sensitivity to investment opportunities separately for well-governed and poorly-governed firms. We utilize the governance index by Gompers, Ishii, and Metrick (2003)aswellasthepresenceofthreeindividualprovisions(poisonpill,classifiedboard,and goldenparachutesprovisions)tosortfirmsintotwosubgroups. Wefindthatgoodgovernance andbadgovernancegroupsexhibitapreviouslydocumentedconcaverelationbetweenaccrual management and investment-Qsensitivity. Better governedfirms show a strongerassociation for some of the governance measures. This result suggests that well-governed firms do not 4

discouragethestrategicusageofaccrualsandeffectivelymanageaccruals inresponsetotheir investmentopportunities. Overall, our findings highlight the importance of managerial discretion to transfer earnings between periods as a better alternative to real earnings management that sacrifices valuable investment projects. Prior literature suggests the effect of accruals on investment decisions can work through both the internal resource allocation channel and external financing channel. First, accruals allows managers to allocate internal funds for valuable investment projectsratherthanformeetingearningsbenchmarks,improvinginternalresourceallocation. Many studies document that firms have been engaging in real earnings management in various ways including price discounts, overproduction, delaying R&D investment, reduction of discretionary expenditures, stock repurchase, and sale of profitable assets.3 The real earnings management has direct real consequences. Ewert and Wagenhofer (2005) argue that firms engaging in real earnings management may deviate from normal business practices and thus experience a decline in their subsequent operating performances. Underperformance followingseasonedequityofferingsismoresevereforfirmsengaginginrealmanagementthanthose managing accruals (Cohen and Zarowin, 2010). Firms seem to sacrifice employment, R&D, and otherinvestmentto financeEPS-increasing stock repurchases (Almeidaet al.,2013). Second, accruals management can influence investment decisions through the channel of external financing. Firms seem to manage discretionary accruals to obtain financing as suggested by abnormally high levels of positive accruals in the periods preceding stock issuances (Chen, Gu, and Tang, 2008; DuCharme, 2004; Friedlan, 1994; Shivakumar, 2000; Teoh, Welch, and Wong, 1998a,b). Bergstresser, Desai, and Rauh (2006) also document increased earnings managements measured by pensions assumptions prior to acquisition activities. However, this evidence alone does not speak to the role of accruals in the efficiency of investment decisions. Linck, Netter, and Shu (2013) take a step toward this direction by examining financially constrained firms with valuable investment projects. They find that these 3Dechow and Sloan, 1991; Baber et al., 1991; Bushee, 1998; Roychowdhury, 2006; Hribar et al., 2006; Cheng, 2004; Almeida et al., 2013; Herrmann et al. 2003; Bartov, 1993; Jackson and Wilcox, 2000; Gunny, 2010 5

firmsusediscretionaryaccrualstocrediblysignalpositiveprospectstoraisecapitalnecessary fortheinvestments. We also contribute to the recent literature linking real investment decisions to earnings management. Zhang (2007), Wu, Zhang and Zhang (2010), Arif et. al (2016), among others, arguethataccrualsreflectrealinvestmentchoicesoffirms. Wu,Zhang,andZhang(2010)link the accrual anomaly, where firms with high accruals earn abnormally low returns on average, to real investment in a Q-theory framework. In their model, discount rates vary negatively with accruals and investment, therefore predicting lower future returns. Arif et. al (2016) show that like real investment, accruals decline significantly when economic uncertainty is high,consistentwiththeviewthataccounting accruals and investmentare stronglylinked. Our investigationis especially relevant in lightof therecent trend of adoptingstricter disclosure rules: The 2001-2002 accounting scandals and the subsequent passage of SOX likely increasedtheexpectedcostoffraudulentfinancialreporting. SOXinstitutedanumberofprovisionsincludingimprovingthecompositionandfunctionofauditcommittees,CEOandCFO financialstatementcertification,restrictionsonnonaudit-relatedworkbythecompany’sauditors, mandatory audit partner rotation, and an annual report on internal controls. Firms make choices between accruals managementand real activitiesmanagement(Cohen, Dey, and Lys, 2008;Cohenand Zarowin,2010;Badertscher, 2011),andthechoicedependsontheirrelative costs(Zang2012). Becauseaccrualsmanagementiseasiertodetectinnaturethanrealactivity manipulation, the heightened scrutiny post SOX is likely to have increased the relative cost ofaccrualsmanagement,reducingaccountingflexibilityinGAAP.Infact,empiricalevidence indicatesthataccrualsmanagementhasdecreasedsincetheimplementationofSOX.Loboand Zhou(2010)documentlowerdiscretionaryaccrualspostSOX.Koh,Matsumoto,andRajgopal (2008)documentthatthepropensityto engageinincome-increasingearnings managementto meet or beat earnings benchmarks has declined. Cohen et al. (2008) and Bartov and Cohen (2009) document that the level of accruals-based earnings management declined in the post- SOX period while the level of real activities manipulation increased, suggesting a shift from accrualsmanagementtorealmanagement. Ourexaminationoftheassociationbetweenaccrualsmanagementandinvestmentdecisionshasimplicationsforunderstandingthereal benefits and costsofcorporatedisclosurepolicies. 6

2. Data and Methodology 2.1. Accruals We utilizetheabsolutevalueofdiscretionary accruals as themeasure ofa moderateearningsmanagement. Weconsideraccrualsmanagementmoderateforthefollowingtworeasons. First,sincethesumofafirm’sincomeoverallyearsmustequalthesumofitscashflows,managersmustatsomepointintimereverseany”excessive”accrualsmadeinthepast. Therefore, it is unlikely to observe an extreme accruals management that persists over time. Second, an accrualsmanagementiswithintheboundaryofGAAPandthereforeisunlikelytobeextreme bydefinition. Ingeneral, anaccruals managementdoesnotincursevereeconomiccostsasdo earnings managements violatingtheGAAP, which are often followedby restatements and, in somecases, SEC investigationsorlawsuits. Total accruals are defined as the difference between earnings and cash flows from operationsandisconstructedbysubtractingCashFlowfromOperations(CompustatitemOANCF) from Net Income (item NI), scaled by beginning-of-year total assets. We decompose total accruals to separate the component that are beyond the control of the managers. We estimate a modified version of Jones model of accruals (Dechow, Sloan, and Sweeney (1995)), which regresses total accruals on changes in revenueand gross property, plant and equipment(PPE) to control for changes in nondiscretionary accruals caused by changing conditions. Total accruals includes changes in working capital accounts, such as accounts receivable, inventory, and accounts payable that depend on changes in revenues to some degree. Thus revenues are used to control for the economic environment of the firm because they capture the firms’ operations before managers’ manipulations. Gross PPE is included to control for the portion of total accruals related to nondiscretionary depreciation expense. To summarize, we estimate thefollowingmodelon oursamplebyeach industrygroup andyear4: 1 TA =b +b +b D REV +b PPE +e , it 0 1 2 it 3 it it A it−1 4WeutilizeFama-French’sdefinitionof48industries 7

whereTAistotalaccrualsscaledbythebeginning-of-yearassets,D REV isthechangeinsales normalized by beginningassets and PPE is gross property plantand equipmentscaled by beginningassets. We thenfeed theseestimatesto thefollowingequation to obtaindiscretionary accrual (DA). 1 DA =TA −b −b −b (D REV −D REC )−b PPE , it it 0 1 2 it it 3 it A it−1 where b is the estimated value of b (j= 0, 1, 2, 3). DA is essentially the discretionary porj j tion of total accruals expressed as a percentage of the lagged assets. Note that the change in accountsreceivable(D REC)issubtractedfromthechangeinrevenuestoallowforthemanipulation of credit sales. The original Jones (1991) Model implicitly assumes that discretion is not exercised over revenues while the modified Jones model (Dechow, Sloan, and Sweeney (1995)) adjusts the change in revenues for the changes in receivables to control for potential revenues manipulation. Our results are qualitativelyunchanged when we employ the original Jones model. Throughout the paper, we utilize absolute value of discretionary accruals since earningsmanipulationinvolvesbothpositiveandnegativevaluesofaccruals. 2.2. Data We consider all firms between 1989 and 2012 that are available in the merged Center for Research onSecurityPrices-CompustatIndustrialAnnualdatabase. Weexcludefinancialservices firms, regulated utilities, and firms with book values smaller than $10 million. We also drop observations with the missing total asset information. These steps result in a sample of 99,528 firm-year observations. The main variables are winsorized at the 1% and 99% level. Panel A of Table 1 summarizes various firm characteristics. Investment and cash flow are scaled by beginning-of-year capital measured by property, plant and equipment. The mean investment rate and mean lagged cash flow are 0.34 and 0.62, respectively. The mean discretionary accrual to total assets ratio (-0.005) is very close to zero as expected, reflecting the intertemporal nature of accruals management. However, its standard deviation is quite large with 0.349, highlightingmanagers’s discretionoverintertemporal shiftsin thefirm’s earning. 8

The absolute value of discretionary accruals is larger with a mean value of 12.2% of total assets. Next, corporate governance measures are drawn from Investor Responsibility Research Center (IRRC), which published detailed listings of corporate governance provisions. We examine the data between 1990 and 2007 because, after IRRC was acquired by Institutional Shareholder Services in 2005, a new data collection methodology was implemented in 2007, makingthepre-andpost-2007dataincomparable(seeKarpoff,Schonlau, andWehrly(2016) for additional detail about discontinuity between pre- and post-2007 data). The IRRC tracks 24 corporate provisionsincluding corporate charters and bylaws. Almostall provisionsgives management a tool to resist different types of shareholder activism, such as calling special meetings,changingthefirm’scharterorbylaws,suingthedirectors, orjustreplacing themall atonce. Gompers,Ishii,andMetrick(2003)dividesthemintofivegroups: tacticsfordelaying hostile bidders (Delay); voting rights (Voting); director/officer protection (protection); other takeover defenses (Other); and state laws(State). They also construct a governance index by assigning one point for the existence (or absence) of each provision and summing the points across the 24 provisions. Well-governed firms tend to have less provisions and, thus, are assigned a lower number of the governance index. For our sample periods, this index has a mean of 9.06 and standard deviation of 2.74. FollowingKediaand Philippon (2009), we also select one provision from each of the three groups defined by Gompers et al. excluding the VotingandStategroups.5 ClassifiedboardischosenfromtheDelaygroup,Goldenparachutes fromtheProtectiongroup,andPoisonpillfromtheOthergroup. Table1showsthat53.4%of our firm-year observations have the Poison Pill provision, 58.6% Classified Board provision, and 61.3%GoldenParachutes provision. PanelBreportsinvestmentratesforsubsamplessortedonlaggedQand|DA|. Thesample is first sorted into four quartiles based on lagged Q, and then each of the four subsamples is furthersortedintofourquartilesbasedon|DA|. Investmentratesincreasewithinvestmentopportunities proxied by lagged Q, consistent with the literature. Investment rates also increase monotonically with | DA |, but the magnitude differs across lagged Q quartiles. Investment 5WedroppedLimitAbilitytoAmendCharterprovisionfromtheVotinggroupbecauseverylittlefractionof oursampleobservationshavetheprovision. 9

rises slowly for low Q quartiles but moves up rapidly for high Q quartiles. For the lowest Q quartile,forexample,investmentratesriseonlyby0.054,from0.182inthelowest|DA|quartile to 0.236 in the highest |DA| quartile. By contrast, for the highest Q quartile, investment rates leap from 0.480 to 0.648, suggesting that accruals are utilized heavily in conjunction withinvestmentsfor firmswithstronggrowthpotentials. 3. Test Results 3.1. Baseline Specification In this section, we investigate our main hypothesis that the accruals management can be utilized to improve the investment-Q relationship. We augment the standard investment regressionspecificationas follows: I =a +b ·|DA |+b ·|DA |·Q +b ·|DA |2·Q +b ·Q +b ·CF +g +e , it i 1 it 2 it it−1 3 it it−1 4 it−1 5 it−1 t it where i indexes a firm and t indexes time. The dependent variable is investment scaled by beginning-of-year capital. | DA | is the absolute value of discretionary accruals. | DA | ·Q it and |DA|2 ·Q are of particularinterest becausethey capturedifferences in investment-Qsensitivity across firms with a varying degree of accruals management. The quadratic term is introduced to account for the possibility that the effect of accruals management may not be linear. Timeandfirmfixedeffectsareincluded. Wealsoreplacefirm fixedeffectswithindustry fixed effects in some specifications. Our industry definition is drawn from Fama/French’s classificationof48 industries. Table 2 reports the estimation results. The first column presents the standard investment regression result as a benchmark. The second regression allows for the possibility of a linear relationship between | DA | and the investment-Q sensitivity. The coefficient of | DA | ·Q is positiveand significant, indicating that investment is more sensitiveto investment opportunities when accruals are actively managed. The third regression introduces a quadratic term, | DA |2 ·Q to allow for the possibility that the marginal effect of | DA |2 ·Q may vary with 10

the size of | DA |. Once the quadratic term is introduced, the coefficient of | DA | ·Q nearly quadruplesfrom0.0153to0.0590andthestatisticalsignificancealsoimproves. Thequadratic term is negative and statistically significant at the 1% level. The quadratic specification fits the databetter than a linear specification, lending support forthe view that moderate accruals management can improve the investment-Q sensitivity but an extreme usage of accruals can ratherhurttheinvestment-Qsensitivity. Thelastcolumnaddscashflow,buttheresultsremain the same. Also note that the coefficients of Q vary little across the four regressions, suggesting that | DA | adds additional explanatory power to the specification. Overall, the results support our hypothesis that accruals management helps managers respond to the investment opportunitiesmoreefficiently. Wenextinvestigatewhethertheassociationbetweenaccrualsmanagementandtheinvestments- Qsensitivitychangeswithinvestmentopportunities. PanelBofTable1showsthatinvestment increases with |DA| but the size of the increase differs considerably across different Q quartiles. WefurtherexaminethisdynamicsbysortingthesampleintotwosubgroupsbasedonQ and estimating the baseline specification separately for the two subsamples. Table 3 reports the estimation results. The first two columns report the benchmark cases without |DA|. The investment-Qsensitivityseems much higher for the low Q firms. The coefficient for the high Q subgroup is only 0.0639 while the coefficient for the low Q subgroup is 0.1488, suggesting that high Q firms may have more room for improvement in their investment response to investment opportunities. The last two regressions present the results of our baseline specification. The effect of discretionary accruals is pronounced in the high-Q subgroups as shown by thelinearand quadratictermsof|DA|. Theseestimatesaresimilarto thoseinfull sample results (Table 2). As before, marginal increases in | DA | improves the investment-Q sensitivity as long as the size of accruals are moderate. The estimates for the low Q subsample are quite different. The quadratic term remains negative and significant, but the linear term, |DA|·Q, is no longer statistically significant. Overall, the documented association seems to be mainly driven by high Q firms. This highlights the importance of strategic accruals management because accruals have bigger effects where they are needed the most. That is, the effects are more pronounced in the subsample with relatively lower investment-Q sensitivity 11

in the benchmark cases (first two regressions). Furthermore, these firms are the ones with stronggrowthpotentials,forwhichinvestmentdecisionsareespecially critical. 3.2. Restatements The concave relationship documented in the previous section suggests that the marginal improvementintheinvestment-Qsensitivitydiminisheswiththesizeofdiscretionaryaccruals. To corroborate this result, we consider a more extreme form of earnings management, financial misreporting that requires restatements in later dates. The degree of misreporting varies considerably among restating firms from a minor misapplication of accounting principles to an outright fraud. While accruals management is a legitimate tool that allows managers to exertdiscretionoverreportedearningsacrosstime,misreportingisaclearviolationofGAAP, resultinginSECinvestigationsorlawsuitsinsomecases. Becausemisreportingismorelikely to be driven by opportunisticearnings management, we do not expect such earnings managementto beassociatedwithimprovementin theinvestment-Qsensibility. We start with the restatement announcement data provided by the United States General Accoutring Office (GAO). The data contain announcements made between January 1997 and June 2006. We then identify the misreporting periods corresponding to the restatement announcement by reading news articles in FACTIVA. Our final sample covers 2284 restating firm-year observations between 1996 and 2004. The distribution of misreporting over the sample period is reported in panel A of Table 4. On average, 6.6% of sample firms misreport each year to restatetheir accounting statementin laterdates. However, there is a strong timeseries trend in the frequency of misreporting. An incidence of misreporting is relatively rare inearlyyearswith46incidencesin1996and98in1997. However,itgraduallyincreasesover timetoreach 421incidences in2004. Wemodifythebaselineinvestmentspecificationbyreplacing|DA|witharestatedummy variableas follows: I =a +b ·Restate +b ·Restate ·Q +b ·Q +b ·CF +g +e , it i 1 it 2 it it−1 3 it−1 4 it−1 t it 12

where Restate is set to one if misreporting that subsequently results in restatements occurs in the given firm-year. Note that a quadratic association cannot be tested in this setting because Restate is an indicator variable. Restate·Q captures differences in the investment-Q sensitivity between misreporting firms and non-restating firms. Panel B of Table 4 report the estimation results. The first column presents a univariate analysis of investment for restating firms and non-restating firms with year and firm fixed effects. The restate dummy is positive, but onlymarginallysignificant (10%). Thenext two regressionsshow thattherestate dummy becomes negative and significant once Restate·Q, Q, and cash flow are controlled for. The mainvariableofinterest,Restate·Q,remainsinsignificant,suggestingthatmisreportingdoes not facilitate a better alignment between investment and investment opportunities. It appears that accruals are utilized strategically to improve investment decisions, but that fraudulent accountingseemsto bemotivatedby ratheropportunisticbehavior. 3.3. Quasi-Natural Experiment: The Sarbanes-Oxley Act and Earnings Management Animportantconcernintheaboveresultsshowingastrong,concaverelationshipbetween earnings management and investment efficiency is the possibleendogeneity coming from the two choice variables. There could be an omitted variable that drives both investment and earnings management. An ideal empiricalsetup wouldprovideexogenousshocks toearnings management for one group of firms and not for another. A comparison of changes in investment around the shock for the two groups of firms would yield a better estimate of the effect of earnings management on investment. In this section, we employ an empirical approach to address the concerns about possible endogeneity by employing a natural experiment in the form oftheSarbanes-Oxley Act of2002. We examine the effects of earnings management on investment by using the passage of SOX. SOX instituted a number of provisions including improving the composition and function of audit committees, CEO and CFO financial statement certification, restrictions on nonaudit-related work by the companys auditors, mandatory audit partner rotation, and an annual report on internal controls. Empirical evidence shows that accruals management de- 13

creased quickly and significantly after SOX. Lobo and Zhou (2010) document lower discretionary accruals post SOX. Koh, Matsumoto,and Rajgopal (2008) documentthat thepropensity to engage in income-increasing earnings management to meet or beat earnings benchmarks has declined. Cohen et al. (2008) and Bartov and Cohen (2009) document that the levelofaccruals-based earningsmanagementdeclinedinthepost-SOXperiodwhilethelevel ofreal activitiesmanipulationincreased, suggestingashiftfrom accruals managementto real management. In our sample, the average amountof discretionary accruals (in absolutevalue) was 19.1% of total assets. After the enactment of SOX, the average value of discretionary accruals fell to 13.2% of assets, representing a 31% decline in the use of discretionary accruals. We use this shock to the use of discretionary accruals to examine the impact of earnings managementon investmentefficiency. An important challenge is that the SOX Act was at the national level and hence affected most firms in the US, complicating the formation of a good control group of firms. To deal with this complication, we estimate changes in investment efficiency around SOX in two ways. First, we estimate investment regressions with firm fixed effects and include a post- SOX dummy variable. In this estimation, the firms in the sample are also the control group, wherethepre-SOX timeyears represents thecontrolperiod andthepost-SOXyears thetreatmentperiod. Specifically, weestimatetheregression I it =a +g +b ·1 +b ·Q +b Q ·1 +b ·CF +e , i t 1 (SOX) 2 i,t−1 3 i,t−1 (SOX) 4 i,t−1 it K i,t−1 where a and g are firm and year fixed effects, 1 is a dummy variable taking a value i t (SOX) of one in the years following the implementation of SOX. The coefficient b captures level 1 changes in investment rates around SOX and b , the main coefficient of interest, captures 3 changes ininvestmentsensitivitytoTobin’sQ inthepost-SOXperiod. The first column of Table 5 reports the estimation results for the post-SOX analysis of investmentefficiency. Thecoefficientontheinteractionbetween Tobin’sQandthepost-SOX dummy variable is negative and statistically significant, representing a decline in investment efficiency followingSOX. The valueof the coefficient, -0.0227, represents a decline of about 14

27%ininvestmentefficiencyfollowingthenegativeshocktotheuseofdiscretionaryaccruals aftertheSarbanes-Oxley Act. An alternative approach to comparing investment investment efficiency before and after SOXwithintheUSistocomparechangesininvestmentforfirmsaffectedbySOXtofirmsthat werenotaffectedaroundthesametimeperiod. Tothis,weemployadifference-in-differences (DID) estimator. The DID methodology we employ compares the effect of SOX on groups affected by the regulation (treatment group) to those that are unaffected (control group). The inferencesaremadebycalculatingthechangesininvestmentlevelsandefficiencyoftreatment firms around the event to the changes around the event for the control firms. We choose to construct a set of control firms using data from Canada and the United Kingdom, as firms in these countries tend to be affected by similar economic shocks as firms in the US but were not subject to the changes brought on by SOX. Assuming that the control firms’ investment policiesarebeingdrivenbysimilardynamicsovertime,itwillallowustocontrolforcommon economic shocks and also to alleviatepotential bias due to other changes in law around SOX thatcould haveaffected thetreatmentgroup. ToinvestigatetheeffectofSOXoninvestmentefficiencyinaDIDframework,weestimate thefollowingregression: I it =a +g +d ·1 +n ·1 ·1 +h ·1 ·1 ·Q +b ·Q +b ·CF +e , i t (SOX) (SOX) (i=T) (SOX) (i=T) i,t−1 1 i,t−1 2 i,t−1 it K i,t−1 wherea andg arefirmandyearfixedeffects,1 isadummyvariabletakingavalueofone i t (SOX) intheyears followingtheimplementationofSOX and zero otherwise,and 1 is adummy (i=T) variableset equal to onefor firms that belong to thetreatment group and zero for firms in the control group. The coefficient h on the interaction between the two indicator variables and Tobin’sQcaptures thedifference-in-differences effecton investmentandisthemainestimate of interest in theregression. The coefficient n picks up the difference-in-differences effect on investmentlevels. 15

A challenge with employing the DID methodology around the passage of SOX is that there are other factors, both observable and unobservable, that may influence investment in the United States and other countries around the enactment of SOX. The DID regression is helpfulinthatitallowsforthecontrolofomittedvariablesthataffectthetreatmentandcontrol group similarly. However, identification of the causal effect of SOX on investment requires controlling for other shocks to the treatment group that may be correlated with the timing of SOX. Forexample,thedeclinein investmentefficiency aroundthepassageofSOX mayhave been more significant for US firms due to different sensitivities to the global business cycle. Weaddressthisandrelatedconcernsinavarietyofways. First,weincludefirmlevelcontrols, particularly Tobin’s Q and cash flow, to control for changing investment opportunities over time. Second, in robustness checks6, we include industry by year fixed effects to control for industry/timevariationand find similarresults. Before reporting the results, we examine whether the use of Canadian and UK firms are appropriate to use as controls. An important assumption in the way we construct treatment and control groups is that the outcome in both groups would follow the same time trend in the absence of the treatment. While this assumption is very difficult to verify, we can look at pre-treatment trends to see if investment followed a similar pattern prior to the enactment of SOX. Figure 1 shows mean investmentrates for the two treatment and control groups around the passage of SOX. The figure shows that investment rates for both treatment and control firms moved roughly in parallel before the policy change. After the enactment of SOX, the treatmentfirmsshowaslowerrateofincreaseininvestmentratescomparedtofirmsinCanada and the UK. Figure 1 supports the assumption that trends in investment rates were similar prior to the passage of SOX. We also examine changes in the full distribution of investment for both treatment and control firms. In Figure 2 we plot the kernel densities of investment rates for bothtreatment and control firms before and after thepolicychange. The distribution ofinvestmentratesshiftstotheleftforthetreatmentfirmsbutnotsignificantlyforthecontrol firms, suggesting the presence of an effect of SOX on corporate investment. The shift in the density for the treatment group is statisticallysignificant as theKolmogorov-Smirnovtest for 6Resultsavailableuponrequest. 16

theequalityofthedistributionsisrejectedatthe1%level. Thefiguresshowthatthereappears tobea changeininvestmentfortreatmentgroupscompared tocontrolgroups. Columns (2) through (4) of Table 5 report the results of the DID regression. The second column compares changes in investment efficiency for US firms compared to both UK and Canadian firms. The coefficient on the interaction term between the treatment effect and Tobin’s Q is negative and statistically significant with a magnitude of -0.0339, suggesting a reduction in investment efficiency for US firms around the passage of SOX relative to firms in Canada and the UK. Columns 3 and 4 repeat the difference-in-difference methodology separately for Canadian and UK firms as control groups and find similar results. We also note that the interaction between treatment/control and the SOX dummy is significant and negative. Thenegativecoefficientof-0.0327suggeststhatinvestmentrateswerealsoaffected negativelybythepassageofSOX.Themagnitudesofchangesininvestmentefficiencyrelative to Canadian and UK firms are similar to the magnitude measured in the US only sample reported in column 1. While the identifying assumptions differ across our approaches, the resultsareconsistentandlendsupporttothehypothesisthatthedeclineintheuseofearnings managementafterSOX reduced investmentefficiency forUS firms. 3.4. Corporate governance and accruals management The recent literaturehas focused on theopportunisticaspect ofearnings managementand haslinkedearningsmanagementtopoorcorporategovernance. Klein(2002)andAgrawaland Chadha (2005) document that independence of audit committee and corporate board is negatively associated with earnings management and restatement. Kedia and Philippon (2009) examine restating firms and report that firms with poor governance are more likely to misreport accounting statements to restate in later dates. Cornett, Marcus, and Tehranian (2008) document that the usage of discretionary accruals is reduced by better governance measured by institutional ownership of shares, institutional investor representation on the board of directors, andthepresenceofindependentoutsidedirectors ontheboard. Cheng (2008)reports thatalargerboardisassociatedwithasmallervariationinaccruals. However,thesestudiesfocus on thesize ofdiscretionary accruals and do notconsidertheirinteraction withinvestment 17

decisions. The size of discretionary accruals alone does not indicate whether they were used for fraudulent accounting or to help align investment opportunities with internal or external resources. While a corporate governance mechanism should be designed to deter accounting fraud, it should not discourage strategic management of accruals to the extent that it is within the GAAP boundary and improves investment efficiency. If the strategic usage of discretionary accruals can improve investment decisions, we expect to observe the documented concaverelationinwellgovernedfirmsaswell. Wetestthishypothesisbyexaminingtheassociation between accruals management and investment sensitivity to investment opportunities separately forwell-governedand poorly-governedfirms. We sort the firms into two subgroups based on the degree of corporate governance. Four measures of governance are employed including the governance index by Gompers, Ishii, and Metrick (2003) and the presence of three individual provisions (poison pill, classified board, and golden parachutes provisions). The absence of each of the individual provisions is considered good governance, as the presence of those provisionsweakens thepower of the shareholders in favor of managers. Similarly, a firm is classified as having good governance if the value of the governance index is lower than the median. Table 6 summarizes the usage of accruals and other firm characteristics for the two subgroups. The first column shows that the differences in total accruals between the two subgroups are very small and statistically insignificant across all four measures. The second column reports that the absolute values of discretionary accruals are somewhat different across the two subgroups. To the extent that these measures capture the quality of governance, better governed firms have higher absolute values of discretionary accruals in three of the four measures. If the usage of discretionary accruals were motivated exclusively by opportunistic reasons, we would expect the absolute values of discretionary accruals to be higher for poorly governed firms. However, it appears that better governed firms do not discourage the usage of discretionary accruals. These firms also seem to differ in other dimensions. Well governed firms invest more, have more investment opportunities proxied by Q, and have more cash flows. The result is consistent with the previous studies documenting a negative correlation between governance measures and Tobin’sQ (Gompers,Ishii,and Metrick,2003;Bebchuk, Cohen, and Ferrell, 2009). 18

We next conduct a multivariate analysis by estimating the baseline specification for the two subgroups separately. The first two columns of Table 7 reports the regressions results for subgroups formed based on the governance index. Both subgroups show a concave relation between accruals managements and investment-Q sensitivity. The next columns sort the firms based on the presence of each of the individual provisions. The results are similar regardless of which of the four governance measures is used. A small exception is that when a golden parachute provision is utilized, the quadratic term is not statistically significant for well governed firms. Overall, the concave relationship is present for both well-governed and poorly-governed firms. The evidence is consistent with our view that accruals management can beutilizedtoenhance thecorporateinvestmentresponsetoinvestmentopportunities. 4. Conclusion Weempiricallyinvestigatetherelationshipbetweenintertemporaltransferofearningsand the efficiency of corporate investment decisions. Using the absolute value of discretionary accruals asameasureofsuchearningsmanagement,wedocumentthatearningsmanagement exhibits a concave relationship with the investment sensitivity to investment opportunities as measured by Tobin’s Q. We find that the relationship between earnings management and investmentefficieny is concentrated among firms with relativelyhigh investmentopportunities. The effect is present among firms with good corporate governance measures, suggesting that bettergovernedfirmsmanageaccrualsstrategically. Theconcaverelationshipsuggeststhatthe marginalimpactofearningsmanagementoninvestmentefficiencydecreases withtheamount of earnings management. Using misreporting that leads to restatement in future dates, we documentthatamoresevereformofearningsmanagementdoesnotimproveinvestmentefficiency. Weimplementadifference-in-differences(DID)methodologytodisentanglecausality inaquasi-experimentalsettingaroundthepassageoftheSarbanes-OxleyAct. Wefindalarge reduction in investment efficiency for US firms (treatment group) relative to those in UK and Canada (control group) around the passage of SOX. Taken together, these results support the view that a moderate amount of earnings management helps improve corporate investment decisionswhilean excessiveamountundoesthepotentialbenefit ofearnings management. 19

References Agrawal, A., Chadha, S., 2005. Corporate Governance and Accounting Scandals, Journal ofLaw andEconomics48, 371–406. Almeida, H., Fos, V., and Kronlund, M., 2016, The Real Effects of Share Repurchases, Journalof FinancialEconomics119,168-185. Arif, S., N. Marshall and T.L. Yohn, 2016, Understanding the relation between accruals and volatility: A real options-based investment approach, Journal of Accounting and Economics 62,65–86. Baber,W.,Fairfield,P.,andHaggard,J.,1991. Theeffectofconcernaboutreportedincomeon discretionary spending decisions: The case of research and development. Accounting Review 66,818-829. Badertscher, B., 2011, Overvaluation and the Choice of Alternative Earnings Management Mechanisms,AccountingReview86,1491–1518. Baker, M., Stein, J., and Wurgler, J., 2003, When Does the Market Matter? Stock Prices and the Investment of Equity-Dependent Firms, Quarterly Journal of Economics 118, 969– 1005. Bartov, E., 1993, The timing of assets sales and earnings manipulation. Accounting Review 68,840–855. Bartov, E., and Cohen, D., 2009, The numbers game in the pre- and post-Sarbanes-Oxley eras, JournalofAccounting,Auditing&Finance24,505–534. Bebchuk, L., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Review of FinancialStudies22,783-827. Benish, B., 1999, Incentives and Penalties Related to Earnings Overstatements that Violate GAAP. AccountingReview74, 425–457. Bergstresser, D., Philippon, T., 2006, CEO Incentives and Earnings Management, Journal ofFinancialEconomics80, 511–529. Bergstresser, D., Desai, M., and Rauh, J., 2006, Earnings Manipulation, Pension AssumptionsandManagerial InvestmentDecisions,QuarterlyJournalof Economics121,157–195. Burns, N., Kedia, S., 2006, The Impact of Performance-Based Compensation on Misreport- 20

ing,Journalof FinancialEconomics79,35–67. Burns, N., Kedia, S., and Lipson, M., 2010, Institutional Ownership and Monitoring: Evidencefrom Financial Misreporting,JournalofCorporateFinance16,443–455. Bushee, B. 1998, The influence of institutional investors on myopic R&D investment behavior,AccountingReview73,305-333. Chen,J.,Gu,Z.,andTang,Y.,2008,CausesorConsequences? EarningsManagementaround Seasoned EquityOfferings,WorkingPaper. Cheng, S., 2004, R&D Expenditures and CEO Compensation, The Accounting Review 79, 305–328. Cheng, S., 2008, Board Size and the Variability of Corporate Performance, Journal of FinancialEconomics87,157–176. Cohen, D., Dey, A., and Lys, T., 2008, Real and accrual-based earnings management in the pre-and post-Sarbanes-Oxleyperiods,AccountingReview83, 757–787. Cohen, D., Zarowin, P., 2010, Accrual-based and real earnings management activitiesaround seasoned equityofferings,JournalofAccountingandEconomics50, 2–19. Cornett,M.,Marcus,A.,andTehranian,H.,2008,CorporateGovernanceandPay-for-Performance: TheImpactofEarningsManagement,JournalofFinancialEconomics87,357–373. Dechow, P., 1994, Accounting earnings and cash flows as measures of firm performance: Theroleofaccountingaccruals, Journalof Accountingand Economics18,3-42. Dechow, P., Sloan, R., 1991, Executive incentives and the horizon problem: An empirical investigation,JournalofAccountingandEconomics14, 51-89. Dechow, P., Kothari, S., Watts, R., 1998, The Relation Between Earnings and Cash Flows, Journalof Accountingand Economics25,133-168. Dechow, P., Sloan, R., and Sweeney, A., 1995, Detecting Earnings Management, AccountingReview70No.2, 193–225. Dechow,P., Sloan,R., andSweeney,A.,1996,Causesand ConsequencesofEarningsManipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC, Comptemporary AccountingResearch13 No.1,1–36. Desai, H., Hogan, C., and Wilkins, M., 2006, The Reputational Penalty for Aggressive Ac- 21

counting: EarningsRestatementsandManagementTurnover,AccountingReview81,83–112. DuCharme, L., Malatesta, P., and Sefcik, S., 2004, Earnings Management, Stock Issues, and ShareholderLawsuits,JournalofFinancialEconomics71,27–49. Efendi, J., Srivastava, A., and Swanson, E., 2007, Why Do Corporate Managers Misstate Financial Statements? The Role of Option Compensation and Other Factors, Journal of FinancialEconomics85,667–708. Gompers, P., Ishii, J., and Metrick, A., 2003, Corporate Governance and Equity Prices, The QuarterlyJournalof Economics118,107-155. Graham, J., Harvey, C., and Rajgopal, S., 2005, The Economic implications of corporate financial reporting,Journalof AccountingandEconomics40,3–73. Graham, J., Li, S., and Qiu, J., 2008, Corporate Misreporting and Bank Loan Contracting, Journalof FinancialEconomics89,44–61. Gunny, K., 2010, The relation between earnings management using real activities manipulation and future performance: Evidence from meeting earnings benchmarks, Contemporary AccountingResearch27, 855–888. Healy, P., Wahlen, J., 1999, A Review of the Earnings Management Literature and Its ImplicationsforStandard Setting, AccountingHorizons13,365–383. Herrmann, D., Inoue, T., and Thomas, W., 2003. The sale of assets to manage earnings in Japan. JournalofAccountingResearch 41,89–108. Hribar, P. and Collins, D., 2002, Errors in estimating accruals: implications for empirical research, JournalofAccountingResearch 40,105–134. Hribar,N.,Johnson,W.,2006. Stockrepurchasesasanearningsmanagementdevice. Journal ofAccountingandEconomics 41,3-27. Jackson, S., Wilcox, W., 2000, Do managers grant sales price reductions to avoid losses and declinesin earningsand sales? QuarterlyJournalofBusinessandEconomics39, 3-20. Jones, J., 1991, Earnings Management During Import Relief Investigations, Journal of AccountingResearch29 No.2,193-228. Karpoff, J., Lee, D., and Martin, G., 2008, The Cost to Firms of Cooking the Books, Journal ofFinancialand QuantitativeAnalysis43, 581–612. 22

Karpoff,J.,R.Schonlau,andE.Wehrly. 2016. Dotakeoverdefenseindicesmeasuretakeover deterrence?, WorkingPaper. Kedia, S., Philippon, T., 2009, The Economics of Fraudulent Accounting, The Review of FinancialStudies22,2169–2199. Klein, A., 2002. Audit committee, board of director characteristics, and earnings management,Journalof AccountingandEconomics33,375–400. Koh, K., Matsumoto, D., and Rajgopal, S., 2008, Meeting or beating analyst expectations inthepost-scandalsworld: changesinstockmarketrewardsandmanagerialactions,ContemporaryAccountingResearch 25,1067–98. Linck, J., Netter, J., and Shu, T., 2013, Can Managers Use Discretionary Accruals to Ease Financial Constraints? Evidence from Discretionary Accruals Prior to Investment, AccountingReview88,2117–2143. Lobo, G., Zhou, J., 2010, Changes in discretionary financial reporting behavior following theSarbanes-Oxleyact, JournalofAccounting,AuditingandFinance25, 1-26. McNichols, M., Stubben, S., 2008, Does Earnings Management Affect Firms’ Investment Decisions?,AccountingReview83, 1571-1603. Palmrose, Z., Richardson, V., and Scholz, S., 2004, Determinants of Market Reactions to RestatementAnnouncements,JournalofAccountingand Economics37,59–89. Plumlee, M., Yohn, T., 2010. An Analysis of the Underlying Causes Attributed to Restatements,AccountingHorizons24,41–64 Roychowdhury, S. 2006. Earnings management through real activities manipulation. Journalof Accountingand Economics42(3): 335-370. Subramanyam, K., 1996, The pricing of discretionary accruals, Journal of Accounting and Economics,199622, 249–281. Teoh, S., Wong, T., 2002, Why New Issues and High-Accrual Firms Underperform: The RoleofAnalysts’Credulity,The Reviewof FinancialStudies15 No.3,869–900. Teoh, S., Welch, I., and Wong, T., 1998a, Earnings Management and the Long-Run Market Performance ofInitialPublicOfferings,TheJournalofFinance53,1935–1974. Teoh, S., Welch, I., and Wong, T., 1988b, Earnings Management and the Underperformance ofSeasoned EquityOfferings,Journalof FinancialEconomics50, 63–99. 23

Wang, T., Winton, A., and Yu, X., 2010, Corporate Fraud and Business Conditions: Evidencefrom IPOs, Journalof Finance65,2255–2292. Wilson, W., 2008, An Empirical Analysis of the Decline in the Information Content of EarningsFollowingRestatements,AccountingReview83. 519–548. Wu, J., Zhang, L., and X. Zhang, 2010, The q-theory approach to understanding the accrual anomaly,JournalofAccountingResearch,48, 177–224. Yu, F., 2008, Analyst Coverage and Earnings Management, Journal of Financial Economics 88,245–271. Zang, A., 2012, Evidence on the trade-off between real activities manipulation and accrualbased earningsmanagement,AccountingReview87, 675–703. Zhang,X,2007,Accruals,investment,andtheaccrualanomaly,AccountingReview82,1333– 1363. 24

Table1 Summary Statistics PanelAsummarizesfirmcharacteristicsforoursamplebetween1989and2012. Investmentandcashfloware scaled by beginning-of-year capital measured by property, plant and equipment. Discretionary accrual (DA) is a discretionary portion of total accruals, which is defined as net income minus cash flow from operations. DA is estimated by a cross-sectional version of the modified Jones model, expressed as percentage of lagged assets. Corporate governancedata coverthe period between1990 and 2007. PanelB reportsinvestmentrates forsubsamplessortedonlaggedQand|DA|. ThesampleisfirstsortedintofourquartilesbasedonlaggedQ, andtheneachofthefoursubsamplesisfurthersortedintofourquartilesbasedon|DA|. Seetheappendixfor variabledescriptions. PanelA:FirmCharacteristics Variable N Mean Q1 Median Q3 Std. Dev. FirmCharacteristics Investment 99,647 0.34 0.11 0.21 0.39 0.42 LaggedQ 99,647 1.92 1.05 1.40 2.14 1.56 LaggedCashFlow 99,647 0.62 0.08 0.27 0.65 1.61 Leverage 99,647 0.25 0.04 0.20 0.37 0.46 DiscretionaryAccruals DA 99,647 -0.005 -0.078 -0.005 0.066 0.349 |DA| 99,647 0.122 0.022 0.055 0.125 0.204 CorporateGovernance GovernanceIndex 26,399 9.06 7.00 9.00 11.00 2.74 Number(%)offirm-yearobservationswiththeprovisions PoisonPill 14,084 53.4% ClassifiedBoard 15,473 58.6% GoldenParachutes 16,185 61.3% PanelB:MeanInvestmentRatesbyLaggedQ−|DA|Quartiles LaggedQ |DA| 1stquartile 2ndquartile 3rdquartile 4thquartile 1stquartile 0.182 0.238 0.304 0.480 2ndquartile 0.189 0.246 0.315 0.492 3rdquartile 0.201 0.259 0.341 0.535 4thquartile 0.236 0.309 0.407 0.648 25

Table2 Accruals Management and Investment Thistablepresentsestimationresultsofthebaselinespecification: I =a +b ·|DA |+b ·|DA |·Q +b ·|DA |2·Q +b ·Q +b ·CF +g +e , it i 1 it 2 it it−1 3 it it−1 4 it−1 5 it−1 t it wherethedependentvariableisinvestmentscaledbybeginning-of-yearcapital. Firmandyearfixedeffectsare included.Standarderrorsareclusteredatthefirmlevelandreportedinparenthesis. (1) (2) (3) (4) Q 0.0751*** 0.0811*** 0.0775*** 0.0664*** it−1 (0.002) (0.003) (0.003) (0.003) CF 0.0775*** 0.0778*** it−1 (0.002) (0.002) |DA | 0.0877*** 0.0837*** 0.0731*** it (0.015) (0.014) (0.014) Q ·|DA | 0.0153** 0.0590*** 0.0635*** it−1 it (0.007) (0.014) (0.013) Q ·|DA |2 -0.0427*** -0.0445*** it−1 it (0.012) (0.012) Constant 0.2337*** 0.2230*** 0.2254*** 0.2055*** (0.007) (0.008) (0.008) (0.008) Observations 99,647 99,647 99,647 99,647 R2 0.475 0.439 0.439 0.478 ***indicates1%significanceand**indicates5%significance. 26

Table3 Subsample Analysis: HighQ vs. Low Q ThistableestimatesthefollowingbaselinespecificationfortwosubsamplessortedonQ. I =a +b ·|DA |+b ·|DA |·Q +b ·|DA |2·Q +b ·Q +b ·CF +g +e , it i 1 it 2 it it−1 3 it it−1 4 it−1 5 it−1 t it wherethedependentvariableisinvestmentscaledbybeginning-of-yearcapital. Firmandyearfixedeffectsare included.Standarderrorsareclusteredatthefirmlevelandreportedinparenthesis. (1) (2) (3) (4) HighQ LowQ HighQ LowQ Q 0.0639*** 0.1488*** 0.0563*** 0.1469*** it−1 (0.003) (0.010) (0.003) (0.011) CF 0.0787*** 0.0754*** 0.0790*** 0.0756*** it−1 (0.003) (0.004) (0.003) (0.005) |DA | 0.1051*** 0.069 it (0.028) (0.050) Q ·|DA | 0.0513*** 0.061 it−1 it (0.016) (0.052) Q ·|DA |2 -0.0399*** -0.0719*** it−1 it (0.013) (0.024) Constant 0.3329*** 0.1023*** 0.2955*** 0.0764*** (0.016) (0.012) (0.017) (0.013) Observations 51,410 51,409 50,412 49,235 R2 0.528 0.478 0.529 0.473 ***indicates1%significanceand**indicates5%significance. 27

Table4 Accounting Restatements PanelAdescribesthedistributionofmisreportingofaccountingstatementsbetween1996and2004.Weidentity the misreported periods for each firm that makes a restatement announcement in the period of January 1997 throughJune2006.PanelBreportsestimationresultsofbaselinespecification: I =a +b ·Restate +b ·Restate ·Q +b ·Q +b ·CF +g +e , it i 1 it 2 it it−1 3 it−1 4 it−1 t it whereRestateissettooneifafirmmisreportsinthegivenfirm-year,andzerootherwise. Notethat|DA | is it replacedbytherestatedummyvariable. Firmandyearfixedeffectsareincluded. Standarderrorsareclustered atthefirmlevelandreportedinparenthesis. PanelA:Distribution ofmisreporting byrestatementdata FiscalYear Misreporting firms Numberofobservations Fraction(%) 1996 46 1.1% 1997 98 2.2% 1998 133 3.0% 1999 200 4.6% 2000 256 6.2% 2001 320 8.2% 2002 396 10.8% 2003 414 11.6% 2004 421 11.7% Mean 254 6.6% PanelB:Investment Regressions (1) (2) (3) Restate ·Q 0.0134 0.0164 it it−1 (0.011) (0.011) Restate 0.0216* -0.0542*** -0.0582*** it (0.012) (0.019) (0.019) Q 0.1064*** 0.0930*** it−1 (0.004) (0.004) CF 0.0937*** it−1 (0.006) Intercept 0.4696*** 0.1411*** 0.1146*** (0.007) (0.008) (0.008) Observations 36,246 36,246 36,246 R2 0.502 0.531 0.567 28

Table5 Investment Efficiency around theSarbanes-Oxley Act of2002 ThistableexaminetheeffectsofaccrualsmanagementoninvestmentbyestimatingchangesininvestmentefficiencyaroundtheimplementationofSOX.Column(1)reportsestimatesfromthefollowingregression: I K it =a i +g t +b 1 ·1 (SOX) +b 2 ·Q i,t−1 +b 3 Q i,t−1 ·1 (SOX) +b 4 ·CF i,t−1 +e it , i,t−1 . Columns(2)through(4)reportestimatesfromthefollowingdifference-in-differences(DID)regression: I K it =a i +g t +d ·1 (SOX) +n ·1 (SOX) ·1 (i=T) +h ·1 (SOX) ·1 (i=T) ·Q i,t−1 +b 1 ·Q i,t−1 +b 2 ·CF i,t−1 +e it , i,t−1 wherea andg arefirmandyearfixedeffects,1 isadummyvariable(Post-SOX)takingavalueofonein i t (SOX) theyearsfollowingtheimplementationofSOXandzerootherwise,and1 isadummyvariablesetequalto (i=T) oneforfirmsthatbelongtothetreatmentgroupandzeroforfirmsinthecontrolgroup. Thetreatmentgroupis U.S.firmsandthecontrolgroupconsistsoffirmslocatedintheUnitedKingdomandCanada. Standarderrors areclusteredatthefirmlevelandreportedinparenthesis. ControlSample (1) (2) (3) (4) USPre-SOX UK/Canada Canada UnitedKingdom Post-SOX -0.0281** 0.0860 0.0832 0.0876 (0.011) (0.077) (0.078) (0.077) Post-SOX×Q -0.0227*** (0.004) Post-SOX×Treatment -0.0327** -0.0649*** -0.0428** (0.015) (0.022) (0.017) Post-SOX×Treatment×Q -0.0339*** -0.0316*** -0.0357*** (0.005) (0.004) (0.005) Q 0.0816*** 0.0461*** 0.0754*** 0.0440*** it−1 (0.002) (0.004) (0.003) (0.004) CF 0.0777*** 0.0178*** 0.0786*** 0.0173*** it−1 (0.002) (0.002) (0.003) (0.002) FixedEffects Firm,Year Firm,Year Firm,Year Firm,Year Observations 99,657 113,951 90,617 105,919 R-squared 0.476 0.385 0.477 0.379 ***indicates 1%significance, **5%significance, and*10%significance. 29

Table6 Firm Characteristics by Corporate GovernanceSubgroups Thistablesortsfirm-yearobservationsintotwosubgroupsbasedoncorporategovernancemeasuresandreport the mean firm characteristics by the subgroups. Four governancemeasuresare employedincluding Gompers, Ishii,andMetrick(2003)’sgovernanceindexandthepresenceofthreeindividualprovisions. Alsoreportedare thedifferencesinfirmcharacteristicsbetweenthetwosubgroups. Thecorrespondingt-statisticsarereportedin parentheses.Seeappendixforvariabledescriptions. Accruals OtherFirmCharacteristics TA |DA| Investment lagQ lagCF GovernanceIndex Low 0.0016 0.0818 0.2871 2.0558 0.6416 High 0.0019 0.0765 0.2395 1.8342 0.4924 Difference(Low-High) -0.0003 0.0053 0.0476 0.2216 0.1492 t-statistic (-0.49) (2.52) (13.97) (11.28) (9.34) PoisonPill No 0.0020 0.0767 0.2730 2.0310 0.6320 Yes 0.0017 0.0805 0.2498 1.8498 0.4972 Difference(No-Yes) 0.0003 -0.0038 0.0232 0.1812 0.1348 t-statistic (0.53) (-1.80) (6.79) (9.24) (8.46) ClassifiedBoard No 0.0021 0.0820 0.2764 2.0243 0.5865 Yes 0.0016 0.0765 0.2483 1.8621 0.5357 Difference(No-Yes) 0.0005 0.0055 0.0281 0.1622 0.0508 t-statistic (0.75) (2.59) (8.16) (8.21) (3.16) GoldenParachutes No 0.0018 0.0810 0.2807 2.0831 0.5861 Yes 0.0018 0.0774 0.2463 1.8273 0.5375 Difference(No-Yes) -0.0001 0.0036 0.0344 0.2558 0.0486 t-statistic (-0.13) (1.68) (9.93) (12.90) (3.01) 30

Table7 Corporate GovernanceandEarnings Management This table reports estimation results of baseline regressions for subsamples sorted by corporate governance measures. Four governance measures are employed including Gompers, Ishii, and Metrick (2003)’s governance index and the presence of three individual provisions. The dependent variable is investmentrate. Timeandfirmfixedeffectsareincluded.Robuststandarderrorsareclusteredatthefirmlevelandreportedinparenthesis.Seetheappendix forvariabledescriptions. GovernanceIndex PoisonPill ClassifiedBoard GoldenParachute Low High No Yes No Yes No Yes Q 0.0389*** 0.0365*** 0.0418*** 0.0380*** 0.0472*** 0.0352*** 0.0449*** 0.0369*** it−1 (0.007) (0.005) (0.007) (0.007) (0.008) (0.006) (0.008) (0.005) CF 0.0881*** 0.0693*** 0.0826*** 0.0686*** 0.0847*** 0.0663*** 0.0825*** 0.0716*** it−1 (0.013) (0.011) (0.015) (0.011) (0.013) (0.011) (0.016) (0.009) |DA | 0.0089 -0.0001 -0.0221 -0.0013 0.027 -0.0343 -0.0147 -0.0124 it (0.031) (0.024) (0.030) (0.036) (0.029) (0.027) (0.043) (0.023) Q ·|DA | 0.0882*** 0.0849*** 0.0976*** 0.0470* 0.0854** 0.0808*** 0.0799*** 0.0606** it−1 it (0.034) (0.026) (0.032) (0.033) (0.036) (0.027) (0.024) (0.027) Q ·|DA |2 -0.0789*** -0.0740*** -0.0647** -0.0836*** -0.0857*** -0.0644*** -0.0606 -0.0638*** it−1 it (0.030) (0.020) (0.027) (0.026) (0.030) (0.019) (0.041) (0.020) Observations 7,463 9,776 7,595 9,644 7,250 9,989 6,874 10,365 R2 0.608 0.483 0.628 0.56 0.582 0.58 0.614 0.605 ***indicates 1%significance, **5%significance, and*10%significance. 31

Figure1. InvestmentRates around SOX: TreatmentandControl Firms Thisfigureplotsaverageinvestmentrates(I/K)forUSfirms(“treatment”)andCanadian/UKfirms(“control”). I/K 0.30 0.28 Out[10]= 0.26 0.24 Year 1998 1999 2000 2001 2002 2003 2004 2005 2006 Treatment Control 32

Figure 2. Kernel DensityEstimation: InvestmentRates This figure plots the Epanechnikovkernaldensity investmentrates for both US firms (“treatment”)and Canadian/UKfirms(“control”)fortheperiodbeforeandafterthepassingoftheSarbanes-OxleyAct.AKolmogorov- Smirnovtestfortheequalityofdistributionsisrejectedatthe1%levelforthetreatmentgroup. ControlGroup TreatmentGroup 4 3 2 1 0 0 1 2 3 x Pre−SOX Post−SOX 3 2 1 0 0 .5 1 1.5 2 x Pre−SOX Post−SOX 33

Appendix: Variable Descriptions Variable Description Investment CapitalExpenditures dividedbybeginning-of-year capitalmeasuredbyproperty, plant, andequipment. Q Bookvalueoftotalassetsminusthebookvalueofequityplusthemarketvalueof equityscaledbythebeginning-of-year totalassets. CashFlow EBITplusdepreciation andamortization minusinterestexpense, taxesand dividends scaledbybeginning-of-year capital. Leverage Totaldebt(long-term andshort-term) scaledbytotalassets GovernanceIndex Gompers,Ishii,andMetrick(2003)’sindexconstructed byassigning onepointfor theexistence (orabsence) ofeachcorporate governance provision andsummingthe pointsacrossall24provisions. Post-SOX Anindicator variablesettooneforyearsfollowing2003implementation ofSOX andzerootherwise. Restate Anindicator variablesettooneifmisreporting thatsubsequently resultsin restatements occursinthegivenfirm-yearandzerootherwise. 34