bédard et al - 2004 - the effect of audit committee expertise, independence and activity on aggressive earning management

13–35 The Effect of Audit Committee Expertise, Independence, and Activity on Aggressive Earnings Management Jean Bédard, Sonda Marrakchi Chtourou, and Lucie CourteauSUMMARY: This study i

Submitted: July 2002 Accepted: November 2003

Vol 23, No 2

September 2004

pp 13–35

The Effect of Audit Committee Expertise,

Independence, and Activity on Aggressive Earnings Management

Jean Bédard, Sonda Marrakchi Chtourou, and Lucie CourteauSUMMARY: This study investigates whether the expertise, independence, and activities

of a firm’s audit committee have an effect on the quality of its publicly released financial information In particular, we examine the relationship between audit committee charac- teristics and the extent of corporate earnings management as measured by the level of income-increasing and income-decreasing abnormal accruals Using two groups of U.S firms, one with relatively high and one with relatively low levels of abnormal accruals in the year 1996, we find a significant association between earnings management and audit committee governance practices.

We find that aggressive earnings management is negatively associated with the financial and governance expertise of audit committee members, with indicators of independence, and with the presence of a clear mandate defining the responsibilities of the committee The association is similar for both income-increasing and income-de- creasing earnings management, suggesting that audit committee members are con- cerned with both types of earnings management and do not exhibit an asymmetric loss function similar to that of auditors.

Keywords: audit committee; financial expertise; earnings management; abnormal

accruals.

Data Availability: The data used is from public sources identified in the manuscript.

INTRODUCTION

Concerns about earnings management (e.g., Levitt 1998) and recent high-profile accounting

scandals have led most of the investing community to call for more effective audit tees as a mean to improve the quality of financial statements (e.g., Blue Ribbon Committee[BRC] 1999; Securities and Exchange Commission [SEC] 2000) In response to these calls, regula-tors have adopted regulations on the functioning of audit committees in a number of areas includingthe expertise of their members, their independence, and their activities The latest example of such

commit-Jean Bédard is Professor at Université Laval, Sonda Marrakchi Chtourou is Assistant Professor at the Faculté des Sciences Economiques et de Gestion de Sfax, and Lucie Courteau is Associate Professor at the Free University of Bozen-Bolzano.

We thank Mark DeFond (associate editor), the two anonymous reviewers, Ann Gaeremynck, as well as the accounting workshop participants at Katholieke Universiteit Leuven, Université Laval, Université Pierre Mendes-France, Université Montesqieu, Universiteit Maastricht, and the participants at the Auditing Section Midyear Meetings for their comments We acknowledge the financial support of the Social Sciences and Humanities Research Council of Canada.

14 Bédard, Chtourou, and Courteau

regulation in the U.S., the Sarbanes-Oxley Act of 2002 (hereafter SOX), requires that at least oneaudit committee member have financial expertise, that all the members be independent from thefirm’s management, and that the committee oversee the accounting and financial reporting processes

as well as the audit of the financial statements

While prior research on actual fraudulent financial reporting deficiencies provides evidence that

is generally consistent with the assertion that some of the practices recommended or required byregulators are associated with lower likelihood of fraud (Beasley 1996; Abbott et al 2004), there arequestions as to whether they also reduce less spectacular forms of earnings management Klein(2002) provides some evidence on this issue In her examination of the association between auditcommittee independence and earnings management for a sample of S&P 500 firms, she finds asignificant association between abnormal accruals and the presence of a majority of independentdirectors on the committee, but “no meaningful relation between abnormal accruals and having anaudit committee comprised solely of independent directors” (Klein 2002, 389) Thus, evidence isneeded on the possible effects on earnings management of SOX requirements (financial expertise,

100 percent of independent members, and oversight) and of other audit committee best practices

We investigate the relation between, on the one hand, the audit committee’s expertise (financial,governance, and firm-specific expertise), independence, and activities and, on the other hand, ag-gressive earnings management on a sample of 300 U.S firms The sample is composed of threegroups, one with aggressive income-increasing earnings management, one with aggressive income-decreasing earnings management, and a third group of firms with low levels of earnings management

in the year 1996 Earnings management is measured as abnormal accruals estimated with a sectional version of the Jones (1991) model

cross-Controlling for specific motivations that firms may have to manage earnings, and for alternativecontrol mechanisms, as well as for variables that have been found to affect the reliability of abnormalaccruals measurement, we test whether recommended governance practices for audit committees areassociated with a lower likelihood that the firm be in one of the groups with high levels of earningsmanagement A 1996 sample has the advantage of allowing us to examine firms that voluntarilyadopted the best governance practices before some of them were mandated by stock exchanges inDecember 1999 Thus, we can test the effectiveness of these practices on a cross-section of firmsduring the same period and increase the power of our tests by limiting the symbolic display ofconformity associated with mandatory rules (Kalbers and Fogarty 1998)

Our results suggest that an audit committee whose members have more expertise is moreeffective in constraining earning management Specifically, we find that the presence of at least onemember with financial expertise, which is now required by SOX, is associated with a lower likeli-hood of aggressive earnings management, and so is the level of governance expertise in the commit-tee The association between the level of firm-specific expertise and the probability of earningsmanagement, however, is significant only for income-decreasing accruals

Regarding independence, our results generally support the SOX requirement that all members ofthe audit committee be independent Contrary to Klein (2002) whose findings suggest that thecritical threshold for the number of independent directors on the audit committee is 50 percent ratherthan 100 percent, we find no significant effect for a committee composed of 50-99 percent indepen-dent members, but a significant reduction in the likelihood of aggressive earnings management when

100 percent of the members are independent We also find that the percentage of stock options thatcan be exercised in the short term by independent audit committee members is associated with ahigher likelihood of aggressive earnings management This result provides some support for the

U.K Combined Code (Financial Reporting Council [FRC] 2003) provision that the remuneration of

outside directors should not include stock options

Two aspects of the audit committee’s activity, its size and the frequency of its meetings, do notseem to affect the likelihood of aggressive earnings management For the third aspect of committee

activity, the responsibility of overseeing both the financial reporting and the audit processes, we find

a significantly negative association with the likelihood of aggressive earnings management This lastresult lends support to the SOX requirement that the role of the audit committee include the oversight

of both financial reporting and the audit process

While most studies of earnings management consider either only income-increasing accruals orthe magnitude of accruals irrespective of their direction, we examine negative (income-decreasing)and positive (income-increasing) earnings management separately We find that except for the auditcommittee oversight responsibilities, the effects are not statistically different between the two groups

of firms This suggests that audit committee members are concerned with both income-increasingand income-decreasing earnings management and do not exhibit an asymmetric loss function (Antleand Nalebuff 1991)

Our results are subject to the inherent limitations of our measure of earnings management.Abnormal accruals are subject to measurement errors that can lead to erroneous inference if themeasurement error is correlated with the audit committee characteristics (Klein 2002; Kothari et al.forthcoming) While we control for possibly omitted variables that are found in prior literature toaffect the reliability of the measure of abnormal accruals, there is always a possibility that our resultsare caused by measurement error

This paper contributes to the literature on the association between audit committee tics and earnings management in three ways First, while most studies on audit committees focus onthe independence of committee members, we also examine their expertise and the extent of theiroversight mandate, two aspects that are emphasized in the Sarbanes-Oxley Act of 2002 Second, weexamine separately firms that, in 1996, showed evidence of income-increasing and income-decreas-ing earnings management Firms often claim that income-decreasing abnormal accruals are indica-tive of conservative reporting behavior and Nelson et al (2002) find that auditors are more likely torequire adjustments to positive than negative accruals Our results contradict both as we find almost

characteris-no significant difference in the association of committee characteristics with the two types of ings management Third, our sample includes firms of various sizes While Klein (2002) studies asample of firms from the S&P 500 for 1992 and 1993 (an average of 346 firms per year), our sampleincludes 300 firms of different sizes for 1996 Our sample firms’ median assets are $51 millionwhereas Klein’s (2002) smallest firm has assets of $179 million Since the audit committee require-ments apply to firms of all sizes, our sample allows us to test the effect of these requirements onsmaller firms, which have also been found to be more prone to earnings management

earn-Overall, our results lend support to the assumptions underlying the SOX requirements that bothexpertise and independence are important characteristics for an audit committee to effectively moni-tor the financial reporting and audit processes They can be used by other regulators that arecontemplating similar rules For example, in Canada the proposed rules on audit committees requirethat all members of the committee be independent, but does not require financial expertise (OntarioSecurities Commission [OSC] 2003)

The remainder of the paper is organized as follow The next section provides the motivation forthe predicted association between audit committee characteristics and earnings management Thethird section discusses sample selection and research design Results are presented in the fourthsection and conclusions in the last section

THE ROLE OF THE AUDIT COMMITTEE IN MITIGATING

EARNINGS MANAGEMENT Earnings Management

Earnings management generally implies a “purposeful intervention in the external financialreporting process, with the intent of obtaining some private gain” (Schipper 1989, 92) Althoughmanagement may intervene in the process to signal private information and make the financial

16 Bédard, Chtourou, and Courteau

reports more informative for users, we concentrate on the negative aspect of earnings management,i.e., “to mislead stakeholders (or some class of stakeholders) about the underlying economic perfor-mance of the firm” (Healy and Wahlen 1999, 368)

The audit committee’s primary role is to help ensure “high quality financial reporting”(PricewaterhouseCoopers 1999, 7) As indicated by the BRC (1999, 7), the audit committee is the

“ultimate monitor of the [financial reporting] process.” The committee may reduce opportunisticearnings management by “evaluating the competence and independence of the external auditors,” byengaging in proactive discussions with company management and outside auditors regarding keyaccounting judgments, and by probing “to find out the nature and extent of issues that managementand the auditors gave considerable attention to” as well as “the outcome of these discussions”(Herdman 2002)

The literature concerning audit committees suggests three main categories of factors that mightaffect their capacity in reducing earning management: the expertise of the members, and the indepen-dence and activity level of the committee Therefore we test the following three hypotheses:

H1: Firms with expert audit committee members are less likely to engage in aggressiveearnings management

H2: Firms with an independent audit committee are less likely to engage in aggressiveearnings management

H3: Firms with an active audit committee are less likely to engage in aggressive earningsmanagement

We consider income-decreasing abnormal accruals separately because they are frequently used

by managers For example, a survey by Nelson et al (2002) indicates that 31 percent of the earningsmanagement attempts are income-decreasing compared to 53 percent that are income-increasing.Managers have various motivations to reduce earnings They may want to reduce the value of thestocks prior to a management buyout (Perry and Williams 1994), to reduce the risk of adversepolitical consequences (Cahan 1992; Jones 1991; Key 1997), or simply to create opportunities toincrease income in future periods (Levitt 1998) For example, the SEC (2003a) alleges that Xeroxcreated “cushion” reserves that it used when necessary to pump up its earnings by nearly $500million in order to meet earnings targets It is also true that the accruals used in one year to increaseearnings must be reversed in the following years, decreasing the earnings by the same amount.Generally, previous research suggests that auditors try to constrain earnings management to agreater extent if it is income-increasing than if it is income-decreasing For example, Nelson et al.(2002) find that auditors’ adjustment rate for income-increasing earnings management attempts ishigher than for income-decreasing attempts and Francis and Krishnan (1999) find that high income-increasing accruals are more likely to result in auditor reporting conservatism This greater attention

to income-increasing earnings management may be associated with professionally mandated cism (Braun 2001) and auditors’ perception that litigation is more likely to occur when income isoverstated (Myers et al 2003) In their study of auditor tenure, however, Myers et al (2003) find that

skepti-if an auditor remains longer with a firm, then he or she is more likely to restrict management frommaking extreme reporting decisions, both income-increasing and income-decreasing

As in the case of auditors, it is possible that the audit committee restrains income-increasing to agreater extent than income-decreasing earnings management because the members have an asym-metric loss function: the likelihood of attracting media attention or being sued could be higher incases of income-increasing earnings management

H4: The ef fect of audit committee characteristics is larger for income-increasing thanfor income-decreasing earnings management

Audit Committee Expertise

In order to fulfill their responsibilities for monitoring internal control and financial reporting,audit committee members should possess the necessary expertise We examine three aspects of theirexpertise: financial, governance, and firm-specific expertise

First, based on the requirement by Section 407 of SOX, we investigate the effect of the presence

of at least one member with financial expertise There is some empirical evidence that this ment is effective for extreme events such as SEC violations and earnings restatement (e.g., McMullenand Raghunandan 1996) While there is some experimental evidence that financial expertise isassociated with a higher likelihood that the committee support the auditor in an auditor-corporatemanagement dispute (DeZoort and Salterio 2001), greater focus on concerns that are critical for thequality of financial reporting, and more structured discussion on reporting quality (McDaniels et al.2002), to our knowledge there is no empirical evidence on the effectiveness of this requirement inpreventing less spectacular cases of earnings management

require-Several authors suggest that the managerial labor market for outside directorships provides anincentive to monitor effectively by rewarding effective outside directors with additional positions asdirectors and disciplining those who have a record of poor monitoring performance (Fama andJensen 1983; Milgrom and Roberts 1992) For example, outside directors of firms charged withaccounting and disclosure violations by the SEC are more likely than others to lose their otherdirectorships (Gerety and Lehn 1997) Additional directorships not only signal outside directors’competence to the managerial labor market, but it also helps them to acquire governance expertiseand to gain knowledge of best board practices On the other hand, if the number of other director-ships is too large, then it may reduce the time the director can devote to the particular firm, thusdecreasing the committee’s governing effectiveness (Morck et al 1988; Beasley 1996) Conse-quently, additional directorships may improve effectiveness up to a point, but beyond that point, thecommittee may be penalized because of the time and effort absorbed by other directorships.The experience of independent directors on the company’s board allows them to develop theirmonitoring competencies while providing them with some firm-specific expertise such as knowledge

of the company’s operations and its executive directors Thus, as their experience increases, theybecome more effective at overseeing the firm’s financial reporting process On the other hand, overtime the audit committee members may become more complacent, offsetting the knowledge effect.Previous research results support the knowledge effect For example, Kosnik (1987) finds that thelonger the average tenure of outside directors, the more likely the company is to resist greenmailpayments and Beasley (1996) finds that the likelihood of financial reporting fraud is a decreasingfunction of the average tenure of outside directors

Audit Committee Independence

To fulfill its oversight role and protect the interest of shareholders, the audit committee must beindependent of the firm’s management We consider two aspects of independence: the number ofnonrelated outside members and whether these members participate in the firm’s stock option plans.Section 301 of SOX requires that all members of the audit committee be independent Severalstudies find an association between the proportion of nonrelated outside directors on the auditcommittee and some indicators of reporting quality such as the probability of SEC enforcementaction (Wright 1996) and the size of abnormal accruals (Klein 2002) Klein (2002) however, finds

that it is the presence of a majority of independent directors on the committee, rather than 100

percent, that seems to have a significant effect on the level of abnormal accruals To provideevidence on this issue, we examine the effect of both thresholds (50 percent and 100 percent) on thecommittee’s effectiveness in monitoring the level of earnings management

Stock option schemes may compromise committee members’ independence While executivestock options are designed to align the manager ’s interests with those of the shareholders, they

18 Bédard, Chtourou, and Courteau

sometimes have the opposite effect For example, Safdar (2003) finds that executives managediscretionary accruals prior to exercising substantial portions of their outstanding stock options.Even if no research has examined the effect of stock options in the specific case of outside directors,

the U.K Combined Code on Corporate Governance (FRC 2003) states that their remuneration

should not include stock options We believe that such a practice is even more important for auditcommittee members because it is their duty to monitor the quality of the financial reports Hence, weexpect that outside directors with options that can be exercised currently or in the short run are lesseffective in curtailing income-increasing earnings management, especially if the options are in-the-money or at-the-money (i.e., the current stock price is higher than or equal to the exercise price of theoptions) However, even if they can be exercised in the short term, Huddart and Lang (1996) showthat options are not necessarily exercised soon after their vesting date If that is the case or if theoptions are out-of-the-money, then it may be in the interest of committee members to allow income-decreasing accruals in order to accumulate reserves to be used in subsequent years, when an increase

in earnings would enhance the value of the options

Audit Committee Activity

Expertise and independence will not result in effectiveness unless the committee is active Weexamine three aspects of its level of activity: the duties it has to perform, the frequency of itsmeetings, and its size The duties of an audit committee can be classified into three categories(Verschoor 1993; Wolnizer 1995): oversight of the financial statements, of the external audit, and ofthe internal control system (including internal auditing) We focus on the first two because they arethe most relevant for income management SOX (U.S House of Representatives 2002, Section 2)states that the purpose of an audit committee is to oversee the accounting and financial reportingprocesses of the company as well as the audit of its financial statements Furthermore, the BRC(1999) recommends that the responsibilities should be memorialized in a formal charter approved bythe board of directors A formal charter not only provides guidance to members as to their duties, but

it is also a source of power for the audit committee Kalbers and Fogarty (1993) find that a formalwritten charter establishing its responsibilities plays an important role in the power of the auditcommittee and that its perceived effectiveness is significantly related to this concept of power.The second dimension of committee activity we examine is the frequency of its meetings Anaudit committee eager to carry out its functions of control must maintain a constant level of activity(National Commission on Fraudulent Financial Reporting [NCFFR] 1987) and best practices sug-gest three or four meetings a year (Cadbury Committee 1992; KPMG 1999) McMullen andRaghunandan (1996) show that the audit committees of firms that are facing SEC enforcementactions or restating their quarterly reports are less likely to have frequent meetings The committees

of only 23 percent of their problem companies met more than twice a year compared to 40 percentfor the other firms Abbott et al (2004) find similar results in a more recent sample

As indicated by the BRC (1999, 26) “Because of the audit committee’s responsibilities and thecomplex nature of the accounting and financial matters reviewed, the committee merits significantdirector resources […] in terms of the number of directors dedicated to [it].” Best practices suggest

at least three members (Cadbury Committee 1992; BRC 1999), which provides the necessary strengthand diversity of expertise and views to ensure appropriate monitoring The benefit of additionalmembers, however, must be weighed against the incremental cost of poorer communication anddecision making associated with larger groups (Steiner 1972; Hackman 1990) The objective is tohave a committee not so large as to become unwieldy, but large enough to ensure effective monitor-ing In general, it is recommended to limit the size of the committee to five (Arthur Andersen 1998)

or six members (National Association of Corporate Directors [NACD] 2000) While limited, theevidence suggests that size may matter For example, Archambeault and DeZoort (2001) find asignificantly negative relationship between committee size and suspicious auditor switches, butAbbott et al (2004) find no significant association between size and earnings misstatements

RESEARCH DESIGN

The objective of this study is to determine whether good audit committee practices reduce thelikelihood of earnings management as measured by abnormal accruals Our sample is drawn from thepopulation of U.S firms whose financial data appear on Compustat in 1996 From this population,

we identify the 100 firms with the highest income-increasing abnormal accruals, the 100 firms withthe highest income-decreasing abnormal accruals, and the 100 with the lowest abnormal accruals

We categorize the first two groups as using “aggressive earnings management” (AEM) and the thirdgroup as having “low earnings management” (LEM) Several studies indicate that all existing modelsmeasure abnormal accruals with error (for example, Dechow et al 1995) By including only firmswith high and low abnormal accruals in the sample, we seek to improve the power of our tests bymitigating the measurement error problem.1

Abnormal Accruals Estimation

Our sample is based on the complete set of firms on Compustat with a December 31, 1996 end and complete accruals data for 1996 We exclude firms from the regulated (SIC 4000 to 4900),financial (SIC 6000 to 6900), and government (SIC 9900) sectors because their special accountingpractices make the estimation of their abnormal accruals difficult The abnormal component of totalaccruals is estimated with the modified Jones (1991) cross-sectional model (DeFond and Jiambalvo1994; Francis et al 1999; Becker et al 1998) This requires the estimation of a cross-sectionalregression for each industry (two-digit SIC codes), so we eliminate industries with less than tenfirms These requirements leave 3,947 observations for the calculation of abnormal accruals

year-Abnormal accruals (AbnAccruals) for each firm i in industry j are defined as the residual from

the regression of total accruals (the difference between cash from operations and net income) on twofactors that explain nondiscretionary accruals, the change in revenue and the level of fixed assetssubject to depreciation All variables are deflated by total opening assets to reduce heteroscedasticity

AbnAccruals ijt = TAC ijt /A ijt–1–[αj (1/A ijt–1) + β1j (∆RE ijt /A ijt–1) + β2j (PPE ijt /A ijt–1)] (1)where:

AbnAccruals ijt = abnormal accruals for firm i from industry j in year t;

TAC ijt = total accruals for firm i from industry j in year t;

A ijt– 1 = total assets for firm i from industry j at the end of year t–1 (Compustat item 6);

∆RE ijt = change in net sales for firm i from industry j between years t–1 and t

39 industries that meet our requirements contains 3,451 firms Abnormal accruals are then computedfor each firm from Equation (1)

Panel A of Table 1 reports descriptive statistics for the entire sample of 3,451 firms The average

1 Focusing on extremes (top and bottom deciles) may render our estimates of abnormal accruals vulnerable to the bias caused by extreme cash flows and earnings We control for this possible bias with a regression approach, including both cash flows and earnings as control variables in the regression models, and with Kasznik’s (1999) matched-portfolio approach.

2 Firms with very large earnings or cash flows from operations have been shown to bias the estimation of discretionary accruals For detecting outliers, we use three criteria: Cook’s distance, Studentized residuals, and hat matrix An observa- tion is excluded from the sample as an outlier if it fails two out of these three tests.

20 Bédard, Chtourou, and Courteau

(median) abnormal accrual is 0.000 (0.017) and 58 percent of the firms have positive abnormalaccruals The average (median) current earnings and cash flows from operations relative to openingtotal assets are –0.099 (0.031) and –0.029 (0.062), respectively Compared to Klein (2002), thepopulation we use to estimate the abnormal accruals includes smaller firms (with median asset of

$82 million whereas Klein’s smallest firm has assets of $179 million), with lower profitability(Klein’s average [median] earnings deflated by opening assets are 0.056 [0.048]) and generatinglower levels of cash flows from operations for each dollar of opening assets (Klein’s sample firmshave average [median] of 0.117 [0.107])

While we take precautions (e.g., dropping outliers and firms with extreme earnings and cashflows) to avoid known bias in the estimation of abnormal accruals, the possibility of measurementerror is always an issue in studies using the modified Jones model to measure discretionary accruals

in tests of earnings management As noted by Klein (2002) any proxy for abnormal accruals yieldsbiased metrics if the measurement error in the proxy is correlated with omitted variables Priorstudies suggest that the measurement error is correlated with current earnings, previous year’searnings, changes in earnings, current cash flows from operations, changes in cash flows, changes intotal accruals, firm size, and previous year ’s return on assets (Kasznik 1999; Jeter and Shivakumar1999; Klein 2002; Kothari et al forthcoming)

Panel B of Table 1 reports the Spearman correlations of total and abnormal accruals with thesevariables The first two columns show that all variables are significantly correlated with the absolutevalues of total and abnormal accruals, suggesting that the measure might be biased We control forthis by including some of the possible omitted variables as control variables in the regression modelsused to test the effect of the audit committee on the measure of abnormal accruals Because we focus

on extremes (top and bottom deciles) in our analysis of the effect of audit committee characteristics,our estimates of abnormal accruals are particularly vulnerable to the bias caused by cash flows andearnings For example, Jeter and Shivakumar (1999) show that the cross-sectional Jones modelyields systematically positive (negative) estimates of abnormal accruals for firms whose cash flowsare below (above) the industry median Panel C shows that the possible correlated variables arecorrelated to each other, suggesting that they capture much of the same processes Consequently,earnings and cash flows are included as control variables in the regression model used to test theeffect of audit committees on the measure of abnormal accruals, along with total assets and previousyear ’s ROA Because the top (bottom) decile of the abnormal accruals distribution is more likely tocontain firms with negative cash flows (earnings), we also include two indicator variables in theregression models: one for the presence of negative cash flows and one for negative earnings.3

Aggressive Earnings Management

In order to detect aggressive earnings management, we rank the 3,451 remaining firms on thesize of their abnormal accruals and select two subsamples of 100 firms each from both ends of thedistribution (i.e., the 100 largest positive and the 100 largest negative abnormal accruals) These 200firms comprise the aggressive earnings management (AEM) subsample We then select anothersubsample of 100 firms with the lowest level of abnormal accruals centered around zero (50 negativeand 50 positive), which form the low earnings management (LEM) subsample Figure 1 shows themean and median levels of abnormal accruals for each decile of the distribution The subsample ofnegative (positive) AEM is drawn from the first (tenth) decile, which has mean and median abnormalaccruals of –0.38 and –0.28 (0.27 and 0.22), respectively The LEM subsample is drawn fromdeciles 4 and 5, which have mean (median) abnormal accruals of –0.02 (–0.02) and 0.01 (0.01),

3 In a second attempt to control for omitted variables, we use Kasznik’s (1999) matched-portfolio method, adjusting the value of each firm’s abnormal accruals by the median abnormal accruals for a portfolio of firms matched on the absolute level of earnings The correlation of possible correlated variables is lower with adjusted than with unadjusted abnormal accruals with the exception of cash flows (for which it is larger) To control for the potential effect of these variables, we keep the same control variables, except earnings, when the adjusted abnormal accruals are used to test the effect of the audit committee The results are substantially the same with the two approaches (see footnote 9).

TABLE 1 Information for the Overall Population Panel A: Descriptive Statistics on Accruals and Abnormal Accruals

Panel B: Spearman Correlations of Total and Abnormal Accruals with Possible Correlated Variables

Panel C: Spearman Correlations Among Possible Correlated Variables

Variable |Earnings t–1| |∆∆∆∆∆Earnings| |CashFlows| |∆∆∆∆∆CashFlows| |∆∆∆∆∆Total Accruals|

All variables except Total Assets and ROA are deflated by opening asset Changes in the value of a variable between 1996

and 1995 are indicated by ∆, the absolute value of a variable by |variable|, t refers to the year 1996 and t–1 to the year

CashFlows = cash flows from operations from cash flows statement;

Ln(Assets) = natural log of total assets; and

ROA = net income before extraordinary items for 1995 divided by total assets at the end of 1995.

22 Bédard, Chtourou, and Courteau

respectively

To collect 100 observations with full governance data in each category, we have to consider 203observations for the positive subsample (AEM+), 286 for the negative subsample (AEM–), and 160for the LEM subsample Observations have to be dropped because of missing proxy statements (45,

113, and 37 firms, respectively), absence of an audit committee (7, 7, and 1), missing information ondirectors’ stock option and stock holdings (4, 5, and 4) and changes in the board of directors during

1996 in firms for which the 1995 proxy statement is not available (45, 59, and 18)

Because it is not randomly selected, our sample is not necessarily representative of the tion Table 2 shows some statistics on the nature of the firms in the population of 3,451 firms and thesample of 300 firms used in this study The distribution of industry composition by two-digit SICcodes shows that the percentages of firms in each sector are similar between our sample and thepopulation, except for the service industry which is overrepresented in the sample The other statis-tics show that the sample is different from the population The sample firms are smaller and theirmean growth is lower but their median growth is larger suggesting that some of the firms in thepopulation have large growth rates.4 The mean loss is larger for the sample, but the median earningsare the same Both the mean and median cash flows are lower in the sample than in the population

popula-Audit Committee Variables

All the audit committee characteristics are hand-collected from proxy statements for 1996 Foreach committee member, we determine whether he or she holds a professional certification inaccounting (CPA) or financial analysis (CFA) or has experience in finance or accounting Ourdefinition of financial expertise is more restrictive than that of the BRC in that it excludes priorexperience as a CEO We consider that the CEO position provides financial literacy but not exper-

tise FinExpertise is coded 1 if at least one audit committee member has accounting or financial expertise, and 0 otherwise The committee members’ competence is also measured by GovExpertise,

which is the average number of directorships held by nonrelated outside directors in unaffiliated

firms, and by FirmExpertise measured as the average number of years of board service for nonrelated

outside committee members

Consistent with prior research and the requirements of SOX, we classify directors as executives,

4 Excluding two extreme observations from the population reduces the mean from 1.20 to 0.56.

FIGURE 1 Mean and Median Abnormal Accruals by Decile over the 3,451 Sample Firms a

a The population consists of 3,451 firms from Compustat with a December 31, 1996 year-end, complete accruals data excluding the regulated, financial, and government sectors after the removal of firms with extreme income and cash flows from operation and outliers The deciles of the distribution are obtained by ranking the remaining firms on their

Abnormal Accruals Abnormal Accruals is the abnormal component of total accruals estimated with the Jones

cross-sectional model (see Equation (1)).

related, or independent outsiders.5 Related directors are those who have business relationships withthe firm or its managers, although they are not employees of the firm Consultants, suppliers,bankers, former employees, and managers’ family members as well as employees of other firms thathave a business relationship with the firm are part of this group The independent outsider groupincludes all directors who seem to have no relation with the firm other than their position asdirectors To clarify the affiliations disclosed in the proxy statements, we obtain the list of each firm’s

affiliated companies in Who Owns Whom (1996) Two dichotomous variables are used to measure independence: 100% Ind is coded 1 if the audit committee is composed solely of nonrelated outside directors, and 0 otherwise, and 50-99%Ind is coded 1 if the committee is composed of more than 50

percent but less than 100 percent of nonrelated outside directors, and 0 otherwise The adverse effect

on independence potentially caused by stock option schemes (StockOptions ) is measured as the ratio

of stock options that can be exercised in the 60 days following fiscal year-end to the total of optionsand stocks held by nonrelated outside committee members

We measure the three dimensions of audit committee activity as follows The presence of a clearmandate defining the responsibilities of the audit committee is measured with the indicator variable

Mandate, which is coded 1 if the proxy statement indicates that the committee is responsible for the

5 In order to be considered as independent for purposes of SOX, an audit committee member may not accept any consulting, advisory, or other compensatory fee from the firm or be affiliated to the firm or any of its subsidiaries in any other way (SOX 2002).

TABLE 2 Descriptive Statistics for the Population and the Sample Panel A: Industry Composition

Panel B: Financial Characteristics

c The sample consists of three groups of 100 firms, each based on their level of abnormal accruals for the year 1996: the group with the highest positive (income-increasing) accruals, the group with the highest negative (income-decreasing) accruals, and the group with the accruals closest to 0.