Engel

ARTICLE IN PRESS

Contents lists available at ScienceDirect

Journal of Accounting and Economics

Journal of Accounting and Economics 49 (2010) 136–154

0165-41

doi:10.1

� Cor

E-m1 Fo

conside

journal homepage: www.elsevier.com/locate/jae

Audit committee compensation and the demand for monitoring ofthe financial reporting process

Ellen Engel a,�, Rachel M. Hayes b, Xue Wang c

a University of Chicago Booth School of Business, USAb David Eccles School of Business, University of Utah, USAc Goizueta Business School, Emory University, USA

a r t i c l e i n f o

Article history:

Received 8 March 2008

Received in revised form

31 July 2009

Accepted 19 August 2009Available online 29 August 2009

JEL classification:

G30

M41

M49

Keywords:

Audit committees

Board of director compensation

Audit fees

01/$ - see front matter & 2009 Elsevier B.V. A

016/j.jacceco.2009.08.001

responding author.

ail address: [email protected] (E.

r example, Klein (2002a) examines the relat

r the relation between audit committee char

a b s t r a c t

We examine the relation between audit committee compensation and the demand for

monitoring of the financial reporting process. We find that total compensation and cash

retainers paid to audit committees are positively correlated with audit fees and the

impact of the Sarbanes-Oxley Act, our proxies for the demand for monitoring. Our

results are robust to the inclusion of audit committee quality, measured as the

committee chair financial expertise. Our results suggest a recent willingness by firms to

deviate from the historically prevalent one-size-fits-all approach to director pay in

response to increased demands on audit committees and differential director expertise.

& 2009 Elsevier B.V. All rights reserved.

1. Introduction

In this paper, we study compensation of audit committee members of public firms. Specifically, we examine how cross-sectional variation in the demand for monitoring of the financial reporting process is associated with compensation foraudit committee members. Audit committees play a crucial role in firms’ financial reporting processes, and thus haveattracted considerable attention from researchers, especially in the wake of recent high profile financial reporting scandals.While there are large literatures both on audit committee characteristics and on compensation for directors, research hasnot, to date, examined how pay for audit committee members varies with the firm’s financial reporting environment.Recent literature on audit committees has focused instead on committee characteristics such as size and composition (Deliand Gillan, 2000; Klein, 2002b), and the relation between these characteristics and various outcomes.1 Meanwhile,research on compensation for directors has tended to ignore much of the within-firm heterogeneity in director pay,focusing instead on compensation for a representative director at a given firm (see, for example, Farrell et al., 2008; Linnand Park, 2005).

We investigate whether firms that face a higher demand for monitoring of the financial reporting process pay highercompensation to their audit committees. In addressing this question, we incorporate the effects of audit committee director

ll rights reserved.

Engel).

ion between audit committee independence and earnings quality, while Carcello and Neal (2000)

acteristics and auditor reporting behavior.

www.elsevier.com/locate/jae

dx.doi.org/10.1016/j.jacceco.2009.08.001

mailto:[email protected]

ARTICLE IN PRESS

E. Engel et al. / Journal of Accounting and Economics 49 (2010) 136–154 137

quality. A key part of our analysis also examines within-firm heterogeneity in director compensation—in particular, auditcommittee pay relative to compensation committee pay (the audit committee pay premium)—and how the auditcommittee pay premium relates to the demand for monitoring of the financial reporting process.

A greater demand for monitoring of the financial reporting process can arise from a number of characteristics of the firmand its environment or from external forces. In developing proxies for the demand for monitoring, we consider both firm-specific and external forces. Firm-specific forces include the complexity of the business and the firm’s organizationalstructure, the strength of internal control systems, financial reporting quality and litigation risk. These factors are amongthose that impact the overall transparency and riskiness of the financial reporting process. We argue that the lower thetransparency and the greater the risk of the financial reporting process, the greater is the demand for monitoring of thefinancial reporting process by investors and other stakeholders. We expect total annual audit fees paid to the external auditfirm for performing audit services to capture these firm-specific and related environmental forces. Accordingly, our firm-specific proxy for the demand for monitoring of the financial reporting process is firm-size-deflated total annual audit fees.

External to the firm, we consider the impact of the requirements of the Sarbanes-Oxley Act (SOX), and the factors that led tothe legislation, on the demand for monitoring of the financial reporting process. SOX addresses widespread concerns ofregulators and investors about the financial reporting process, including the role of the audit committee. Specifically, SOXrequires that all members of the audit committee be independent and that the company’s annual report disclose whether amember of the audit committee is a financial expert. These provisions, along with overall investor pressure to improvecorporate governance, are likely to create an increased demand in many firms for audit committee members to monitor thefinancial reporting process more diligently. Accordingly, our second proxy, an indicator variable for the post-SOX time period, isdesigned to capture factors relating to the enhanced monitoring environment, greater board accountability and increasedcredential requirements directed at audit committee members that are not expected to be fully reflected in the audit feevariable. This variable allows us to capture a potentially important external shift in the demand for monitoring of the financialreporting process that is triggered by the provisions of SOX and concurrent higher levels of investor scrutiny.

We expect that the factors that lead to greater demand for monitoring of the financial reporting process require anincreased commitment of time and effort by audit committee members. We hypothesize that, holding director qualityconstant, this effect leads to higher compensation for these directors. We control for director quality because auditcommittee compensation is likely related to both the demand for monitoring of the financial reporting process and thequality of the audit committee. Including an audit committee quality measure—specifically, the financial expertise of theaudit committee chair—allows us to investigate both effects.

To investigate our research questions, we collect compensation data for outside directors from proxy statements for thesample of ExecuComp firms between 2000 and 2004. We also collect data on the financial expertise of audit committeechairs during these years. Our analyses consider two main time periods: the pre-SOX period including 2000 and 2001, andthe post-SOX period from 2002 through the end of 2004. Our descriptive analyses document that total compensation foraudit committees increased significantly in the post-SOX period, with notable increases in the cash retainer and meetingfee components. We also find that a significantly larger percentage of firms have audit committee chairs with financialexpertise, and especially accounting expertise, in the post-SOX period.

We conduct two primary regression analyses. Our first set of regressions examines the determinants of the level of auditcommittee compensation, while our second set of regressions explains the difference between compensation of auditcommittee members and compensation committee members on the same board. In each regression, we include audit fees,as well as the post-SOX indicator, to capture the demand for monitoring of the financial reporting process. While ourprediction is that audit committee compensation is positively related to audit fees because the two variables areresponding similarly to underlying economic factors, we note that it is also possible that audit fees are endogenous in thecompensation regression. Consequently, before we perform our regression analyses, we instrument for audit fees in a two-stage least squares regression and conduct a Hausman test to determine whether audit fees are an endogenous regressor.The Hausman test fails to reject that audit fees are exogenous for all specifications. Thus, the use of ordinary least squaresregressions to test our hypotheses will yield consistent coefficient estimates. In the levels regressions, we find that totalcompensation (excluding meeting fees) and cash retainers paid to audit committee members are positively correlated withannual audit fees, our proxy for the demand for monitoring of the financial reporting process. These compensationmeasures are also positively associated with our measure of audit committee chair expertise. Further, our empiricalevidence indicates that the level of compensation for audit committees has increased substantially in the post-SOX periodcompared to the pre-SOX period.

Our regressions examining the differences in compensation between audit committees and compensation committeesare motivated by the idea that audit committees are more likely to be affected by the demand for monitoring of thefinancial reporting process, including the implications of SOX, than compensation committees. By using compensationcommittees as a control group, we control for systematic shocks to the market for outside directors and firm-specificfactors. The regression analyses using the difference in compensation support the predictions that firms with a higherdemand for monitoring of the financial reporting process are likely to pay a higher level of total compensation to auditcommittees relative to compensation committees. Further, these analyses document statistically significant increases inthe differences in total compensation and in the cash retainer component between the two committees from the pre- to thepost-SOX period. Additional analyses indicate that the significant relations between the audit committee pay premium andthe demand for monitoring of the financial reporting process exist only in the post-SOX period.

ARTICLE IN PRESS

E. Engel et al. / Journal of Accounting and Economics 49 (2010) 136–154138

In studying pay for audit committee members, we control for firm-wide factors that affect the overall level of directorcompensation by comparing pay for audit committee members to pay for compensation committee members. Ourregression analyses suggest a pay premium of $1,248 in total compensation earned by audit committee members relative tocompensation committee members in the post-SOX period, in contrast to almost identical pay of the two committees in thepre-SOX period. We note that this finding is surprisingly small in economic magnitude given the importance thatregulators, politicians, and the media have placed on audit committees in the post-SOX period. However, we argue that themere presence of within-firm director pay heterogeneity in the post-SOX period is of interest. Prior studies of directorpay—see, for example, Farrell et al. (2008)—have noted that firms design outside director compensation for a group ofdirectors, rather than tailoring compensation packages to the specific characteristics or outside opportunities of individualdirectors. Variation in annual director pay at a given firm typically comes through service on different committees or ascommittee chair or lead director.2 This approach to pay-setting seems to be quite different from that used for employees; inmany firms, there is substantial across-employee pay heterogeneity even within narrowly defined job categories (see Bakeret al., 1994). We also find that following SOX, the premiums paid to audit committee members relative to compensationcommittee members are related to the firm-specific proxy for the demand for monitoring of the financial reporting processand director-specific expertise in financial reporting. These results suggest that firms are showing a willingness to deviatefrom the historically prevalent one-size-fits-all approach to director pay, and to adjust director compensation packages tothe specific characteristics of individual directors. These findings may be particularly notable in light of recent accountingscandals that have raised questions about the effectiveness of audit committees. Given the role played by the auditcommittee in overseeing the firm’s financial reporting process, the incentives provided by compensation arrangements forthese directors seem to be taking on an increased importance.

The remainder of the paper is organized as follows: In Section 2 we review prior studies on compensation for outsidedirectors, and develop hypotheses. We discuss the sample and research methodology and provide descriptive statistics inSection 3. We discuss our empirical approach and present the main empirical results in Section 4. Section 5 concludes the paper.

2. Related research and hypothesis development

2.1. Related research

Our research relates to two strands of literature: research on outside directors, and research on audit committees.The vast majority of research on outside directors has focused on the determinants of board characteristics such as

composition or size and on how these board characteristics affect firm performance or observable actions of the board suchas CEO turnovers, the takeover market and executive compensation (see Hermalin and Weisbach, 2003, for a review).Another line of research on outside directors, motivated by Fama and Jensen’s (1983) conjecture that reputation and themarket for directors are the primary incentive mechanisms for outside directors, focuses on directors’ accumulation ofseats on additional boards and directors’ turnover around circumstances such as financial distress. Recently, with the trendtoward more equity-based compensation for outside directors, researchers have begun to examine compensation foroutside directors. Notably, Yermack (2004) studies incentives for outside directors by following a sample of directors forfive years after election and tracking director compensation, changes in equity ownership by directors, other board seatsobtained and departures from the board. In contrast to Fama and Jensen’s (1983) hypothesis, he finds that incentives fromcompensation and ownership account for more than half of total incentives for outside directors. This indicates theimportance of understanding compensation arrangements for outside directors.

Research on audit committees parallels research on boards of directors, and focuses on the determinants of auditcommittee characteristics and how these characteristics are related to financial reporting quality. Following the passage ofSOX, a growing number of papers have examined another feature of audit committees—the inclusion of a director withfinancial expertise. For example, Defond et al. (2005) examine the market reaction to the appointment of financial expertson audit committees. However, few papers study the compensation and incentives of audit committee members. Oneexception is Srinivasan (2005), who investigates whether audit committee members are held accountable for financialreporting failure by looking at director turnover and the loss of board positions in other companies when their companiesexperience accounting restatements. In general, however, the incentives provided specifically to audit committee membershave received little attention in the literature on directors. This is likely because, as noted above, differences in within-firmdirector pay were small and relatively uncommon until recently.

Recently, Linck et al. (2009) extend the literature on outside directors by studying the effects of SOX on variousdimensions of corporate boards, including director workload, structure, risk, compensation and turnover. Our paper issimilar to Linck et al. (2009) in that we explore the impact of regulatory rules on director compensation. However, giventhat the primary focus of SOX is on strengthening firms’ financial reporting processes, the effects appear to weigh mostheavily on audit committees. Thus, we are able to investigate the broader issue by examining a more targeted group of

2 These within-firm sources of pay variation have typically been ignored in prior research. For example, Farrell et al. (2008) and Linn and Park (2005)

analyze pay for a representative director at each firm, and Yermack (2004) notes that he ignores committee fees and meeting fees in order to keep the data

collection and analysis tractable.

ARTICLE IN PRESS


outside directors. More importantly, our paper also addresses a more general question of the relation betweencompensation for audit committees and demand for monitoring of the financial reporting process.

2.2. Hypothesis development

Our primary objective is to examine cross-sectional variation in compensation for audit committees. In particular, weare interested in how compensation arrangements of audit committees vary with the demand for monitoring of thefinancial reporting process in general and relative to the compensation of members of other board committees. In thissection, we discuss the channels through which the demand might affect the level of audit committee compensation, anddevelop hypotheses about this relation. We also discuss the effects of audit committee director quality on audit committeecompensation.

We expect that the demand for monitoring of the financial reporting process is high when a firm has complex businessoperations and is subject to great risk of financial misstatement. This setting requires outside directors who understandfinancial accounting issues in order to effectively monitor the financial reporting process. One might expect that thesefirms have a high demand for qualified directors and that the overall supply of such directors may be relatively limited. Weexpect these forces to lead to higher compensation for directors serving on audit committees of firms with a high demandfor monitoring of the financial reporting process.

In addition to being financially literate, the outside directors serving on audit committees of firms with a high demandfor monitoring of the financial reporting process are likely to spend more time and effort learning, communicating, andunderstanding firms’ financial reporting processes. Managerial productivity theory (Rosen, 1992) predicts a higherreservation wage for directors serving on audit committees in this case.

Another channel through which a high demand for monitoring of the financial reporting process might affect auditcommittee compensation is the increased risk of misstatement that is likely to be present in such firms. Previous research(Srinivasan, 2005) has suggested that audit committee members suffer reputation penalties as a result of financialreporting failure. Agency theory predicts that a risk-averse agent requires a higher risk premium when risk exposure isgreater. These arguments all suggest that there will be a positive relation between compensation paid to audit committeesand the demand for monitoring of the financial reporting process.

The effects described above are likely to be exacerbated by recent financial reporting failures and the concurrentpassage of SOX. As mentioned in the Introduction, SOX and the events leading to its passage have raised the demand formonitoring of the financial reporting process for all public companies: (1) SOX focuses on financial reporting and auditcommittees, requiring that all members of the audit committee be independent, and that the company’s annual reportdisclose whether a member of the audit committee is a financial expert. These requirements are likely to lead to a decreasein supply and an increase in demand for qualified directors serving on audit committees. (2) SOX and recent investorpressure have increased the workload of audit committees substantially. Linck et al. (2009) find that audit committeesmeet more than twice as often post-SOX as they did pre-SOX, and that the average number of directorships held by adirector on the audit committee decreased significantly. (3) SOX and the financial reporting failures leading to SOX havealso increased the risk exposure of audit committees. Again, Linck et al. (2009) document that average Director and Officerinsurance premiums increased by more than 150%, and the proportion of firms that experienced audit committee turnoverhas increased significantly in the post-SOX period compared to the pre-SOX period. We therefore expect that compensationfor audit committees increases in the post-SOX period.

Finally, it seems reasonable to expect audit committee compensation to be related to both the demand for monitoring ofthe financial reporting process and the quality or expertise of audit committees. If this is true, one might argue that thepositive relation we hypothesize between audit committee compensation and the demand for monitoring of the financialreporting process could also reflect a relation between compensation and audit committee quality. Thus, we include ameasure of audit committee financial expertise in addition to the demand for monitoring of the financial reporting proxiesin order to separately examine these effects in our regressions. If the relation is actually driven by committee financialexpertise, we expect this effect to be captured by the expertise variable. We hypothesize that both audit committeefinancial expertise and the overall demand for monitoring of the financial reporting process will be positively related toaudit committee compensation.

3. Sample, research methodology and descriptive statistics

3.1. Sample and data

We identify the sample of outside directors using ExecuComp firms covering the period from 2000 to 2004.3 Followingprevious studies on outside directors (Adams, 2003; Yermack, 2004), we exclude utilities (2-digit SIC 49) and financialinstitutions (1-digit SIC 6) from the sample because these firms tend to have different corporate governance structures thanfirms in non-regulated industries (Macey and O’Hara, 2003).

3 2000 is the first year that audit fee data are publicly available for US public companies.

ARTICLE IN PRESS


We then collect data on fees paid to members of audit and compensation committees of the board from corporate proxystatements. Specifically, we collect data on the annual cash retainer and equity retainer, the number of stock option grants,the number of stock grants, and the meeting fees paid to directors serving as chair of the audit committee, members of theaudit committee, chair of the compensation committee, and members of the compensation committee, respectively.4

Following Yermack (2004), we assume that when directors can choose between cash retainer and equity retainer, theychoose the maximum amount of cash pay permitted.5 For equity compensation awarded upon election as directors, weassume that the awards are made equally through the terms of the directorship. However, we exclude one-time/special/discretionary equity-based awards made on an irregular schedule, such as the one-time equity award when a director joinsthe board. We also do not consider insurance plans, retirement plans and charity matching contributions in thecompensation packages. We value stock grants by multiplying the number of shares by the closing stock price. For the valueof stock option awards, if the proxy statement discloses the number of stock options awarded, we use the Black–Scholesmethod to compute the value of option awards assuming that firms issue at-the-money options at the end of the fiscal yearand with a seven-year maturity. Volatility and dividend yields for each firm-year are obtained from the ExecuCompdatabase. If the proxy statement only discloses the dollar value of options, then we use the dollar value. In addition, weobtain information on CEO characteristics (i.e., tenure, equity ownership and role on board), other board-related data (e.g.,director independence and audit committee chair flag) from IRRC (the corporate governance database maintained by theInvestor Responsibility Research Center), and collect number of committee meetingsduring the year, audit fee data, and thebackground information of audit committee chairs from proxy statements.

To be included in the sample, financial and stock return data must be available from Compustat and CRSP, and data onthe parameters of the Black–Scholes option pricing model must be available from ExecuComp. This screening procedureleaves us with an initial sample of 5,465 firm-year observations. We conduct our basic regressions of the key relations withcontrols for various firm-specific characteristics on this broader sample. For our analyses that also control for auditcommittee financial expertise and those variables suggested by bargaining theory to explain compensation for auditcommittees, we also require that certain CEO and board member data be available from IRRC. This reduces the sample to3,295 firm-year observations, an average of roughly 660 firms per year over the five-year sample period.6

3.2. Research methodology

3.2.1. Variable measurement

In this section, we describe the measurement of audit and compensation committee compensation (our dependentvariables), and our proxies for the demand for monitoring of the financial reporting process and audit committee quality.

Committee compensation measures. Annual compensation for audit and compensation committees is composed of fivecomponents: cash retainer, equity retainer, stock grants, option grants and meeting fees. In our empirical analysis, we focuson total compensation, along with the cash component of total compensation, cash retainer. Total compensation in theregression analyses is measured as the sum of these components with the exception of meeting fees.7 Since thecompensation variables are highly right skewed, we use a logarithmic transformation. We collect compensationinformation for both the audit and compensation committee chairs and the other members of these committees. In manyinstances, committee chairs receive additional compensation recognizing the typically substantial additional effort chairsexert in preparation for meetings. In our main analyses, we focus on the chairs of audit committees (and chairs ofcompensation committees, in the difference regressions). Results based on committee members are similar.

Demand for monitoring of the financial reporting process proxies. We use two proxies to capture the demand formonitoring of a firm’s financial reporting process. Our first proxy is the amount of audit fees charged by the outsideauditors for performing the firm’s annual audit services (AuditFee). Our second proxy, PostSOX, captures the demand formonitoring of the financial reporting process relating to a regulatory shift in the overall demand for monitoring of thefinancial reporting process—the passage of the Sarbanes-Oxley Act. We elaborate on these proxies below.

4 Stock option awards include options and SARs; stock awards include common stock, restricted stock, deferred stock units and phantom stock units;

meeting fees include attending both board and committee meetings, but special committee meeting fees are excluded. The following committees are

defined as audit committee: audit committee, audit and legal committee, audit review committee, audit and ethics committee, audit and affiliated

transactions committee, audit and compliance committee, audit and public policy committee. The following committees are defined as compensation

committee: compensation committee, executive salary committee, compensation and stock option committee, compensation policy committee,

organization and compensation committee, compensation and leadership committee, compensation, benefits and stock option committee, compensation

and incentive committee, etc.5 The retainer paid for special committees formed for special reasons is excluded. We also note that results are not sensitive to instead assuming that

directors choose the maximum amount of equity pay permitted.6 We note that the use of the ExecuComp and IRRC databases results in a sample of firms that are, on average, larger in size (market value and sales)

than the average Compustat firm. We note that our sample firms, with a mean (median) market value of $7,600 (1,313) million and a mean (median) sales

of $5,000 (1,156) million, are considerably larger than the CRSP/Compustat firms, whose mean (median) market value and sales are $2,475 (98) million

and $1,923 (99) million, respectively. The use of this sample, however, allows for comparability with the large number of related studies that draw from

the ExecuComp and IRRC databases. Nonetheless, our results must be interpreted in the context of our relatively large sample firms.7 We exclude meeting fees from the compensation variables to avoid a mechanical relation with one of our key control variables, the number of audit

committee meetings. We discuss this and other issues relating to the audit meetings variable in Section 4.

ARTICLE IN PRESS


As discussed earlier, our main hypothesis is that audit committee compensation is positively related to the demand formonitoring of the firm’s financial reporting process. One way to examine this hypothesis would be to identify factors thataffect the demand for monitoring and regress compensation on those factors. While some of these factors areobservable—for example, size and industry—others, such as the risk of financial misstatement and the complexity of thefirm’s organizational structure, are more difficult for researchers to capture. As a result, a regression of audit committeecompensation on the observable firm-specific factors that determine the demand for monitoring is likely to omit somerelevant factors.

To incorporate unobservable factors that influence the demand for monitoring, we consider the relation between auditcommittee compensation and the audit fees paid to auditors. We argue that both are likely to be positively related tofactors that increase the demand for monitoring. For example, our earlier arguments suggest that firms with high risk ofmisstatement or complex organizational structures will likely ask for more effort from audit committee members, and thuspay higher compensation. Similarly, we note that the extent of work and billings by the audit firm is expected to be linkedwith many firm-specific forces that give rise to the demand for monitoring of the financial reporting process, including thecomplexity of the business and the firm’s organizational structure, the strength of internal control systems, financialreporting quality and litigation risk.8 These factors influence the overall transparency and riskiness of the financialreporting process, which impacts the scope and complexity of the annual audit activities, and the risk premium required byauditors. A regression of audit committee compensation on audit fees is therefore predicted to yield a positive coefficient,as audit fees proxy for firm-specific variation in the demand for monitoring of the financial reporting process.

Such a regression may suffer from simultaneity bias if auditor effort and audit committee effort are substitutes orcomplements in the firm’s objective function. For example, auditor effort might make audit committee effort morevaluable; in this case, the firm’s optimal choice of audit committee effort would be an increasing function of auditor effort.This would suggest a direct causal link between audit committee compensation and audit fees. Econometrically, thisscenario would require modeling audit committee pay and audit fees as a system of simultaneous equations. AsWooldridge (2002) points out, however, a system of equations need not be simultaneous, even when two choice variablesare regressed on similar economic factors.9 It is possible that ordinary least squares (OLS) is appropriate when both auditfees and audit committee compensation are positively related to factors that increase the demand for monitoring.Accordingly, as we discuss in more detail later, we perform a Hausman test to assess whether the endogeneity of audit feesin our setting poses a threat to the OLS estimates.

In determining annual audit fees from corporate proxy statements, we include only fees directly related to auditingservices (including internal control analyses) and exclude fees related to supplemental and tax work. We scale audit fees bythe square root of total asset value of the firm based on the findings of Simunic (1980) and Kinney et al. (2004) that thesquare root function best captures the relation between audit fees and assets. Size deflation allows us to ensure that ourfindings are not driven solely by the well established association between compensation and size (see, for example, Rosen,1992; Murphy, 1985).10

In selecting total assets as a size deflator, we are careful not to purge our demand for monitoring proxy of its keyattributes. Audit fees can be viewed as the product of quantity (Q) and price (P) of audit labor hours, where Q captures thenumber of audit hours in the engagement and P reflects the audit firm’s price per unit of auditing. The price component isfurther impacted by both the mix of labor units (i.e., class of labor) and billing rates for each class of labor chosen by theaudit firm. We can thus expand the basic audit fees equation to be

Total Audit Fees ¼Xn

i¼1

Hours of labor units for class of labor i|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}Quantity

� Billing rate per hour for class of labor i|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}Price

where i ¼ audit class of labor (e.g., staff, senior, partner).We argue that both the quantity and price components of audit fees chosen by the audit firm are impacted by firm size,

complexity and risk. The empirical findings of O’Keefe et al. (1994) largely support these relations. Thus, deflating audit feesby firm size preserves the key elements of interest—firm complexity and risk—through both the price and quantity in ourproxy for the demand for monitoring of the financial reporting process.

Our sample period includes the years 2000–2004. This sample period allows us to develop an additional proxy for thedemand for monitoring of financial reporting arising from an important external regulatory event—the passage of theSarbanes-Oxley Act. While audit fees in the post-SOX period can capture some aspects of increased demand for financial

8 See, for example, prior work by O’Keefe et al. (1994), Bell et al. (2001) and Danielsen et al. (2007).9 See chapter 9 of Wooldridge (2002) for a discussion of structural models and systems of simultaneous equations. Wooldridge (2002) cites Biddle

and Hamermesh (1990) as an example of a study that considers the responses of two choice variables to the same factors. Biddle and Hamermesh (1990)

conduct an OLS regression of one choice variable (minutes per week spent sleeping) on another (minutes per week working). They note that their results

suggest that the time spent sleeping is changed by the indirect effects of the economic factors on decisions about work time.10 Because size may also capture the complexity of the firm’s financial reporting system, one might argue that the undeflated audit fee variable is an

appropriate demand proxy. While our primary model includes a size variable (i.e., market value) in the regressions, we address this issue by also

conducting our analyses using the logarithm of undeflated audit fees as the demand proxy. All regression results using the undeflated audit fees measure

are qualitatively similar to those in Tables 5–7.

ARTICLE IN PRESS


monitoring due to SOX—for example, requirements relating to internal control disclosures and testing—it is possible thatfactors relating to the enhanced monitoring environment, greater board accountability, and increased credentialrequirements directed at audit committee members are not fully reflected in the audit fee variable. Following SOX andthe events leading to its passage, audit committee members are also likely to spend more time communicating withoutside auditors, as well as additional time and effort communicating with and monitoring internal auditors and financialexecutives of the company. SOX was signed into law on July 31, 2002; thus we define our PostSOX indicator variable to equalzero for years 2000 and 2001, and to equal one for years 2002 through 2004.11

Audit committee expertise in monitoring of the financial reporting process. We develop a measure, AC_Expertise, thatcaptures the extent of audit committee expertise based on information about the audit committee chair’s backgrounddescribed in corporate proxy statements. In classifying the background of audit committee chairs, we are guided in part byboth the proposed and final definitions of financial expertise of audit committee members under the Sarbanes-Oxley Act.Initial SOX proposals defined a financial expert as a director with prior or current experience as a public accountant,auditor, principal or chief financial officer, controller or chief accounting officer. Final regulations expanded this definitionto include a broader set of experience, including supervision or oversight of employees with financial reporting duties. Thisbroader category would include corporate presidents and chief executive officers. We also consider the findings of Defondet al. (2005), which highlight the distinction between accounting and non-accounting financial experts. Using a sample of702 announcements of audit committee director appointments over the period 1993–2002, Defond et al. (2005) documenta significant positive market reaction to announcements of accounting financial expert appointments, while the marketreaction to similar announcements of non-accounting financial experts is not significant.

Using biographies included in corporate proxy statements, we categorize the employment experience and backgroundof audit committee chairs as follows:

�

wh

pos

The

Non-financial director (Non_Financial_Expert)—Audit committee chairs with no direct financial training or experience.This category includes directors with responsibility to oversee the financial reporting process but who do not havedirect experience in a financial position themselves. AC_Expertise takes a value of 1 for these directors.
� Finance financial expert (Expert_Finance)—Audit committee chairs with financial training and experience, including
chief financial officers, vice-presidents of finance, and finance professors. AC_Expertise takes a value of 2 for thesedirectors.
� General accounting financial expert (Expert_Accounting_General)—Audit committee chairs with accounting training and
experience, including certified public accountants (CPAs), chief accounting officers, controllers, accounting professorsand those who served on accounting standard or other oversight boards. Accounting experts with employmentexperience with a Big 5(4) accounting firm are excluded; these directors are separately considered in the next category.AC_Expertise takes a value of 3 for these directors.
� Accounting expert with Big 5(4) employment experience (Expert_Accounting_Big 5)—Audit committee chairs with
employment experience at a Big 5(4) accounting firm.12 This category is designed to capture a higher quality accountingexpert, as Big 5(4) accountants are more likely to have direct experience with the financial reporting process of publicfirms. AC_Expertise takes a value of 4 for these directors.

Our classification highlights firms that have explicitly chosen an audit committee chair with financial reportingexpertise. We note that our classifications of both finance and accounting experts are narrower than those of both theproposed and final SOX regulations. We have attempted to identify firms that have chosen directors with greater ability tospecifically monitor the financial reporting process vs. those with more broad finance skills and training.

3.2.2. Regression specifications

Our first set of analyses explores factors associated with the level of audit committee compensation. Since our tests usepanel data, we include firm fixed effects in our regressions and estimate robust standard errors to mitigate potentialproblems with the panel data. We include AuditFee and PostSOX to capture the demand for monitoring of the financialreporting process.

Similar to Linck et al. (2009), we select control variables mainly based on agency theory and bargaining theory. Whileagency problems are generally defined as the conflict between shareholders and managers, agency problems could alsoexist between shareholders and outside directors. Agency theory suggests that compensation should provide incentives toalign the interests of the parties. Thus, it predicts relations between compensation and effort, firm performance, jobcomplexity, and leverage.

We use the number of times the audit committee of the board of directors met during the year (AuditMeet) disclosed inthe firm’s proxy statement to reflect the effort and workload of audit committees. One might argue that the number of

11 The use of annual data determines how we partition the sample period. We find qualitatively similar results for all of our regression analyses (1)

en we exclude year 2002 observations in the empirical tests to address the concern that the year 2002 is a transition year, and (2) when we define the

t-SOX period as years 2003 and 2004.12 Big 5(4) refers to the five (four) largest global professional accounting firms that perform a substantial majority of audits of publicly traded firms.

number of global firms was reduced from five to four during our sample period due to the collapse of Arthur Andersen in 2002 following Enron.

ARTICLE IN PRESS


audit committee meetings is also an indicator of the overall demand for monitoring of the firm’s financial reportingprocess. Firms with greater demand for monitoring of the financial reporting process are those that are willing to incur theincreased costs of a more active audit committee. Menon and Williams (1994) document that the number of annual auditcommittee meetings is positively associated with firm size, suggesting a connection between monitoring complexity andthe need for greater audit committee involvement. The number of audit committee meetings is also often used as a proxyfor audit committee diligence.13

We use industry-median-adjusted stock returns and industry-median-adjusted ROA (return on assets) to control for thepotential impact of firm performance on compensation. We control for firm size (measured as the logarithm of marketvalue of equity), research and development costs (measured as R&D expenditures deflated by total assets), and market-to-book ratio (measured as book value of assets minus book value of equity plus market value of equity divided by book valueof assets). We define leverage as short term and long term debt scaled by total assets, and we capture cash constraints andthe tax advantages of options by including a zero dividend dummy and a net operating loss carryforward dummy (Yermack,2004).14 Finally, in regressions without firm fixed effects, we include a control for an increased demand for monitoring thatis driven by the strength of the firm’s internal controls over the financial reporting process. MatlWeak takes a value of 1 forall sample firm-years if the firm disclosed the existence of a material weakness in internal control under SOX.15

Alternatively, bargaining theory suggests that director compensation is determined by the negotiation process betweenthe board and the CEO (see Hermalin and Weisbach, 1998 for theory, and see Ryan and Wiggins, 2004 for empiricalevidence). Following that literature, we use the percentage of independent directors on the board and the ownership ofoutside directors to measure the board’s negotiation power. To capture the CEO’s bargaining power, we use CEO tenure,ownership, and a dummy variable equal to one if the CEO chairs the board.

Our second set of regression analyses examines audit committee compensation relative to compensation committeemembers. One potential challenge in conducting multivariate analyses is the possibility of correlated omitted variables.Because some data are unobservable or unavailable, we cannot include all firm-specific variables in the regression model. Itis possible that these unobservable factors might contaminate the statistical relationship. To address this concern, we rundifference regressions that analyze the compensation variables for audit committees relative to compensation committees.Thus, we are able to use outside directors serving on compensation committees as a control group to control forunobservable firm-specific factors that may be associated with board member compensation across firms.

It is possible that other factors occurring around the time of the SOX legislation, including general macroeconomicconditions, pressure from more active institutional investors, and changes in the public’s view of corporate governance,may have contributed to the passage of SOX. However, given that the primary focus of SOX is on firms’ financial reportingprocesses, directors on the audit committee are likely to be affected more by SOX. As such, another appealing feature of thedifference specification is the use of outside directors serving on the compensation committee as a control group to controlfor these factors.16

3.3. Descriptive statistics

Table 1 presents summary statistics relating to the compensation of audit committee members.17 Panel A presents thelevel of compensation and the key components of the compensation packages for outside directors serving as chair of theaudit committee and other members of the audit committee. We observe that the equity component (equity retainer,options and equity grants) of total compensation is, on average, greater than the cash component (cash retainer andmeeting fees). We also note that the average and median compensation (in total and by component) is slightly greater forthe audit committee chair than for other members of the audit committee. For example, the average (median) totalcompensation is $140,912 ($97,000) for the audit committee chair and $135,451 ($91,459) for other members.

Panel B shows the time series variation in compensation and its components from 2000 to 2004 for the chair of theaudit committee. The average (median) total compensation for the audit committee chair is $138,474 ($104,651) in the

13 See Vera-Munoz (2005) and DeZoort et al. (2002) for insightful discussions and reviews of the related literature on audit committee effectiveness,

including diligence.14 The zero dividend dummy is equal to one if a firm pays a dividend (Compustat data26). The net operating loss carryforward dummy is equal to one

if a firm has a net operating loss carryforward (Compustat data52).15 We thank Doyle, Ge and McVay for sharing their dataset of firms disclosing material internal control weakness under Sections 302 and 404 of SOX.

Doyle et al. (2007) use SEC electronic filings of annual 10-Ks and subscription data from Compliance Week to identify material weakness disclosures by

firms over the period August 2002 through the end of our sample period.16 While SOX focuses mainly on the financial reporting process, it is possible that other SOX requirements (e.g., independence of compensation

committee members) could impact the level of compensation committee pay in some firms. We would expect any such effect to be less substantial than

the effect on audit committees and not linked with the demand for monitoring of the financial reporting process. However, if SOX requirements do

increase compensation committee pay, any pay differential for audit committee members would be reduced, making it harder for us to document a link

between the audit committee pay differential and the demand for monitoring. We note that the recent expanded Compensation Discussion and Analysis

disclosures in corporate proxy statements required by the SEC, which would likely have a greater impact on the risk and required effort of compensation

committee members, were implemented subsequent to our 2000–2004 sample period (specifically, these disclosures were required in proxy statements

filed after December 15, 2006).17 All compensation, audit fees, and price-level variables are in real (vs. nominal) terms, adjusted using the GDP price deflator index with 2004 as the

base year.

ARTICLE IN PRESS

Table 1Summary statistics: audit committee compensation.

Panel A: Compensation statistics for audit committees

N Q 1 Mean Median Q 3 Std.

Audit committee chair

Total compensation (incl. meeting fees) 5,465 57,176.40 140,912.20 97,000.00 156,488.13 290,335.08

Total compensation (no meeting fees) 5,465 51,000.00 134,544.21 89,360.02 149,774.98 290,384.75

Cash retainer 5,465 16,500.00 29,449.76 26,666.60 39,762.80 19,847.70

Equity retainer 5,465 0.00 6,346.22 0.00 0.00 17,355.51

Options 5,465 0.00 90,104.43 37,272.22 97,451.42 291,276.91

Equity grants 5,465 0.00 8,643.79 0.00 0.00 23,309.42

Meeting fees 5,465 0.00 6,367.99 4,746.72 9,295.20 7,221.82

Audit committee member

Total compensation (incl. meeting fees) 5,465 54,000.00 135,451.46 91,459.27 150,241.28 289,279.54

Total compensation (no meeting fees) 5,465 48,028.55 129,728.28 84,377.57 144,084.60 289,393.44

Cash retainer 5,465 15,492.00 26,249.10 24,787.20 35,000.00 17,385.20

Equity retainer 5,465 0.00 5,845.21 0.00 0.00 16,216.99

Options 5,465 0.00 88,990.18 36,901.63 96,700.35 291,236.71

Equity grants 5,465 0.00 8,643.79 0.00 0.00 23,309.42

Meeting fees 5,465 0.00 5,723.18 4,315.20 8,454.40 6,229.34

Panel B: Time series compensation statistics for audit committee chair

2000 2001 2002 2003 2004 Pre-SOX Post-SOX Pre-/Post-SOX difference

Total compensation Mean 150,278.61 141,236.05 109,560.99 150,629.69 156,107.93 145,307.52 138,473.83 �6,833.69

(incl. meeting fees) Median 73,615.09 88,187.52 79,260.00 111,489.86 124,337.60 82,515.48 104,650.52 22,135.04a


(no meeting fees) Median 70,175.11 81,848.57 72,070.42 102,887.18 111,804.13 77,228.09 95,933.63 18,705.54a

Cash retainer Mean 24,090.15 24,186.47 26,497.22 32,754.78 38,170.97 24,143.10 32,393.72 8,250.61a

Median 22,020.00 21,576.00 24,306.40 30,596.70 35,000.00 22,020.00 30,000.00 7,980.00a

Equity retainer Mean 5,464.07 5,776.79 5,851.98 6,572.32 7,839.45 5,635.98 6,740.24 1,104.25b

Median 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Options Mean 111,511.84 101,957.36 65,970.46 93,170.06 84,578.62 106,259.32 81,142.25 �25,117.07b

Median 27,711.35 38,639.65 29,971.20 47,743.77 40,944.41 33,626.40 38,474.90 4,848.50

Equity grants Mean 5,523.75 5,365.89 5,399.17 9,897.05 16,203.06 5,436.97 10,422.83 4,985.86a

Median 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0:00a

Meeting fees Mean 3,688.81 3,949.53 5,842.16 8,235.49 9,315.83 3,832.14 7,774.80 3,942.65a

Median 3,303.00 3,236.40 4,755.60 7,229.60 8,000.00 3,236.40 6,340.80 3,104.40a

N 878 1,072 1,193 1,178 1,144 1,950 3,515

Total compensation includes cash retainer, equity retainer, options, equity grants. Meeting fees are included in the first total compensation measure but

not the second. Options include stock options and SARs. Equity grants include common stock, restricted stock, deferred stock units and phantom stock

units. Meeting fees include fees for attending both board and committee meetings; special committee fees are excluded. All compensation data are price-

level adjusted using the GDP price deflator index with 2004 as the base year. Pre-SOX = years 2001 and 2001. Post-SOX = years 2002–2004. t-tests used to

test differences in the pre- and post-SOX mean; Wilcoxon two-sample tests used to test differences in the pre- and post-SOX median. a, b, and c denote

significance of coefficients at the 1%, 5%, and 10% levels, respectively.


post-SOX period, compared to $145,308 ($82,515) in the pre-SOX period. Despite the decline in average total compensation,the components of total compensation generally increase over time. We note significant increases (at better than the 1%level) from the pre-SOX to the post-SOX period in the levels of each of the components of total compensation, with theexception of the value of options, which experienced a decline in value from the pre-SOX to the post-SOX period. Thisreduction in the value of options granted to committee members is due primarily to the significant drop in the value ofstock options granted in 2002. The component with the largest percentage increase is the cash retainer: the average totalcash retainer is $32,394 in the post-SOX period, compared to $24,143 in the pre-SOX period.

Table 2 provides summary compensation statistics for compensation committees. Panel A presents the level of totalcompensation excluding meeting fees and the level of total cash retainer for outside directors serving as chair of thecompensation committee and other members of the compensation committee. When we compare these statistics forcompensation committees with those for audit committees in Table 1, we note that outside directors on audit committeesare in general paid more than outside directors on compensation committees. For example, mean (median) totalcompensation for the chair of the compensation committee is $133,311 ($87,946), compared to $140,912 ($97,000) for theaudit committee chair.

Panel B shows the time series variation in compensation variables from 2000 to 2004 for the chair of the compensationcommittee and the difference between audit and compensation committee chairs. We define the difference variables ascompensation variables for audit committees less compensation variables for compensation committees. When wecompare the inflation-adjusted levels of compensation for audit committees and compensation committees, we note thatwhile the level of compensation for compensation committees is comparable to that for audit committees in the pre-SOX

ARTICLE IN PRESS

Table 2Summary statistics: compensation committee compensation.

Panel A: Compensation statistics for compensation committees


Compensation committee chair

Total compensation 5,465 50,090.80 133,311.32 87,945.73 147,640.00 290,272.40

Cash retainer 5,465 16,182.00 28,566.38 25,891.20 38,500.00 19,344.51

Compensation committee member

Total compensation 5,465 47,931.43 12,9,542.01 84,401.74 14,3,921.16 28,9,441.09

Cash retainer 5,465 15,492.00 26,169.62 24,772.50 35,000.00 17,519.67

Panel B: Time series compensation statistics


Compensation committee chair


Median 70,175.11 81,596.33 71,808.54 102,127.34 109,984.97 77,046.06 94,723.92 17,677.86

Cash retainer Mean 24,314.02 24,192.78 26,202.07 31,363.21 35,513.94 24,247.37 30,962.41 6,715.04a

Median 22,020.00 21,576.00 23,778.00 28,402.00 33,166.65 22,020.00 28,000.00 5,980.00a

Difference between audit and compensation committee chairs

Total compensation Mean �280.71 253.70 469.04 1,943.56 3,376.85 13.08 1,909.59 1,896.51a

Median 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0:00a

Cash retainer Mean �223.87 �6.30 295.15 1,391.58 2,657.03 �104.27 1,431.31 1,535.57a

Median 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0:00a

N 878 1,072 1,193 1,178 1,144 1,950 3,515

Total compensation includes cash retainer, equity retainer, options, and equity grants, but excludes meeting fees. Options include stock options and SARs.

Equity grants include common stock, restricted stock, deferred stock units and phantom stock units. All compensation data are price-level adjusted using

the GDP price deflator index with 1,998 as the base year. Pre-SOX = years 2001 and 2001. Post-SOX = years 2002–2004. t-tests used to test differences in

the pre- and post-SOX mean; Wilcoxon two-sample tests used to test differences in the pre- and post-SOX median. a, b, and c denote significance of

coefficients at the 1%, 5%, and 10% levels, respectively.


period, the level of compensation for audit committees exceeds that of compensation committees in the post-SOX period.The mean (median) difference in total compensation excluding meeting fees between the two committees is $13 ($0) in thepre-SOX period, and $1,910 ($0) in the post-SOX period. The increase in the difference from the pre-SOX period to the post-SOX period is statistically significant at better than the 1% level. We note that while the $1,910 premium is small inmagnitude, the presence of within-firm director pay heterogeneity in the post-SOX period appears to represent a shift indirector pay patterns. We also note that the pay premium in those firms that do deviate from the historically prevalentuniform pay to directors is larger than the mean of $1,910, given that the median of the pay premium is zero.

Table 3 displays the summary statistics (Panel A) and the time-series trend (Panel B) of audit fees, board/committeemeetings, and audit committee chair expertise. Panel A notes that the average (median) number of audit committeemeetings is 6.6 (6.0) per year, which exceeds that of the compensation committee (4.3 (4.0)). Consistent with Linck et al.(2009), we find overall increases in the number of annual committee meetings over the sample period (2000–2004). PanelB shows that the average number of total committee meetings is 16.88 in the post-SOX period, significantly higher than the11.38 meetings in the pre-SOX period. The number of audit committee meetings shows a similar impressive increase: theaverage number of audit committee meetings is 7.71 in the post-SOX period, compared to 4.52 in the pre-SOX period.18

These differences are statistically significant at better than the 1% level. While we also observe a significant increase in thenumber of compensation committee meetings from the pre-SOX to the post-SOX period, the magnitude of the increase ismuch smaller (average of 3.8 meetings pre-SOX vs. 4.6 in the post-SOX period) than that of the audit committee. Anothernotable trend is in audit fees: Audit fees have increased significantly from an average of $1.164 million in the pre-SOXperiod to $2.289 million in the post-SOX period, and the ratio of audit fees to the square root of total assets has increasedsignificantly from 0.61 in the pre-SOX period to 1.17 in the post-SOX period (see Ghosh and Pawlewicz, 2008; Griffin andLont, 2008 for similar findings on changes in audit fees around SOX).

Table 3 also reports significant increases in the number of audit committee chairs that have both finance and accountingexpertise, with the largest increase in Expert_Accounting_Big5, the category of audit committee chairs with employmentexperience at a Big 5(4) accounting firm. While the vast majority of sample firms continue to have audit committee chairswithout financial expertise, the percentage of firms with non-financial chairs has declined significantly from 83% in thepre-SOX period to 73% in the post-SOX period. The percentage of firms with finance financial experts increased from 11% inthe pre-SOX period to 15% in the post-SOX period. The percentage of firms with accounting experts (both general and Big5(4)) doubled from 6% to 12% from the pre-SOX period to the post-SOX period with the largest increase in the Big 5 expert

18 The increase in the number of audit committee meetings per year is likely due, in part, to the new required communication between auditors and

audit committees.

ARTICLE IN PRESS

Table 3Summary statistics: firms’ audit fees and board committee meetings.

Panel A: Summary statistics


Audit fees and board meetings

Audit fees (undeflated) 5,465 377.58 1,887.5 779.64 1,859.04 3,830.97

AuditFee 5,465 0.43 0.97 0.72 1.22 0.89

Total board meetings 5,465 9 14.92 14 19 7.98

Audit committee meetings 5,465 4 6.57 6 8 3.38

Compensation committee meetings 5,465 3 4.31 4 6 2.55

Audit committee chair expertise

Non_Financial_Expert 3,473 1 0.76 1 1 0.43

Expert_Finance 3,473 0 0.14 0 0 0.35

Expert_Accounting_General 3,473 0 0.04 0 0 0.21

Expert_Accounting_Big5 3,473 0 0.06 0 0 0.23

AC_Expertise 3,473 1 1.40 1 1 0.81

Panel B: Time series statistics


Audit fees and board meetings

Audit fees (undeflated) Mean 1,121.74 1,198.81 1,641.66 1,968.5 3,293.51 1,164.11 2,288.81 1,124.70a

Median 509.21 530.68 685.93 835.73 1,645.1 523.79 978.06 454.27a

AuditFee Mean 0.59 0.63 0.84 0.97 1.71 0.61 1.17 0.56a

Median 0.47 0.52 0.69 0.79 1.47 0.49 0.9 0.41a

Total board meetings Mean 11.06 11.64 14.18 17.2 19.37 11.38 16.88 5.51a

Median 10 10 13 16 18 10 16 6.00a

Audit committee meetings Mean 4.24 4.75 6.52 7.83 8.84 4.52 7.71 3.20a

Median 4 4 6 8 8 4 7 3.00a

Compensation committee meetings Mean 3.87 3.75 4.04 4.56 5.2 3.8 4.59 0.79a

Median 4 4 4 4 5 4 4 0.00a

N 878 1,072 1,193 1,178 1,144 1,950 3,515

Audit committee chair expertise

Non_Financial_Expert Mean 0.85 0.81 0.8 0.75 0.66 0.83 0.73 �0.10a

Median 1 1 1 1 1 1 1 0.00a

Expert_Finance Mean 0.1 0.12 0.13 0.14 0.18 0.11 0.15 0.04a

Median 0 0 0 0 0 0 0 0.00a

Expert_Accounting_General Mean 0.03 0.03 0.03 0.05 0.07 0.03 0.05 0.02a

Median 0 0 0 0 0 0 0 0.00a

Expert_Accounting_Big5 Mean 0.03 0.04 0.04 0.06 0.09 0.03 0.07 0.03a

Median 0 0 0 0 0 0 0 0.00a

AC_Expertise Mean 1.23 1.29 1.32 1.42 1.6 1.27 1.45 0.18a

Median 1 1 1 1 1 1 1 0.00a

N 385 657 746 826 859 1,042 2,431

Audit fees are in thousands of dollars and include annual fees directly related to auditing services. Audit fees do not include audit-related fees, tax fees,

and all other fees. AuditFee ¼ audit fees/square root of total assets. Audit fees and total assets are price-level adjusted using the GDP price deflator index

with 2004 as the base year. Non_Financial_Expert ¼ 1 if the audit committee chair has no direct financial training or experience. Expert_Finance ¼ 1 if the

audit committee chair has financial training and experience. Expert_Accounting_General ¼ 1 if the audit committee chair has accounting training and

experience, but excludes those with Big 5(4) accounting firm employment experience. Expert_Accounting_Big5 ¼ 1 if the audit committee chair has

employment experience at a Big 5(4) accounting firm. AC_Expertise ¼ 1 if Non_Financial_Expert ¼ 1; 2 if Expert_Finance ¼ 1; 3 if

Expert_Accounting_General ¼ 1; and 4 if Expert_Accounting_Big5 ¼ 1. Pre� SOX ¼ years 2001 and 2001. Post� SOX ¼ years 2002–2004. t-Tests used to

test differences in the pre- and post-SOX mean; Wilcoxon two-sample tests used to test differences in the pre- and post-SOX median. a, b, and c denote

significance of coefficients at the 1%, 5%, and 10% levels, respectively.


category (increased from 3% to 7%). The increases in the percentage of experts in each category result in the aggregateexpertise variable, AC_Expertise, increasing from 1.27 in the pre-SOX period to 1.45 in the post-SOX period. These increasesare statistically significant at better than the 1% level, suggesting that more boards have chosen to increase the level offinancial expertise of their audit committees by appointing directors with finance and, especially, accounting experienceafter the passage of SOX.

Table 4 lists summary statistics for the control variables capturing the determinants of directors’ compensationsuggested by contracting theory and bargaining theory. With regard to the determinants related to contracting theory, wefirst note that our sample firms perform better than their industry peers, with an average industry-adjusted accountingreturn of 6% and industry-adjusted stock return of 17%.19 On average, our sample firms spend 4% of their total assets on

19 Industry adjustments are computed using CRSP/Compustat firms as a comparison group, defining industry based on two-digit industry codes.

ARTICLE IN PRESS

Table 4Summary statistics: control variables.


ROA_Ind 5,465 0.00 0.06 0.05 0.11 0.19

Return_Ind 5,465 �0.15 0.17 0.07 0.34 0.60

RD 5,465 0.00 0.04 0.00 0.05 0.07

MV 5,465 527.00 7,600.34 1,312.73 4,153.60 26,876.59

Leverage 5,465 0.01 0.19 0.17 0.29 0.18

DivDum 5,465 0.00 0.46 0.00 1.00 0.50

NOLDum 5,465 0.00 0.36 0.00 1.00 0.48

M=B 5,465 1.21 2.06 1.59 2.32 1.59

MatlWeak 5,465 0.00 0.16 0.00 0.00 0.37

IndDirOwn% 3,295 0.19% 3.80% 6.84% 2.39% 9.57

CEOTen 3,295 3.00 7.59 5.00 10.00 7.03

CEOOwn% 3,295 0.09% 1.93% 0.28% 1.00% 5.50

CEO=Chair 3,295 0.00 0.66 1.00 1.00 0.47

BdInd% 3,295 57.14% 68.18% 70.00% 80.00% 15.65

ROA_Ind ¼ industry-median-adjusted return on assets. Return_Ind ¼ industry-median-adjusted stock return. RD ¼ research and development

expenditures/total assets. MV ¼ market value ($millions, price-level adjusted using the GDP price deflator index with 2004 as the base year).

Leverage ¼ ðshort term and long term debt)/total assets. DivDum ¼ 1 if firm pays a dividend. NOLDum ¼ 1 if firm has a net operating loss carry-

forward. M=B ¼ market value/book value. MatlWeak ¼ 1 if firm disclosed material weaknesses in internal controls in response to SOX Sections 302

and 404. IndDirOwn % ¼ percentage ownership of independent directors. CEOTen ¼ CEO tenure. CEOOwn % ¼ percentage ownership of CEO. CEO=Chair ¼ 1 if

CEO is chairman of board. BdInd % ¼ percentage of independent directors on board.


research and development. The mean (median) market value is $7.60 ($1.31) billion, and the mean (median) market-to-book is 2.06 (1.59). In addition, the average firm leverage is 0.19, 46% of the sample firms issue dividends, and 36% of thesample firms have a net operating loss carryforward. Finally, 16% of our sample firms disclosed material weaknesses ininternal controls after the passage of SOX.

Turning to the variables suggested by bargaining theory literature, we find that the mean percentage of independent directorson the board is 68%, and these outside directors hold on average 3.8% of the equity of the firm. Table 4 also notes that the averagetenure of a CEO in the sample is 8 years, 66% of CEOs also chair the board, and these CEOs’ average ownership is 1.93%.

4. Empirical results

4.1. Hausman test for endogeneity

As discussed earlier, we include audit fees in the audit committee compensation regression as a proxy for the demandfor monitoring of the financial reporting process. Audit fees and audit committee compensation are predicted to respond tothe same set of economic factors. We note, however, that OLS is inappropriate if audit fees are endogenous in the auditcommittee compensation regression. In this case, audit fees would be correlated with the residual from the regression,resulting in inconsistent estimates of the regression coefficients. To address this possibility, we run a Hausman test todetermine whether audit fees are endogenous in our compensation regressions.

The Hausman test provides a formal test for the endogeneity of a regressor by comparing the OLS estimate with a twostage least squares (2SLS) estimate. As Wooldridge (2002) notes, if the potentially endogenous variable is uncorrelated withthe residual from the regression, then OLS and 2SLS estimates should differ only by sampling error. Consequently, the firststep in running a Hausman test is to identify one or more exogenous instruments that will allow us to run 2SLS. Aninstrumental variable needs to satisfy two requirements: the instrument should be partially correlated (after the otherexogenous variables in the regression have been netted out) with the potentially endogenous variable, and the instrumentshould be uncorrelated with the regression residual (Wooldridge, 2002; Greene, 2000). While the first requirement can betested statistically, the second requirement needs to be satisfied on theoretical grounds.

The endogeneity concern in our setting stems from the possibility that audit fees are correlated with the residual fromthe audit committee compensation regression. In considering potential instruments, we note that many firm-levelvariables are correlated with audit fees, but they are also likely to be correlated with audit committee compensation. Forexample, we would expect a company with M&A or restructuring transactions to pay both higher audit fees and higheraudit committee compensation as a result of these transactions. Consequently, we will have to go beyond the traditionalaudit fee model to identify an instrument.

As our instrument, we use an indicator variable for whether the firm has a December fiscal year end.20 As prior work onaudit fees has found (see, for example, Francis, 1984; Craswell and Francis, 1999), audit fees are lower, all else equal, for

20 Variation in fiscal year ends has been used as an identification technique by other researchers, most notably Oyer (1995). He uses variation in fiscal

year ends to identify incentive effects of nonlinear incentive contracts.

ARTICLE IN PRESS


firms with ‘‘off-peak’’ fiscal year ends. Francis (1984) hypothesizes that the higher peak fees relate to the auditor productionfunction; higher audit fees for these firms appear to reflect the increased auditor workload during the busy season. Amajority of U.S. firms (63% in our sample) have December fiscal year ends, so we use December fiscal year end to captureauditors’ peak period. We do not expect a corresponding response to calendar year-end busy season in audit committeecompensation. Unlike auditors, directors tend to serve on relatively few firms’ boards (for example, Yang and Krishnan,2005, find that audit committee members average 1.3 outside directorships in their sample of randomly chosen Compustatfirms). Thus, while auditors’ busy season is determined by the fiscal year ends of auditees, which tend to be concentrated inDecember, any additional effort required by audit committee members will be related to the firm’s own fiscal year end.

In the first stage of 2SLS, we regress audit fees on all of the explanatory variables from the second stage (i.e., theexplanatory variables other than audit fees in our tabulated regressions) and the December fiscal year end instrument.21

The December fiscal-year indicator is significantly positive with a t-statistic of 10.04 in this regression, indicating that theinstrument is partially correlated with the potentially endogenous variable, as required for a valid instrument.22 The R2 inthe first stage regression is 0.20. In the second stage, we include the predicted value of AuditFee in the compensationregression and compute the 2SLS estimates. We then conduct a Hausman test that compares the OLS and 2SLS estimates ofAuditFee to test the null hypothesis that this variable is exogenous. The Hausman test fails to reject the null hypothesis thataudit fees are exogenous in both the levels and difference regressions and for both total compensation and cash retainer.For example, in the levels (difference) regression of total compensation that controls for audit committee expertise, the teststatistic is 0.113 (0.037), corresponding to a p-value of 0.74 (0.85), which indicates that the audit fee variable is uncorrelatedwith the residual from the audit committee compensation regression.23 Similar results hold for the Hausman tests onAuditFee in our other specifications. These results from Hausman tests indicate that OLS will generate consistent estimatesin the audit committee compensation regression. As such, we use OLS in the analysis that follows (see Wooldridge, 2002,2004).

4.2. Levels regression results

We proceed by estimating an OLS regression for the compensation variables of interest. Columns (1)–(3) ofTable 5 present our first set of regression analyses, which examine the relation between the level of total compensation andproxies for the demand for monitoring of the financial reporting process. Given that there is a mechanical relation betweenmeeting fees and AuditMeet, we use the level of total compensation excluding meeting fees in all regression analyses.24 Wepredict that firms with a higher demand for monitoring of the financial reporting process are likely to pay highercompensation to audit committees. In Column (1), we consider the demand for monitoring of the financial reportingprocess proxies and also include firm level control variables suggested by contracting theory. We find a significantlypositive relation between the level of total compensation and our firm-specific proxy for the demand for monitoring of thefinancial reporting process, AuditFee. We also document a significantly positive relation between the level of totalcompensation for audit committees and the post-SOX indicator variable, PostSOX, which proxies for an external shift in theoverall demand for monitoring of the financial reporting process.

In addition to evaluating the statistical significance of our results, we can use the coefficients from the regression tocalculate economic effects. Using the estimates in Column (1), we compute that the predicted total compensation of theaudit committee chair is $78,523 for a firm in the 10th percentile of AuditFee (0.28), and $83,756 for a firm in the 90thpercentile of AuditFee (1.89), where all other explanatory variables take their median values.25 Again using the estimates inColumn (1) to illustrate the economic effects, the predicted total compensation of the audit committee chair of a medianfirm (i.e., holding all other explanatory variables at their medians) is $71,084 in the pre-SOX period, and $79,907 in thepost-SOX period. While one might argue that these numbers are small in economic magnitude compared to the totalwealth of these directors, Adams and Ferreira (2008) present evidence that directors are more likely to attend boardmeetings when board meeting fees are higher, suggesting that corporate directors appear to perform for even very smallfinancial rewards. As such, our results are consistent with the prediction that firms respond to the increased demand formonitoring of the financial reporting process from both firm-specific and external forces by providing more financialrewards.

We also note a significantly positive coefficient on AuditMeet in Column (1), consistent with the notion that auditcommittees are paid for their time and effort. While AuditMeet is not our main variable of interest, its economic effects on

21 As noted in Larcker and Rusticus (2008) (footnote 8), if we are only interested in one equation in a potentially simultaneous system, we do not need

to estimate both equations with exogenous instruments. The simultaneous equation estimator is identical to an independent application of the 2SLS

procedure to each equation. See also Wooldridge (2002, p. 192).22 Because fiscal year ends may be similar for firms in a given industry, we also run 2SLS with industry fixed effects. The t-statistic on December fiscal

year end is 7.73 when indicators for two-digit industry are included in the first stage.23 We conduct the Hausman test on a levels regression with industry fixed effects rather than firm fixed effects, since the December fiscal year end

indicator would be subsumed by firm fixed effects.24 We obtain qualitatively similar results when we include meeting fees in the level of total compensation as the dependent variable.25 Because we calculate these values with all other explanatory variables at their median values, these numbers appear low in comparison to the

descriptive statistics presented earlier. Similar calculations using the means of other variables produce values more similar to the means of our descriptive

statistics.

ARTICLE IN PRESS

Table 5OLS regressions of audit committee compensation.

Total compensation Cash retainer

(1) (2) (3) (4) (5) (6)

AuditFee 0:040a 0:037a 0:031a 0:072a 0:069a 0:064a

(3.71) (3.15) (2.85) (5.79) (4.16) (4.05)

PostSOX 0:117a 0:121a 0:106a 0:146a 0:131a 0:128a

(7.76) (6.54) (5.61) (13.76) (9.28) (8.88)

AC_Expertise 0:041b 0:037b 0:032b 0:029b

(2.47) (2.19) (2.20) (2.05)

AuditMeet 0:011a 0:014a 0:014a 0:017a 0:015a 0:014a

(3.96) (3.97) (4.19) (7.12) (4.90) (4.61)

ROA_Ind �0.040 �0:247c�0:218c �0.020 �0:264b

�0:235b

(0.61) (1.81) (1.65) (0.52) (2.46) (2.23)

Return_Ind 0:023c 0.026 0.025 �0:025a�0:029b

�0:030b

(1.78) (1.26) (1.16) (3.13) (1.99) (2.09)

R&D �0:534b �0.171 �0.107 �0.028 �0.264 �0.230

(2.43) (0.52) (0.33) (0.23) (1.01) (0.91)

LogMV 0:471a 0:463a 0:455a 0:112a 0:141a 0:149a

(22.52) (15.30) (14.72) (8.23) (5.88) (6.54)

Leverage 0:263a 0:234c0:259b 0.021 0.087 0.093

(2.75) (1.87) (2.01) (0.36) (0.92) (0.96)

DivDum �0.014 0.041 0.032 0:130a 0:185a 0:188a

(0.36) (0.88) (0.66) (4.86) (4.55) (4.15)

NOLDum 0:061b 0.009 �0.007 0:040b 0.028 0.029

(2.42) (0.30) (0.23) (2.52) (1.35) (1.36)

M=B 0:069a 0:079a 0:082a �0.007 �0.001 �0.003

(8.19) (5.08) (5.50) (1.18) (0.08) (0.20)

IndDirOwn% 0.002 �0.001

(1.42) (0.60)

LogCEOTen 0.010 0.010

(0.71) (0.88)

CEOOwn% �0.003 �0.001

(0.74) (0.42)

CEO=Chair �0.043 �0:065a

(1.46) (3.22)

BdInd% 0:341a 0:218a

(3.42) (2.79)

Intercept 7:543a 7:406a 7:247a 9:048a 8:810a 8:643a

(51.52) (33.83) (30.94) (94.46) (50.86) (49.34)

N 5,465 3,473 3,295 5,002 3,263 3,100

Adjusted R2 0.81 0.82 0.83 0.81 0.82 0.82

The dependent variables are the logarithm of total compensation in Columns (1)–(3), and the logarithm of cash retainer in Columns (4)–(6). Total

compensation includes cash retainer, equity retainer, options, and equity grants, but excludes meeting fees. AuditFee ¼ audit fees/square root of total

assets. PostSOX ¼ 1 for years 2002–2004. AuditMeet ¼ number of yearly audit committee meetings. See Table 3 for AC_Expertise definition, and Table 4 for

other control variable definitions. Regressions are estimated with firm-level fixed-effects. t-Statistics calculated using heteroskedasticity robust standard

errors are in parentheses. a, b, and c denote significance of coefficients at the 1%, 5%, and 10% levels, respectively.


audit committee compensation are similar to those of AuditFee. Using the estimates in Column (1), we compute that thepredicted total compensation of the audit committee chair is $77,313 for a firm in the 10th percentile of AuditMeet (3), and$84,425 for a firm in the 90th percentile of AuditMeet (11), where all other explanatory variables take their median values.Recall that the relation between audit committee total compensation and AuditMeet is not simply a mechanical effect, sincemeeting fees paid to committee members are excluded from the computation of total compensation.26 We observed earlierthat AuditMeet could be viewed as an alternative proxy for the demand for monitoring of the financial reporting system.The number of audit committee meetings captures the firm’s response to a need for additional monitoring by the auditcommittee. We note, however, that this conclusion should be interpreted cautiously due to potential concerns about the

26 We also consider the more subtle mechanical effect that some firms may not have explicit meeting fees, but instead may pay a cash retainer based

on the expected number of audit meetings. Approximately 20% of our sample firm-years do not have a separate disclosure of meeting fees, suggesting that

these firms either do not provide additional compensation for meeting attendance or the meeting fees are implicit in the cash retainer. In the latter case,

our measure of total compensation excluding explicit meeting fees may still include implicit meeting fees, confounding our goal of measuring total

compensation without meeting fees. Since firms that do not pay or separately disclose meeting fees do so uniformly for all board members, our primary

difference regressions effectively control for this concern. The issue, however, may be present in our levels regression. As a sensitivity test, we conduct our

levels regression analyses after excluding the observations for which meeting fees are not separately reported in proxy statements. We find that our

results are robust to the exclusion of these firm-years.

ARTICLE IN PRESS


endogeneity of the number of audit meetings. To assess these concerns, we also consider a version of our model of the auditcommittee compensation/demand for monitoring of financial reporting relation that excludes AuditMeet. We find that thecoefficients are similar and the statistical significance of both AuditFee and PostSOX remains unchanged when we excludeAuditMeet from our models in our primary analyses. This suggests that the inclusion of AuditMeet does not have a materialadverse econometric impact on the interpretation and significance of the coefficients of our key demand proxies. Finally,with regard to other control variables, we find that larger firms (high LogMV) and firms with higher investmentopportunities (high M/B) and leverage award their audit committees higher total compensation.27

As discussed before, it seems reasonable to expect both the demand for monitoring of the financial reporting processand the quality of audit committees to be related to compensation arrangements of audit committees. To separatelyexamine these effects, we additionally include a measure of audit committee quality, AC_Expertise, in Column (2). Theresults show that AC_Expertise is significantly and positively related to the level of total compensation. At the same time,the coefficients on the key proxies for the demand for monitoring of financial reporting remain significantly positive. Theseresults support the notion that both audit committee quality and the overall demand for monitoring of the financialreporting process are positively related to the level of audit committee compensation.

We also include control variables suggested by bargaining theory in Column (3). Consistent with prior research (Ryanand Wiggins, 2004), our results indicate that directors serving on outsider dominated boards (high BdInd%) receive morecompensation. This result is consistent with the notion that director compensation varies systematically with barriers toeffective monitoring. We note that the inclusion of these control variables does not alter the results of the proxies for thedemand for monitoring of financial reporting and the quality of audit committees.

We continue our analysis by examining the level of total cash retainer. Columns (4)–(6) report our results. In all threecolumns, we find a positive and statistically significant relation between the level of total cash retainer for auditcommittees and the two proxies for the demand for monitoring of the financial reporting process, AuditFee and PostSOX. Wealso note that the coefficients in Columns (4)–(6) are higher and more significant than those in Columns (1)–(3). Using theestimates in Column (4) to present economic effects, we compute (1) the predicted cash retainer of the audit committeechair is $24,698 for a firm in the 10th percentile of AuditFee (0.29), and $27,761 for a firm in the 90th percentile of AuditFee

(1.91), where all other explanatory variables take their median values; and (2) the predicted cash retainer for the auditcommittee chair of a median firm (i.e., holding all other explanatory variables at their medians) is $22,033 in the pre-SOXperiod, and $25,496 in the post-SOX period.

Similar to the total compensation analyses, we find that AC_Expertise is significantly and positively related to the level ofcash retainer. Regarding other explanatory variables, we continue to observe significantly positive coefficients on AuditMeet

and firm size (LogMV). We note the following differences in control variables from those reported earlier: (1) The coefficienton industry-adjusted market returns is now negative and significant, suggesting that the weak positive correlation betweentotal compensation and market returns is driven by the equity compensation component. (2) The coefficient on M/B is nownegative and loses significance, in contrast to the positive and significant coefficients reported earlier, supporting the viewthat growth firms are more likely to use equity-based compensation relative to cash compensation. (3) The zero dividenddummy and the leverage variable are positively associated with the level of total cash retainer, supporting predictions fromfinancial contracting theory (John and John, 1993) that firms award more equity compensation when they face a scarcity ofcash (DivDum equal to 0), and when the conflict between creditors and shareholders is not severe (low Leverage). (4) TheCEO/Chair dummy is now negative and significant. These results suggest that the demand for monitoring of the financialreporting process also plays an important role in determining the cash retainer component of audit committee pay.

4.3. Difference regression results

In our next set of regressions, we employ a unique feature of our dataset to control for correlated omitted variables.Since we have compensation data for both audit committees and compensation committees, we are able to usecompensation committees to control for unobservable firm-specific variables. Thus, we conduct a difference regression inwhich the dependent variables now capture the differences between audit committee compensation and compensationcommittee compensation. In our base case, we include both the proxies for demand for monitoring of the financialreporting system and control variables suggested by contracting theory (Column (1)). Similar to Table 5, we also estimatethe difference model with AC_Expertise (Column (2)) and other firm-specific controls suggested by bargaining theory(Column (3)).

Columns (1)–(3) of Table 6 present the results when the difference in the level of total compensation excluding meetingfees is the dependent variable. In all models, we find that the difference in the level of total compensation is positivelyrelated to AuditFee, the firm-specific proxy for the demand for monitoring of the financial reporting process, and hasincreased significantly in the post-SOX period compared to the pre-SOX period. These results support the prediction thatfirms with a higher demand for monitoring of the financial reporting process are likely to pay a higher level of totalcompensation to audit committees relative to compensation committees.

27 As a robustness check, we have replaced LogMV with MV, MV-Squared, and MV-Cubed. We obtain qualitatively as those reported in the paper,

suggesting that our results are not affected by the functional form of MV.

ARTICLE IN PRESS

Table 6Difference regressions.


(1) (2) (3) (4) (5) (6)

AuditFee 0:004a0:003b 0:004a 0:009a 0:009a 0:010a

(3.35) (2.34) (2.92) (3.65) (2.89) (3.17)

PostSOX 0:014a 0:013a 0:013a 0:030a 0:027a 0:027a

(5.78) (4.20) (4.01) (8.12) (5.52) (5.15)

AC_Expertise 0:008a 0:008a 0:018a 0:018a

(2.95) (2.73) (4.86) (4.65)

AuditMeet 0:001a0:001b 0:001b 0:003a 0:003a

0:002b

(4.28) (2.54) (2.28) (4.66) (2.74) (2.48)

MatlWeak 0.002 0.002 0.001 �0.006 �0.007 �0.009

(0.68) (0.64) (0.27) (1.20) (0.97) (1.28)

ROA_Ind �0.007 �0.023 �0.018 �0.007 �0.029 �0.019

(1.44) (1.60) (1.25) (0.91) (1.23) (0.78)

Return_Ind �0.002 �0.002 �0.001 �0:004c �0.003 �0.003

(1.47) (0.66) (0.54) (1.76) (0.73) (0.65)

RD �0.024 �0.041 �0.028 �0.034 �0.015 0.016

(1.56) (1.33) (0.92) (1.14) (0.31) (0.32)

LogMV �0.001 0.000 0.001 �0:003b �0.002 0.000

(1.25) (0.06) (1.14) (2.22) (0.94) (0.10)

Leverage �0.003 �0.006 �0.007 �0:017c�0:028c

�0:027c

(0.58) (0.63) (0.71) (1.77) (1.81) (1.75)

DivDum �0:004c �0.005 �0.004 �0:012a�0:013b

�0:013b

(1.66) (1.40) (1.08) (2.94) (2.06) (2.05)

NOLDum �0.001 �0.004 �0.003 0:007c 0.005 0.006

(0.49) (1.43) (1.17) (1.89) (1.11) (1.16)

M=B 0.001 0.003 0.001 0:003c 0.005 0.003

(0.60) (0.92) (0.43) (1.90) (1.52) (0.84)

IndDirOwn% 0:000c 0:001a

(1.75) (2.77)

LogCEOTen 0:003b 0:005c

(2.29) (1.93)

CEOOwn% 0.000 0.001

(1.55) (1.60)

CEO=Chair �0:015a�0:024a

(4.99) (4.79)

BdInd% �0:013c �0.001

(1.87) (0.06)

Intercept 0.001 �0:017c �0.015 0.003 �0:025c�0:032b

(0.16) (1.67) (1.63) (0.34) (1.79) (1.99)

N 5,464 3,472 3,295 4,990 3,258 3,097

Adjusted R2 0.02 0.03 0.04 0.05 0.05 0.06

The dependent variables are differences between audit committee compensation and compensation committee compensation: differences in logarithm of

total compensation in Columns (1)–(3), and differences in logarithm of cash retainer in Columns (4)–(6). Total compensation includes cash retainer, equity

retainer, options, and equity grants, but excludes meeting fees. AuditFee ¼ audit fees/square root of total assets. PostSOX ¼ 1 for years 2002–2004.

AuditMeet ¼ number of yearly audit committee meetings. See Table 3 for AC_Expertise definition, and Table 4 for other control variable definitions. t-

Statistics calculated using heteroskedasticity robust standard errors are in parentheses. a, b, and c denote significance of coefficients at the 1%, 5%, and 10%

levels, respectively.


Similar to the level regressions, we use the coefficients from the regression to calculate economic effects. Using theestimates in Column (1), we compute that the predicted difference in total compensation between audit committee chairsand compensation committee chairs is $1,089 for a firm in the 10th percentile of AuditFee (0.28), and $1,675 for a firm in the90th percentile of AuditFee (1.89), where all other explanatory variables take their median values. Again using the estimatesin Column (1) to illustrate the economic effects, the predicted difference in total compensation between audit committeechairs and compensation committee chairs of a median firm (i.e., holding all other explanatory variables at their medians)is close to 0 in the pre-SOX period, and $1,248 in the post-SOX period.

We also include our measure of audit committee quality, AC_Expertise, in the difference regressions. Becauseit is not clear ex ante how audit committee quality will affect the differences in compensation across committees, includingthis variable allows us to address two plausible scenarios. First, we note that the definition of audit committeequality is driven by recent regulatory calls for financial experts on the audit committee. Although recent regulation hasspecified that compensation committee chairs be independent, there is not a comparable measure of compensationcommittee chair quality. As a result, the quality issue appears to be specific to audit committees, so a higher quality audit

ARTICLE IN PRESS

Table 7Difference regressions: pre- and post-SOX.


Pre-SOX Post-SOX p-Value Pre-SOX Post-SOX p-Value

AuditFee �0.005 0:005a 0.052 �0.003 0:010a 0.038

(0.83) (3.06) (0.48) (3.03)

AC_Expertise 0.003 0:009a 0.045 0.004 0:021a 0.001

(1.65) (2.58) (1.59) (4.50)

AuditMeet 0.001 0:001a 0.375 0.001 0:003a 0.166

(0.69) (2.20) (0.39) (2.48)

MatlWeak 0.003 �0.001 0.575 �0.002 �0.012 0.447

(0.63) (0.13) (0.26) (1.35)

ROA_Ind 0.009 �0.038 0.085 0.016 �0.032 0.258

(0.79) (1.53) (0.67) (0.91)

Return_Ind �0.001 �0.001 0.944 �0.003 �0.002 0.992

(0.50) (0.35) (0.49) (0.35)

RD 0:061a �0.071 0.007 0:108a �0.022 0.112

(3.05) (1.61) (3.40) (0.29)

LogMV 0.004 0.001 0.294 0.003 �0.001 0.352

(1.50) (0.51) (0.95) (0.27)

Leverage �0.005 �0.006 0.968 �0.016 �0.033 0.575

(0.35) (0.54) (0.75) (1.56)

DivDum �0:012b �0.001 0.136 �0.013 �0.013 1.000

(2.08) (0.20) (1.63) (1.57)

NOLDum �0:013b 0.000 0.054 �0:019b 0:016b 0.001

(2.09) (0.16) (2.37) (2.48)

M=B �0:003b 0.004 0.226 �0.004 0.007 0.107

(1.99) (0.70) (1.57) (1.11)

IndDirOwn% 0.000 0.000 0.849 0.000 0:001b 0.317

(0.79) (1.48) (1.22) (2.46)

LogCEOTen �0.002 0:005a 0.014 �0.002 0:008b 0.038

(0.86) (2.72) (0.78) (2.01)

CEOOwn% 0.000 0.000 0.697 0.000 0.001 0.294

(1.11) (1.25) (0.69) (1.58)

CEO=Chair �0:007c�0:018a 0.047 �0.006 �0:031a 0.004

(1.83) (4.58) (1.02) (4.68)

BdInd% �0:019b �0.011 0.549 �0:029b 0.011 0.112

(2.38) (1.13) (2.08) (0.52)

N 969 2,326 894 2,203

Adjusted R2 0.01 0.03 0.01 0.05

Dependent variables are differences between audit committee compensation and compensation committee compensation. Total compensation includes

cash retainer, equity retainer, options, and equity grants, but excludes meeting fees. Pre-SOX includes observations from years 2000–2001. Post-SOX

includes observations from years 2002–2004. p-Value is presented for test of difference between coefficients for pre- and post-SOX regressions (one-sided

test for AuditFee; AC_Expertise, and AuditMeet, and two-sided for other variables). AuditFee ¼ audit fees/square root of total assets. AuditMeet ¼ number of

yearly audit committee meetings. See Table 3 for AC_Expertise definition, and Table 4 for other control variable definitions. t-Statistics calculated using

heteroskedasticity robust standard errors are in parentheses. a, b, and c denote significance of coefficients at the 1%, 5%, and 10% levels, respectively.


committee may be associated with a differential compensation level between the two committees. Second, one mightargue that firms with higher quality audit committees are more likely to also have higher quality compensationcommittees. In this case, the audit committee quality variable will not be related to differential audit committeecompensation.

Column (2) presents the results when we include AC_Expertise in the difference regression model. The results show thatAC_Expertise is significantly and positively related to the difference in the level of total compensation, while the coefficientson the proxies for the demand for monitoring of financial reporting remain significantly positive. These results areconsistent with the expectation that firms with high quality audit committees pay a higher level of total compensation toaudit committees relative to compensation committees. When we include other control variables in Column (3), ourinferences regarding the proxies for the demand for monitoring of financial reporting and audit committee quality areunchanged.

The results for the difference in the level of cash retainer are presented in Columns (4)–(6). We find similar results asthose presented in the total compensation regressions of Columns (1)–(3), while we note that the coefficients on AuditFee,PostSOX, and AC_Expertise in Columns (4)–(6) are of higher magnitudes and significance levels. We also observe that thecoefficients on AuditMeet continue to be positive and significant in both the total compensation (Columns (1)–(3)) and cash

ARTICLE IN PRESS


retainer models (Columns (4)–(6)), consistent with the notion that audit committee pay captures the time and effort ofcommittee members.

The results in Table 6 indicate that audit committee compensation is significantly different from compensationcommittee compensation in the post-SOX period. In our final set of regressions, we investigate whether these post-SOXchanges in the audit committee pay differential can be explained, in part, by AuditFee and AC_Expertise. In other words, dofirms with a greater firm-specific demand for monitoring of financial reporting and with a higher director-specific expertisein financial reporting offer a larger audit committee pay premium in the post-SOX period? We conduct separate regressionsfor the pre-SOX and post-SOX periods and report these results in Table 7. We find positive and significant coefficients onAuditFee and AC_Expertise only in the post-SOX period for total compensation and cash retainer. The pre- and post-SOXcoefficients are significantly different from each other in both regressions.

The results in Table 7 suggest that the significant relations between the audit and compensation committee paypremium and AuditFee, and between the pay premium and AC_Expertise reported in Table 6, are attributable to boardcompensation decisions in the post-SOX period. While the insignificant coefficient on audit fees in the pre-SOX period inthe difference regressions likely reflects the historical lack of within-firm pay variation for directors, these results provideevidence of firms’ increased willingness to offer differential compensation to board members following SOX and the eventsleading up to its passage. Further, this post-SOX pay differential is related to variation in firm-specific demand formonitoring and director-specific expertise in financial reporting.

5. Conclusion

In this paper, we have examined how cross-sectional variation in the demand for monitoring of the financial reportingprocess is associated with compensation for audit committees. Audit committees have become an important part of firms’financial reporting process; thus it is important to investigate the internal governance mechanisms for audit committees.We add to the literature on audit committees by exploring the determinants of audit committee compensation. We alsocontribute to research on board compensation by documenting an increasing trend toward differences in annual payamong board members within a firm.

We focus our analyses on the demand for monitoring of the financial reporting process. Specifically, we predict apositive association between audit committee compensation and the demand for monitoring of the financial reportingprocess. In conducting the empirical analyses, we control for director quality because audit committee compensation islikely related to both the demand for monitoring of financial reporting and the quality of the audit committee.

We examine audit committee compensation, both in absolute terms and relative to compensation committeecompensation. We find that total compensation and cash retainers are positively correlated with proxies for the demand formonitoring of the financial reporting process and measures of audit committee quality. Our empirical evidence isconsistent with the notion that the demand for monitoring of the financial reporting process is an important determinantof the compensation paid to audit committees. In addition, our findings that pay is increasingly different for audit andcompensation committee members suggest a new-found willingness on the part of boards and firms to acknowledge thedifferential contributions and outside opportunities of board members.

We view our study as a first step toward understanding the compensation arrangements for audit committees. Oneimportant area for future research is to examine how compensation incentives interact with reputation incentives toimpact the effectiveness of the audit committees in monitoring financial reporting.

Acknowledgment

We thank an anonymous reviewer, Brian Cadman, Jeff Chen (AAA discussant), Mike Lemmon, Clive Lennox, ScottSchaefer, Peter Wysocki (discussant), Jerry Zimmerman, and workshop participants at the University of Notre Dame, the2007 Southeast Summer Accounting Research Conference, the 2008 American Accounting Association Annual Meeting, andthe 2008 Journal of Accounting and Economics Conference for comments, and Jeff Wooldridge for helpful correspondenceon econometric methods. We thank Chun Wang for excellent research assistance.

References

Adams, R.B., 2003, What do boards do? Evidence from board committee and director compensation data. Working Paper.Adams, R.B., Ferreira, D., 2008. Do directors perform for pay?. Journal of Accounting and Economics 46, 154–171.Baker, G., Gibbs, M., Holmstrom, B., 1994. The internal economics of the firm: evidence from personnel data. Quarterly Journal of Economics 109, 881–919.Bell, T.B., Landsman, W.R., Shackelford, D.A., 2001. Auditors’ perceived business risk and audit fees: analysis and evidence. Journal of Accounting Research

39, 35–43.Biddle, J.E., Hamermesh, D.S., 1990. Sleep and the allocation of time. Journal of Political Economy 98, 922–943.Carcello, J.V., Neal, T.L., 2000. Audit committee composition and auditor reporting. Accounting Review 75, 453–467.Craswell, A.T., Francis, J.R., 1999. Pricing initial audit engagements: a test of competing theories. Accounting Review 74, 201–216.Danielsen, B.R., Van Ness, R.A., Warr, R.S., 2007. Auditor fees, market microstructure, and firm transparency. Journal of Business Finance and Accounting

34, 202–221.

ARTICLE IN PRESS


Defond, M.L., Hann, R.N., Hu, X., 2005. Does the market value financial expertise on audit committees of board of directors?. Journal of AccountingResearch 43, 153–193.

Deli, D.N., Gillan, S.L., 2000. On the demand for independent and active audit committees. Journal of Corporate Finance 6, 427–445.DeZoort, F.T., Hermanson, D.R., Archambeault, D.S., Reed, S.A., 2002. Audit committee effectiveness: a synthesis of the empirical audit committee

literature. Journal of Accounting Literature 21, 38–75.Doyle, J., Ge, W., McVay, S., 2007. Determinants of weaknesses in internal control over financial reporting. Journal of Accounting and Economics 44,

192–223.Fama, E.F., Jensen, M.C., 1983. Separation of ownership and control. Journal of Law and Economics 26, 301–325.Farrell, K.A., Friesen, G.C., Hersch, P.L., 2008. How do firms adjust director compensation?. Journal of Corporate Finance 14, 153–162.Francis, J.R., 1984. The effect of audit firm size on audit prices: a study of the Australian market. Journal of Accounting and Economics 6, 133–151.Ghosh, A., Pawlewicz, R.J., 2008. The impact of regulation on auditor fees: evidence from the Sarbanes-Oxley Act. Working Paper.Greene, W.H., 2000. Econometric Analysis, fourth ed. Macmillan.Griffin, P.A., Lont, D.H., 2008. The effects of auditor dismissals and resignation on audit fees: evidence based on SEC disclosures under Sarbanes-Oxley.

Working Paper.Hermalin, B.E., Weisbach, M.S., 1998. Endogenously chosen boards of directors and their monitoring of the CEO. American Economic Review 88, 96–118.Hermalin, B.E., Weisbach, M.S., 2003. Boards of directors as an endogenously determined institution: a survey of the economic literature. Economic Policy

Review 9, 7–26.John, T.A., John, K., 1993. Top-management compensation and capital structure. Journal of Finance 48, 949–974.Kinney, W.R., Palmrose, Z.-V., Scholz, S., 2004. Auditor independence, non-audit services, and restatements: was the U.S. government right?. Journal of

Accounting Research 42, 561–588.Klein, A., 2002a. Audit committee, board of director characteristics, and earnings management. Journal of Accounting and Economics 33, 375–400.Klein, A., 2002b. Economic determinants of audit committee independence. Accounting Review 77, 435–452.Larcker, D., Rusticus, T.O., 2008. On the use of instrumental variables in accounting research. Working Paper.Linck, J.S., Netter, J.M., Yang, T., 2009. The effects and unintended consequences of the Sarbanes-Oxley Act on the supply and demand for directors. Review

of Financial Studies 22, 3287–3328.Linn, S.C., Park, D., 2005. Outside director compensation policy and the investment opportunity set. Journal of Corporate Finance 11, 680–715.Macey, J.R., O’Hara, M., 2003. The corporate governance of banks. Economic Policy Review 9, 91–107.Menon, K., Williams, J.D., 1994. The use of audit committees for monitoring. Journal of Accounting and Public Policy 13, 121–139.Murphy, K.J., 1985. Corporate performance and managerial remuneration: an empirical analysis. Journal of Accounting and Economics 7, 11–42.O’Keefe, T.B., Simunic, D.A., Stein, M.T., 1994. The production of audit services: evidence from a major public accounting firm. Journal of Accounting

Research 32, 241–261.Oyer, P., 1995. Fiscal year ends and nonlinear incentive contracts: the effect on business seasonality. Quarterly Journal of Economics 113, 149–185.Rosen, S., 1992, Contracts and the market for executives. In: Werin, L., Wijkander, H. (Eds.), Contract Economics, Blackwell, Cambridge, MA.Ryan, H.E., Wiggins, R.A., 2004. Who is in whose pocket? Director compensation, board independence, and barriers to effective monitoring. Journal of

Financial Economics 73, 497–524.Simunic, D.A., 1980. The pricing of audit services: theory and evidence. Journal of Accounting Research 18, 161–190.Srinivasan, S., 2005. Consequences of financial reporting failure for outside directors: evidence from accounting restatements and audit committee

members. Journal of Accounting Research 43, 291–334.Vera-Munoz, S.C., 2005. Corporate governance reforms: redefined expectations of audit committee responsibilities and effectiveness. Journal of Business

Ethics 62, 115–127.Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA.Wooldridge, J.M., 2004. Statistical significance is okay, too: comment on ‘Size Matters’. Journal of Socio-Economics 33, 577–579.Yang, J.S., Krishnan, J., 2005. Audit committees and quarterly earnings management. International Journal of Auditing 9, 201–219.Yermack, D., 2004. Remuneration, retention, and reputation incentives for outside directors. Journal of Finance 59, 2281–2308.

Engel

Documents

Transcript of Engel