vafeas - 2005 - audit committees, boards, and the quality of reported earnings

Contemporary Accounting Research

Vol. 22 No. 4 (Winter 2005) pp. 1093–122 © CAAA

Audit Committees, Boards, and the
Quality of Reported Earnings

*

NIKOS VAFEAS,

University of Cyprus

Abstract

I use data on 252 U.S. firms between 1994 and 2000 to study the relationship between audit
committees and boards of directors with financial reporting quality. I initially document
several changes in committee and board profile during the sample period. Results from
logistic regressions suggest that measures of audit committee and board structure are related
to earnings quality in a manner that is generally consistent with the predictions of agency
theory. This study contributes to extant knowledge by employing different earnings quality
measures from prior studies, and by expanding the range of audit committee attributes
deemed important in determining audit committee performance.

Keywords

Audit committee; Board of directors; Earnings quality

JEL Descriptors

G30, G38

Comités de vérification, conseils d’administration
et qualité des résultats publiés
Condensé

Compte tenu du nombre croissant de scandales touchant l’information financière, le rôle des
comités de vérification dans la gouvernance d’entreprise a fait l’objet d’un débat nourri chez
les responsables de l’élaboration des politiques, les gestionnaires, les investisseurs et les uni-
versitaires. Au cours des vingt dernières années, ce débat a engendré une série de rapports
sur la gouvernance d’entreprise, prescrivant les structures souhaitables des comités de vérifica-
tion, dont le plus récent est l’œuvre du Blue Ribbon Committee (BRC). Tous ces rapports
s’articulent essentiellement autour d’un thème central : la possibilité pour les sociétés
d’accroître la qualité de leurs états financiers en améliorant la structure et le fonctionnement
de leurs comités de vérification. En outre, puisque la qualité des comités de vérification est
fondamentalement liée à celle des conseils d’administration d’entreprise dont ils sont issus,
les responsables de l’élaboration de politiques et les universitaires confèrent également à la
structure du conseil d’administration une importance déterminante dans la qualité des états
financiers.

* Accepted by Michel Magnan. I have benefited from numerous useful comments and suggestions
by Michel Magnan (associate editor), and two anonymous reviewers. Thanks also are due to Irene
Karamanou for providing valuable input on a programming issue, Maria Christodoulou for
competent research assistance, and First Call Corporation, a Thompson Financial Company, for pro-
viding forecast financial data. This project was partly funded by a University of Cyprus research
grant.

1094 Contemporary Accounting Research

CAR

Vol. 22 No. 4 (Winter 2005)
La simple observation semble indiquer que les comités de vérification et les conseils
d’administration présentent une grande variété transversale de structures et de modes de
fonctionnement. Compte tenu de cette variété, la nature de la relation entre les structures du
comité de vérification et du conseil d’administration et la qualité des résultats publiés (qualité
des résultats dans la suite) est un sujet de recherche primordial dont les répercussions sur les
politiques sont évidentes. Les données empiriques à cet égard confirment, pour la plupart, le
point de vue selon lequel les comités de vérification et les conseils d’administration structurés
de manière plus appropriée produisent de l’information de meilleure qualité en ce qui a trait
aux résultats. Les chercheurs précédents ont démontré l’existence d’un lien entre les structures
des conseils d’administration et des comités de vérification et les mesures de la qualité des
résultats fondées sur la comptabilité d’exercice, les retraitements des états financiers et les
propriétés des prévisions de résultats de la direction.
La présente étude élargit ce champ de recherche en traitant du lien entre les attributs
des comités de vérification et des conseils d’administration et deux variables substituts de la
qualité des résultats : la probabilité que les sociétés évitent un fléchissement des résultats et
la probabilité qu’elles préviennent des résultats négatifs inattendus. L’étude enrichit les con-
naissances existantes de différentes façons. Premièrement, elle puise dans un bassin de
recherche qui évolue à grands pas dans le domaine de la finance et qui met en lumière
d’importantes différences dans les motivations, la capacité et la volonté des administrateurs
externes de surveiller la direction. Le thème sous-jacent à ce champ de recherche veut qu’il
soit trop simpliste de cibler l’indépendance des administrateurs et que d’autres facteurs
puissent être liés à la performance d’un administrateur externe en matière de surveillance,
notamment les incitatifs au rendement sous forme d’actions, les fonctions assumées au sein
de conseils d’administration d’autres sociétés, la durée de l’appartenance au conseil d’admi-
nistration, et les responsabilités dans d’autres comités. L’auteur avance l’existence d’un lien
entre ces caractéristiques des administrateurs et la performance des comités de vérification,
et il procède à l’examen empirique de leur relation avec la qualité des résultats.
Deuxièmement, l’auteur s’abstient d’employer les mesures fondées sur les régularisations

inhabituelles utilisées dans les travaux précédents — et auxquelles peuvent être associées
des perturbations — en ayant recours aux faibles hausses des résultats et aux rendements
négatifs inattendus pour évaluer la qualité des résultats. La qualité des résultats étant une
notion difficile à cerner et aucune mesure ne s’étant véritablement démarquée par sa supé-
riorité, l’adoption d’une autre perspective méthodologique pour définir la qualité des résultats,
et pour produire des données complémentaires et corroboratives relativement à la question
étudiée, est un autre aspect de l’intérêt que présente l’étude. Enfin, l’auteur emploie un
échantillon constant de données, qui s’échelonnent sur sept ans, grâce auquel il peut analyser
les questions pertinentes sur une période plus longue. Les données étant structurées sous
forme d’échantillon constant, l’auteur soumet également à des tests les changements dans la
gouvernance qui réduisent les risques d’omission de variables souvent présents dans les
tests transversaux. L’étude des tendances chronologiques est également importante, car le
débat sur la gouvernance d’entreprise et les réformes politiques qui en ont résulté dans les
années 1990 ont entraîné d’importants changements dans les conseils d’administration et les
comités de vérification d’entreprise. L’échantillon de l’auteur, qui couvre la majorité des
années 1990, fait abstraction de l’hypothèse selon laquelle les structures de gouvernance
restent inchangées dans le temps.

Audit Committees, Boards, and the Quality of Reported Earnings 1095

CAR

Vol. 22 No. 4 (Winter 2005)
Les deux variables substituts de la qualité des résultats utilisées comme variables
dépendantes dans la présente étude permettent de déceler les faibles hausses des résultats et
les résultats inattendus. Les caractéristiques du comité de vérification et du conseil d’admi-
nistration retenues par l’auteur sont les suivantes : pourcentage des membres du comité
ayant un lien avec l’entreprise ; pourcentage des membres du comité assumant des fonctions
de cadres dans d’autres sociétés ouvertes ; pourcentage des membres du comité siégeant
aussi à un autre comité de vérification ; taille du comité de vérification ; nombre de réunions

(annuelles) du comité de vérification ; fraction cumulative des actions ordinaires de l’entre-
prise détenues par les membres du comité ; durée moyenne (en années) de l’appartenance
des membres du comité au conseil d’administration ; nombre moyen de sièges occupés par
les membres du comité dans d’autres conseils d’administration ; nombre moyen d’autres
comités du conseil auxquels siègent les membres du comité de vérification ; pourcentage
des actions ordinaires de l’entreprise détenues par l’ensemble des cadres et des administra-
teurs ; pourcentage d’administrateurs externes ; taille du conseil d’administration. Enfin, le
modèle englobe des contrôles relatifs aux pertes, au risque de litiges, à la taille de la société,
aux actionnaires institutionnels et à la croissance de la société.
L’échantillon initial de l’auteur est constitué de sociétés figurant au palmarès

Fortune 500

de 1995. Les institutions financières et les sociétés de services publics sont retirées de
l’échantillon du fait que le régime réglementaire particulier auquel elles sont soumises est
susceptible d’influer sur le rôle des systèmes d’information financière dans le processus de
gouvernance d’entreprise. Les sociétés à l’égard desquelles les données relatives à la gouver-
nance et les données financières sont insuffisantes sont également retirées. L’échantillon
définitif se compose de 1 621 observations relatives à 252 sociétés, pour la période 1994–2000.
L’auteur estime deux équations de régression logistique qui relient la probabilité des
deux mesures de la qualité des résultats aux structures du comité de vérification et du conseil
d’administration. Pour tenir compte de la possibilité que les caractéristiques du comité de
vérification et les pratiques de gestion du résultat soient en corrélation pour chaque société
dans le temps, l’auteur adapte aux données un modèle des effets aléatoires. Ce dernier est
estimé à partir de la totalité des 1 621 observations relatives aux 252 sociétés de l’échantillon.
Les résultats de l’analyse semblent indiquer que les structures des comités de vérification
et des conseils d’administration ont beaucoup évolué entre 1994 et 2000. Les conseils
d’administration de sociétés sont maintenant de taille plus restreinte et sont plus indépendants.
Les comités de vérification sont aussi plus indépendants et plus actifs, et ils comptent
davantage de cadres et d’administrateurs siégeant aussi à un autre comité de vérification. Le

total des actions de la société détenues par les membres du comité a également augmenté
sensiblement. Cette transformation cadre généralement avec les pressions venant de la
presse financière et les recommandations du BRC. Fait intéressant, la majorité de ces change-
ments ont précédé les recommandations du BRC, ce qui donne à penser que maintes sociétés
ont réagi par anticipation, peut-être pour détourner l’attention malvenue du public de la qualité
de leur information financière et pour éviter les coûts politiques qu’aurait occasionnés cet
intérêt.
L’auteur recense donc les liens qui existent entre les structures des comités de vérification
et des conseils d’administration et les faibles hausses des résultats de même que la prévention
des résultats négatifs, les deux variables substituts de la qualité des résultats. Ses observations
découlent de régressions logistiques multivariées expliquant la probabilité de faibles hausses

1096 Contemporary Accounting Research

CAR

Vol. 22 No. 4 (Winter 2005)
des résultats et de la prévention des résultats négatifs. Des régressions des changements
observés dans les conseils d’administration et les comités de vérification sur les indicateurs de
qualité des résultats livrent d’autres données probantes à cet égard. En somme, la présente
étude met au jour des récurrences empiriques qui élargissent l’éventail des explications
relatives aux modalités de l’influence exercée par les comités de vérification et les conseils
d’administration sur la qualité de l’information financière. Les facteurs tels que l’actionnariat
des membres du comité de vérification, les fonctions assumées au sein d’autres comités et, à
plus faible échelle, la durée de l’appartenance et les sièges occupés dans d’autres conseils
d’administration semblent jouer un certain rôle dans l’explication de la qualité des résultats.
Les conclusions de l’étude quant aux faibles hausses des résultats sont conformes à
celles des travaux précédents. Les membres du comité de vérification ayant un lien avec
l’entreprise sont associés à une qualité des résultats inférieure ; les membres du comité
assumant des fonctions de cadres dans d’autres sociétés protègent la direction plus que les

actionnaires ; et la fréquence des réunions du comité est associée à une qualité des résultats
supérieure. L’auteur note, en outre, que le fait que des membres du comité siègent aussi à un
autre comité de vérification est associé à une moindre occurrence de faibles hausses des
résultats. Les observations relatives à la prévention des résultats négatifs inattendus se com-
parent plus modérément aux conclusions des travaux précédents, puisqu’elles indiquent que
l’augmentation des incitatifs au rendement sous forme d’actions et la durée de l’apparte-
nance au conseil d’administration sont associées à une qualité des résultats inférieure. Dans
leur ensemble, ces constatations pourraient être utiles aux responsables de l’élaboration des
politiques qui envisagent des réformes de la gouvernance, aux gestionnaires et aux action-
naires qui s’intéressent à l’amélioration de la qualité des états financiers, et aux chercheurs
qui étudient la gouvernance d’entreprise et la qualité des résultats.
Enfin, les conclusions de l’auteur paraissent indiquer qu’en général, au sens usuel, les
comités de vérification et les conseils d’administration d’entreprise dont les structures et le
fonctionnement sont appropriés peuvent contribuer à l’amélioration de la qualité de l’infor-
mation financière. Les comités de vérification subissent actuellement d’autres changements
d’importance, par suite des nouvelles règles adoptées par la SEC et les Bourses de valeurs.
L’incidence à long terme de ces changements sur la qualité des états financiers des sociétés
est une question qui mérite de retenir l’attention des chercheurs dans les années à venir.

1. Introduction

Amid a growing number of financial reporting scandals, the role of audit committees
in corporate governance has been the topic of an active debate among policymakers,
managers, investors, and academics. Over the past 20 years, this debate fueled a
series of corporate governance reports prescribing desirable audit committee struc-
tures, the most recent by the Blue Ribbon Committee (BRC 1999).

1

The main

theme running through these reports is that firms can improve the quality of their
financial statements by structuring and operating their audit committees more
appropriately. Moreover, given that the quality of the audit committee is funda-
mentally linked to the quality of the corporate board from which the committee
originates, policymakers and academics have also posited board structure as a
determinant of financial statement quality (e.g., Pagano, Schwartz, Wagner, and
Marinelli 2002).

Audit Committees, Boards, and the Quality of Reported Earnings 1097

CAR

Vol. 22 No. 4 (Winter 2005)

Casual observation suggests that there is wide cross-sectional variation in the
way audit committees and boards are structured and operate. Given such variation,
the nature of the relation between audit committee and board structures and earn-
ings quality is a fundamental research question with clear policy implications. The
empirical evidence addressing this question has been mostly consistent with the
view that more appropriately structured audit committees and boards produce
earnings information of higher quality. Klein (2002), Xie, Davidson, and DaDalt
(2003), and Bédard, Chtourou, and Courteau (2004) draw on BRC recommenda-
tions addressing director independence and financial knowledge to document a
link between board and audit committee structures with accruals-based measures
of earnings quality. Abbott, Parker, and Peters (2004) similarly link audit commit-
tee characteristics to financial restatements, and Larcker and Richardson (2004)
find that corporate governance characteristics mitigate the relation between nonaudit
fees and accruals. Finally, Karamanou and Vafeas (2005) document a relation
between board and audit committee structures with the properties of management
earnings forecasts.

The present study extends this line of work by addressing the link between
audit committee and board attributes with two alternative earnings quality proxies:
the likelihood that firms avoid an earnings decline and the likelihood that firms avoid
a negative earnings surprise. This study adds to extant knowledge in the following
ways. First, it draws on a rapidly evolving body of research from the finance litera-
ture that illuminates important differences in the incentives, ability, and will-
ingness of outside directors to monitor management (e.g., Ferris, Jagannathan, and
Pritchard 2003; Yermack 2004; Perry and Peyer, 2005). The underlying theme of
this line of work is that a focus on director independence is overly simplistic, and
that additional factors may be related to an outside director’s monitoring perform-
ance, such as the director’s equity incentives, board seats in other corporations,
length of board tenure, and other committee responsibilities. In this paper I propose a
link between these director characteristics and audit committee performance,
empirically examining their relation to earnings quality.
Second, this paper abstracts from the potentially noisy abnormal-accruals-
based measures used in prior studies by relying on small earnings increases and
negative earnings surprises to measure earnings quality. Given that earnings quality
has been an elusive concept to pin down, and that no single measure has emerged
as definitively superior to others, using an alternative methodological perspective to
capture earnings quality and to provide complementary and confirmatory evidence
of the issue is also a potential contribution of this study. (In a similar spirit, recent
studies by Leuz, Nanda, and Wysocki 2003 and Lang, Raedy, and Yetman 2003 use
a variety of measures, in addition to accruals, to capture earnings quality.) Finally,
this paper employs a seven-year panel of data that allows me to study the relevant
issues over a longer time period. Given the panel structure of the data, I also per-
form tests of changes in governance that reduce the likelihood of omitted-variables
problems that are often present in cross-sectional tests. The study of trends over
time is also important because the corporate governance debate and resulting pol-
icy reforms of the 1990s have induced notable changes in corporate boards and

1098 Contemporary Accounting Research

CAR

Vol. 22 No. 4 (Winter 2005)

audit committees. My sample, covering most of the 1990s, abstracts from the
assumption that governance structures remain unchanged through time. Next, I
develop the study’s research proposition.
The remainder of this paper is organized as follows. Section 2 lays out the
research proposition. I describe the measures of audit committee effectiveness and
board structure in section 3. The data and methodology are discussed in section 4
and the results in section 5. Section 6 presents the summary and conclusion.

2. Research proposition

Corporate boards in general, and audit committees in particular, are responsible for
monitoring the information contained in financial reports. As suggested earlier,
differences in board and audit committee structures across firms, coupled with dif-
ferent levels of effectiveness of the various governance structures, suggest that the
quality of board monitoring over financial reports is likely to vary across firms.
Evidence addressing this notion empirically has clear policy implications. Out of
equilibrium, evidence of higher earnings quality in firms with properly structured
and functioning audit committees would justify policy efforts by those arguing that
uniform committee structures be mandated. In contrast, lack of such evidence
would be more consistent with an efficient contracting view of the firm, suggesting
that reform may lead to monitoring that is redundant and costly because firms can
choose an optimal mix of control mechanisms on their own.
Motivated by these policy concerns, and building on earlier work, the present
study addresses the link between audit committee structure and audit committee

functioning with the quality of reported earnings empirically. Given that audit
committees are the principal liaison between management and auditors and are
chiefly responsible for reporting on earnings quality to the board of directors, I
expect that their monitoring performance should partly determine the extent of
earnings manipulation by managers. Moreover, the quality of the audit committee
is fundamentally linked to the quality of the corporate board because all audit com-
mittee members are also members of the board, and are appointed by the board
itself, while audit committee decisions have to be ratified by the board as a whole.
Accordingly, I also hypothesize that well-structured and functioning corporate
boards are associated with improved financial reporting quality.
To study the link between boards and audit committees with earnings quality,
this paper considers two complementary measures of earnings quality, following
Frankel, Johnson, and Nelson 2002 and Ashbaugh, LaFond, and Mayhew 2003.
First, it focuses on the tendency of firms to manage earnings so as to avoid reporting
a negative earnings change by isolating all firm-years with a marginally positive
earnings change. This measure is based on prior evidence (e.g., Burgstahler and
Eames 2003) that there is a discontinuity in the distribution of earnings changes
around zero, suggesting that firms that report marginally positive earnings changes are
likely trying to avoid reporting bad news to the market. The second measure is
based on work by Matsumoto 2002, who suggests that firms try to meet analyst
expectations benchmarks in order to avoid negative earnings surprises, apparently
by managing earnings. These earnings quality measures have two potential

Audit Committees, Boards, and the Quality of Reported Earnings 1099

CAR

Vol. 22 No. 4 (Winter 2005)

advantages over accruals-based measures: they are likely to be less noisy and

incorporate the effect of cash flow manipulation. Extending this work, I suggest
that the likelihood that a firm will report earnings of poor quality in order to avoid
an earnings decline or a negative earnings surprise will be inversely related to the
effectiveness of its audit committee and board of directors.
R

ESEARCH

P

ROPOSITION

.

The likelihood of reporting a marginal earnings
increase or artificially avoiding a negative earnings surprise will be lower
in firms with properly structured audit committees and boards of directors.

3. Measures of audit committee and board structure

To measure audit committee effectiveness, I initially draw on the BRC recommen-
dations to suggest a set of five audit committee characteristics that are likely to be
related to audit committee performance. I then expand this set by examining the
role of four additional audit committee characteristics and three general govern-
ance characteristics that are likely to be associated with earnings quality.

Audit committee characteristics highlighted by the Blue Ribbon Committee

Monitoring the financial reporting process and ensuring high-quality financial
statements is one of the prime tasks bestowed on corporate boards in general, and

independent outside directors in particular. In line with this, the BRC report and prior
evidence suggest that the presence of inside directors on the audit committee is
likely to be related to poorer financial reporting choices (Klein 2002; Bédard et al.
2004), because insiders often have incentives to tolerate a lower quality of reported
earnings. I therefore expect lower earnings quality in committees with insider
participation.
Second, the BRC recommends that all committee members be financially literate,
and that at least one be a financial expert. Empirical evidence by DeZoort 1998,
Xie et al. 2003, and Bédard et al. 2004 is consistent with the notion that committee
member financial expertise enhances audit committee performance. Following Beasley
and Salterio 2001, I suggest that business executives and directors who serve on
the audit committee of another firm are likely to be financially knowledgeable,
given that such knowledge is normally necessary for appointment to such positions.
Conversely, it is possible that business executives are more likely to identify with,
and thus befriend, management who underperform their monitoring role (DeZoort
and Salterio 2001). Thus, ex ante, the link between earnings quality and business
executives is undetermined, whereas a positive relation is expected between earn-
ings quality and committee members with experience in other audit committees.
A third potential determinant of audit committee performance is committee
size. Ex ante, adding more directors to a committee is likely to have a nonlinear
effect on committee performance. Initially, adding more members to the commit-
tee enhances performance because there are more people on whom to draw. When
committees grow too large, performance declines because of process losses and
diffusion of responsibility. Reflecting concerns about understaffed committees, the
BRC (1999) suggests that audit committees should have at least three members, to

1100 Contemporary Accounting Research

CAR

Vol. 22 No. 4 (Winter 2005)

ensure a minimum required knowledge base and a lower likelihood that the com-
mittee as a whole will be “handled” by management. Xie et al. (2003) find an
insignificant relation between audit committee size and discretionary accruals.
Given somewhat mixed predictions, I tentatively expect a positive relation between
earnings quality and audit committee size.
An important objective for audit committees is to provide their members with
sufficient time to carry out their duties. Vafeas (1999) suggests that board meetings
in general are a credible measure of board activity. Focusing on audit committees in
particular, Menon and Williams (1994) and Xie et al. (2003) suggest that meetings
are a reasonable proxy for committee effectiveness. In its sample audit committee
charter, the BRC (1999) suggests that committees should hold at least four meetings
per year. Accordingly, I expect that audit committees that meet more frequently
also operate better, thereby leading to a higher quality of reported earnings.

Additional audit committee characteristics

One implicit assumption underlying research on audit committees to date has been
that the monitoring performance of financially knowledgeable outside directors is
more or less uniform. Committee members are assumed to perform similarly to one
another provided that they are outsiders and have the financial knowledge to cope
with the demands of audit committee service. The BRC report (1999) reinforces this
view. Such research, however, can be enriched by a growing body of evidence from
the finance literature highlighting other important differences among directors. This
evidence suggests that, in addition to affiliation and knowledge, director perform-
ance is likely to be related to equity incentives, total board seats, other committee
service, and length of tenure on the board (e.g., Ferris et al. 2003; Yermack 2004;
Perry and Peyer 2005). Drawing on this work, I posit the notion that these director
characteristics, detailed below, will also be significant in evaluating audit commit-

tees and explaining differences in performance across committee member directors
and, ultimately, differences in performance across audit committees.
The first such characteristic is audit committee member equity ownership. To
align the interests of shareholders and outside directors, who are themselves agents
of shareholders, firms routinely grant equity to outside directors. Yermack (2004)
finds that director equity awards are made systematically, consistent with the predic-
tions of agency theory. Moreover, there is growing evidence that outside directors
who own more equity in the firm protect shareholder interests more effectively (for
example, by reducing the likelihood of fraud litigation; see Ferris et al. 2003). In
the context of audit committees, higher equity ownership by committee members
is likely to reduce the danger of these directors colluding with management to
manipulate earnings (for example, to inflate executive pay) because such collusion
would ultimately harm their own interests as well. I therefore expect that the
greater the fraction of equity that audit committee members own, the greater will
be their motivation to monitor the financial reporting process more effectively, thus
providing earnings information of higher quality.
Second, director performance may be related to length of tenure on the board.
There are two competing views on the impact of director tenure lengths on audit

Audit Committees, Boards, and the Quality of Reported Earnings 1101

CAR

Vol. 22 No. 4 (Winter 2005)

committee effectiveness. More seasoned directors gain more experience and better
knowledge about the firm’s operations and thus are able to exercise more effective
decision control on management’s financial reporting choices than are junior, less
experienced directors. In contrast, excessively lengthy board service might com-
promise independence because senior directors may be more likely to befriend

management (Vafeas 2003), becoming less critical of the quality of its financial
reports (Beasley 1996). Yermack (2004) does not find a link between outside direc-
tor turnover with either director tenure length or audit committee membership.
Given conflicting predictions, the link between earnings quality and director tenure
length is addressed empirically.
The market for directorships provides corporate directors with a powerful
incentive to perform their duties well. The more board seats directors hold, the
more sensitive to reputation effects they are likely to be (e.g., Yermack 2004).
News of poor audit committee performance, as conveyed by a financial reporting
scandal, is likely to stigmatize the overseeing audit committee members, damaging
their reputation and ability to acquire additional board seats. Furthermore, Ferris et
al. (2003) do not find any evidence that holding more board seats compromises the
directors’ monitoring performance in any way. Consistent with this finding, Perry
and Peyer (2005) find that outside directorships by an executive enhance a firm’s
value. I thus expect a positive relation between the average number of directorships
held by audit committee members and earnings quality.
Finally, additional committee service may have a bearing on director perform-
ance in the audit committee. More committee service provides directors with
greater knowledge of the firm’s affairs and more experience in monitoring manage-
ment. Furthermore, appointment to a committee may signal a director’s ability to
monitor management. Vafeas (1999) finds that firms that operate more board com-
mittees also have more active and, presumably, more effective boards. Also, direc-
tors who hold additional board seats, and thus have a greater reputation capital,
serve at least as much as other directors in board committees (Ferris et al. 2003).
Thus, I tentatively expect a positive relation between the number of other com-
mittee memberships held by audit committee members and audit committee per-
formance, as captured by earnings quality.

General governance characteristics

The fraction of shares owned by officers and directors as a group reflects the incen-
tives the management team has to protect the interests of shareholders. Warfield,
Wild, and Wild (1995) find that higher managerial ownership is associated with a
greater information content of earnings, presumably because manager-shareholders
secure better financial information for investors. I similarly expect that a greater
fraction of insider ownership will be associated with higher earnings quality.
Second, there is a vast body of literature suggesting that a greater proportion
of outsiders serving on corporate boards is associated with improved board moni-
toring over shareholders. Consistent with this notion, Beasley (1996), Dechow,
Sloan, and Sweeney (1996), and Klein (2002) find that a higher representation of
outside directors on the board is associated with higher financial reporting quality.

1102 Contemporary Accounting Research

CAR

Vol. 22 No. 4 (Winter 2005)

I similarly expect the degree of board independence to be positively related to
earnings quality.
Finally, larger boards have been found to be associated with lower firm value
because of the higher coordination costs and process losses they face (Yermack
1996). I also expect that larger boards make the monitoring process more difficult,
thereby leading to financial reports of lower quality, and hypothesize that earnings
quality is lower for firms with larger boards of directors.

4. Data and methodology

The data

My initial sample comprises firms that are listed in the 1995

Fortune 500

survey.
Financial institutions (Standard Industrial Classification [SIC] codes 6000–6999)
and utilities (SIC codes 4900–4999) are deleted from the sample, because their
special regulatory environment is likely to influence the role of financial reporting
systems in the corporate governance process. Also deleted are firms with no elec-
tronic filings with the Securities and Exchange Commission (SEC) between 1994
and 2000, firms for which COMPUSTAT provides insufficient financial data for
1995, and foreign firms. This leaves a final sample of 252 firms.
I extend the sample period one year before 1995 and five years forward, creating
a seven-year panel of data from 1994 to 2000. Firms are allowed to exit the panel
as they merge, go private or bankrupt, or otherwise cease to exist in their initial
form, thereby limiting the effects of survivorship bias. Thus, the sample comprises
an unbalanced panel of data over a seven-year period, ranging from 252 firms in
1995 to 182 firms in 2000, with a total of 1,621 firm-year observations.
Information on audit committee and board structures is collected from the
electronic filings of proxy statements with the SEC in EDGAR for each firm and
each year. Financial information for each firm is collected from COMPUSTAT.
Information on institutional holdings is collected from Compact Disclosure.
Finally, information on analyst forecasts and on corresponding actual earnings per
share (EPS) figures is collected from Institutional Brokers Estimate System (I/B/E/S).
The two earnings quality proxies used as dependent variables in this study
capture small earnings increases and earnings surprises. The audit committee and
board characteristics employed are:

percentage of committee insiders

,

percentage
of active business executives

,

percentage of members with other audit committee
experience

,

audit committee size

,

audit committee meetings

,

2



stock ownership of
committee members

,

mean tenure per committee member

,

mean directorships per
committee member

,

mean committee memberships per committee member

,

3



inside
ownership

,

percentage of board outsiders

, and

board size

. Finally, the model
includes five control variables that are considered by both Frankel et al. 2002 and
Ashbaugh et al. 2003: a

loss dummy

; a

litigation risk dummy

; firm size, proxied by

equity capitalization

;

institutional holdings

; and firm growth, proxied by the

market-
to-book ratio

. (See the appendix for detailed definitions of the variables.)

Audit Committees, Boards, and the Quality of Reported Earnings 1103

CAR

Vol. 22 No. 4 (Winter 2005)

Methodology

In sum, I estimate two logistic regression equations that link the likelihood of low
earnings quality, approximated by the likelihood of a small earnings increase and
the likelihood of meeting or just beating analyst expectations, respectively, using
the following model:
P(

low earnings quality

)

=

α

0



+



α

1

insiders

+



α

2

executives



+



α

3

expertise



+



α

4

size

+



α

5

meetings



+



β

1

ln(

stock

)

+



β

2

tenure

+



β

3

directorships



+



β

4

committees

+



γ

1

inside ownership



+



γ

2

board outsiders

+



γ

3

board size



+



δ

1

loss



+



δ

2

litigation

+

δ

3

ln(

equity cap.

)

+



δ

4

institutions

+



δ

5

market-to-book



+

u

.
To account for the possibility that audit committee characteristics and earnings-
management practices are correlated for each firm through time, I fit a random-
effects model to the data (or, more accurately, a random-intercepts model because
the intercept is the only random parameter). To this end I use the GLIMMIX macro
in SAS (see Wolfinger and McConnell 1993 for the theory behind this macro, and
Littell, Milliken, Stroup, and Wolfinger 1996 for a description of the macro). The
random-effects model illuminates how between-firm variation in boards and audit
committees helps to explain poor earnings quality. More specifically, the random-
effects model assumes that the slope coefficients are constant, while the intercept
varies over individual firms and over time. Also, the model assumes that the firm
and time effects are random variables, and that they are uncorrelated with the
model’s independent (governance and control) variables. Finally, the model assumes
that the number of sample firm-years is sufficiently large. Given the small number
of years examined here, the empirical models in this paper employ a random effect
for firms but fixed effects for time (year dummies). The models are estimated using
the full sample of 1,621 observations for 252 firms.
Table 1 provides a summary of the expected relations among each of the audit
committee, board, and control variables to the likelihood of avoiding reporting a
small earnings increase or a negative earnings surprise. Ex ante, each relation is
denoted as positive (

+

), negative (

−), or undetermined (?).
5. Results
The trend in audit committee and board characteristics
between 1994 and 2000
To understand the trends in audit committee structure and activity and board struc-
ture for my sample firms, in Table 2 I trace their audit committee and board profile
for the period 1994–2000. First, I observe a decline in the relative number of insid-
ers serving on the committee, from 3.69 percent in 1994 to 1.86 percent in 2000
(t = −2.33; p < 0.03), in line with committees becoming more independent of man-
agement’s influence through time. The largest decline occurs in the last sample
1104 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
TABLE 1
Expected relations between explanatory variables and the likelihood of reporting small earnings increases or avoiding negative earnings surprises
Audit committee characteristics highlighted by the BRC
Percentage of committee insiders + Lower earnings quality
Percentage of active business executives ? Undetermined
Percentage of members with other audit committee experience − Higher earnings quality
Audit committee size − Higher earnings quality
Audit committee meetings − Higher earnings quality
Additional audit committee characteristics
Median stock ownership of committee members − Higher earnings quality
Mean tenure per committee member ? Undetermined
Mean directorships per committee member − Higher earnings quality
Mean committee memberships per committee member − Higher earnings quality
General governance characteristics

Inside ownership − Higher earnings quality
Percentage of board outsiders − Higher earnings quality
Board size + Lower earnings quality
Other control variables
Loss dummy − Higher earnings quality
Industry with high litigation risk + Lower earnings quality
Equity capitalization + Lower earnings quality
Institutional holdings + Lower earnings quality
Market-to-book ratio ? Undetermined
Note:
All variables are defined in the appendix.
Explanatory variables Expected sign Lower earnings quality
Audit Committees, Boards, and the Quality of Reported Earnings 1105
CAR Vol. 22 No. 4 (Winter 2005)
year, from 3.18 percent to 1.86 percent. Furthermore, there is a substantial increase
in the percentage of active business executives serving on the audit committee
(t = 4.13; p < 0.01) and in the level of audit committee activity as approximated by
committee meetings (t = 2.87; p < 0.01). Finally, committees have not changed
meaningfully in size over the sample period.
4
In general, these findings are consist-
ent with the recommendations of the Blue Ribbon Committee, emphasizing the
need for more independent, informed, and active audit committees. Furthermore,
the timing of these changes suggests many firms acted in anticipation of the BRC
recommendations, because such changes precede the issuance of the BRC report.
Notably, the median fraction of common stock owned by all audit committee
members together increased steadily and significantly over the sample period, from
0.017 percent in 1994 to 0.039 percent in 2000 (Wilcoxon z = 3.41; p < 0.01). This
finding reflects the recent trend for more stock-based pay for corporate directors.
(Nevertheless, average ownership declined slightly during the sample period, as

indicated by the t-statistic, because of the undue influence of a few extreme obser-
vations.) Third, the percentage of outside directors serving on boards increased
steadily over the sample period, from 75.1 percent in 1994 to 80.3 percent in 2000
(t = 4.81; p < 0.01), consistent with public pressures for more independent boards
of directors. Also, boards have shrunk modestly in size, from 11.7 directors in
1994 to 11.3 directors in 2000 (t = −1.89; p < 0.07), consistent with concerns that
larger boards are less effective monitors of management. Finally, firms report a
small earnings increase and a small positive earnings surprise in about 25 percent
of the sample years.
The relationship between audit committees and boards with the likelihood of
small earnings increases and avoidance of negative earnings surprises
Table 3 presents Pearson (Spearman) pair-wise correlations above (below) the diag-
onal among the governance and earnings quality variables. In general, it appears
that board and audit committee measures are positively correlated, and are thus
complements in monitoring management, whereas ownership concentration is neg-
atively correlated to effective boards and audit committees, and is thus a substitute
for these mechanisms in monitoring management.
Next, I estimate the relation among audit committee structure and activity,
board structure, and control variables to the likelihood of low earnings quality in a
multivariate logistic regression model. The model is estimated twice, employing
small earnings increases and negative surprise avoidance as the dependent vari-
able, respectively. Table 4 presents the results. Importantly, each of the models in
Table 4 is jointly significant at the 0.001 level, as signified by the Pearson
χ
2
statistic.
Results from the model on small earnings increases suggest that a greater
number of insiders on audit committees increases the likelihood of lower earnings
quality. This finding is consistent with Klein 2002 and BRC 1999, and suggests
that the pressures exerted on policymakers and top management teams to make

audit committees more independent are justified. The results are very similar when
I introduce a dummy that is set to 1 for audit committees with no insiders, and 0
otherwise.
1106 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
TABLE 2
The trend in audit committee and board characteristics for 252 sample firms as reported in proxy statements between 1994 and 2000
Audit committee characteristics highlighted
by the BRC
Percentage of committee insiders 3.69 3.85 3.40 3.11 3.75 3.18 1.86 3.32 −2.33
*
−1.48
Percentage of active business executives 48.47 51.16 52.73 54.38 54.97 56.63 59.16 53.66 4.13
†
3.73
†
Percentage of members with other audit
committee experience 20.75 20.99 22.12 21.37 21.76 21.09 19.57 21.15 −0.39 −0.20
Audit committee size 4.38 4.50 4.61 4.58 4.56 4.61 4.61 4.55 1.39 1.31
Audit committee meetings 3.51 3.49 3.59 3.57 3.60 3.73 3.90 3.61 2.87
†
3.00
†
Additional audit committee characteristics
Median stock ownership of committee
members 0.017 0.019 0.021 0.023 0.029 0.036 0.040 0.026 −0.97 3.41
†
Mean tenure per committee member 8.80 9.05 9.12 9.04 8.89 8.65 8.60 8.89 −0.36 −0.06
Mean directorships per committee member 2.25 2.19 2.15 2.16 2.15 2.07 2.06 2.15 −1.49 −1.39
Mean committee memberships per

committee member 1.33 1.31 1.29 1.29 1.27 1.28 1.29 1.30 −1.12 −1.04
(The table is continued on the next page.)
1994 1995 1996 1997 1998 1999 2000 Total
Sample size 243 252 248 244 237 215 182 1,621 t-test Wilcoxon z
Audit Committees, Boards, and the Quality of Reported Earnings 1107
CAR Vol. 22 No. 4 (Winter 2005)
TABLE 2 (Continued)
General governance characteristics
Insider ownership 7.00 6.98 6.15 5.96 5.88 5.34 6.11 6.23 −0.96 0.50
Percentage of board outsiders 75.1 76.2 77.1 77.4 78.3 79.0 80.3 77.5 4.81
†
4.17
†
Board size 11.7 11.8 11.9 11.6 11.5 11.6 11.3 11.6 −1.89 −2.27
*
Firm-years with a small earnings increase 0.32 0.32 0.19 0.25 0.20 0.20 0.28 0.25
Firm-years with earnings surprise avoidance 0.23 0.28 0.27 0.31 0.24 0.23 0.21 0.25
Notes:
All variables are defined in the appendix. t- and Wilcoxon z-statistics denote comparisons between the respective variable values in 1994 and 2000.
*
Significant at the 0.05 level.
†
Significant at the 0.01 level.
Sample size
1994 1995 1996 1997 1998 1999 2000 Total
243 252 248 244 237 215 182 1,621 t-test Wilcoxon z
1108 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
TABLE 3
Pearson (Spearman) correlations above (below) the diagonal among variables approximating audit committee and board structure, and the likelihood

of loss avoidance and negative surprise avoidance for 252 firms between 1994 and 2000 (n = 1,621)
1. Small earn. increase 1.00 0.09 0.07 0.02 −0.04 0.02 −0.08 0.01 0.04 −0.00 −0.05 0.02 −0.06 0.05
2. Surprise avoidance 0.09 1.00 −0.04 −0.06 −0.06 0.03 −0.00 −0.04 0.07 0.02 −0.02 0.04 −0.03 0.04
3. Com. insiders 0.06 −0.04 1.00 −0.28 −0.09 −0.07 −0.04 0.06 0.27 −0.15 0.02 0.07 −0.34 0.03
4. Com. executives 0.02 −0.07 −0.24 1.00 0.03 −0.04 0.03 0.02 −0.20 0.08 −0.01 −0.02 0.14 −0.02
5. Com. experience −0.05 −0.04 −0.09 0.02 1.00 0.04 0.07 −0.11 −0.04 0.33 −0.01 −0.12 0.17 0.02
6. Com. size 0.02 0.03 0.02 −0.04 0.09 1.00 0.01 0.00 −0.06 0.17 0.13 −0.20 0.31 0.34
7. Com. meetings −0.07 0.02 −0.02 0.01 0.10 0.03 1.00 −0.02 −0.06 0.07 0.01 −0.04 0.07 0.05
8. Dir. stock ownership −0.03 −0.06 0.19 −0.10 −0.20 −0.03 −0.07 1.00 0.04 −0.11 −0.00 0.37 −0.01 0.04
9. Director tenure 0.04 0.08 0.18 −0.18 −0.00 0.00 −0.04 0.18 1.00 −0.07 0.04 0.09 −0.15 0.01
10. Directorships −0.00 0.02 −0.13 0.08 0.39 0.24 0.09 −0.30 −0.04 1.00 −0.08 −0.13 0.20 0.14
11. Com. memberships −0.05 −0.02 −0.01 −0.01 −0.05 0.15 −0.01 0.08 0.05 −0.07 1.00 0.03 −0.19 −0.10
12. Inside ownership 0.00 0.01 −0.07 −0.06 −0.19 −0.24 0.11 0.52 0.14 −0.23 0.04 1.00 −0.20 −0.08
13. Board outsiders −0.06 −0.04 −0.24 0.10 0.19 0.29 0.08 −0.06 −0.16 0.22 −0.19 −0.20 1.00 −0.05
14. Board size 0.06 0.04 0.06 −0.02 0.07 0.35 0.07 −0.20 0.04 0.19 −0.10 −0.17 0.02 1.00
(The table is continued on the next page.)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
Audit Committees, Boards, and the Quality of Reported Earnings 1109
CAR Vol. 22 No. 4 (Winter 2005)
Contrary to the notion that business executives enhance earnings quality, audit
committees composed of directors who are business executives have a greater like-
lihood of reporting small earnings increases, despite the fact that these executives
are presumably more likely to be financially literate than other directors. Therefore,
it appears that managers also tilt earnings figures in their favor through captive out-
side audit committee members, even when audit committees are, on the surface,
composed solely of outside directors. This finding is also consistent with DeZoort
and Salterio 2001 and the notion that such executives identify with, and are thus
more empathetic to, management’s wishes.
5
Consistent with expectations, I find that the likelihood of reporting a small

earnings increase decreases as the number of committee meetings increases. This
suggests that greater activity by the committee improves monitoring, thereby lead-
ing to financial reports of better quality. This result is also in line with the work of
Menon and Williams 1994 and Xie et al. 2003. Finally, the loss dummy is nega-
tively related to the likelihood of small earnings increases.
Results from the model studying the likelihood of avoiding negative earnings
surprises are generally weaker. Importantly, increasing committee stock ownership
reduces the likelihood of avoiding negative earnings surprises, suggesting that
equity investment aligns the interests of audit committee members with the interests
TABLE 3 (Continued)
Notes:
A small earnings increase is a positive change of up to 2 percent over last year’s net income
over market value. Firms avoid earnings surprises if they meet or just beat analysts’
expectations; that is, if an earnings surprise is between 0.00 and 0.01 cents over the
consensus (mean) analyst forecast, measured as the last forecast prior to the
announcement of annual earnings. Committee insiders is the fraction of firm
executives, committee members who were employees of the firm within the past
three years and their relatives, and committee members with a fiduciary relation to
the firm either directly or through their principal employer. Committee executives is
the fraction of committee members who are executives in other public firms.
Committee experience is the fraction of committee members who served in the audit
committee of another Fortune 500 firm during the sample period. Committee size is
the number of directors serving in the audit committee at fiscal year-end. Committee
meetings is the number of meetings held by the audit committee during the fiscal
year. Director stock ownership is the fraction of common stock cumulatively owned
by committee members. Director tenure is the average number of years committee
members have served on the board. Directorships is the average number of additional
board seats held by committee members. Committee memberships is the average
number of additional board committees per audit committee member. Inside
ownership is the percentage of common stock owned by officers and directors as a

group. Board outsiders is the fraction of outside (non-executive) directors serving on
the board. Board size is the number of directors serving on the board at fiscal year-
end. The correlation coefficients in bold are significant at p < 0.05.
1110 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
of other shareholders. Among the control variables, litigation risk, equity capitaliza-
tion, and institutional holdings are positively related to negative earnings avoidance.
These results are very similar to results reported by Frankel et al. 2002 and Ash-
baugh et al. 2003.
6
In sum, the analysis presented in Table 4 offers evidence that
audit committee structure and operation and board structure are related to financial
reporting quality.
TABLE 4
Random-effects logistic regressions of the relationship between audit committee and
board characteristics with earnings quality proxies
Intercept −2.130
†
−6.015
†
(−2.67) (−5.51)
Percentage of committee insiders + 1.400
*
−1.663
(2.27) (−1.88)
Percentage of active business executives ? 0.557
*
−0.412
(2.37) (−1.31)
Percentage of members with other audit −−0.247 −0.55

committee experience (−0.86) (−1.51)
Audit committee size − 0.070 0.059
(1.30) (0.89)
Audit committee meetings −−0.119
*
0.021
(−2.56) (0.37)
Ln(median stock ownership of committee − 0.672 −8.71
*
members) (0.21) (−1.99)
Mean tenure per committee member ? 0.005 0.029
(0.35) (1.44)
Mean directorships per committee member −−0.006 0.006
(−0.10) (0.07)
Mean committee memberships per committee −−0.408 0.134
member (−1.72) (0.42)
Ln(inside ownership) − 0.024 0.256
*
(0.29) (2.20)
Percentage board outsiders −−0.070 −0.072
(−1.41) (−1.17)
Board size + 0.064 0.085
(1.65) (0.85)
(The table is continued on the next page.)
Explanatory variables
Expected
sign
Small
earnings
increases

Avoiding
negative
earnings
surprises
Audit Committees, Boards, and the Quality of Reported Earnings 1111
CAR Vol. 22 No. 4 (Winter 2005)
One possible explanation for the difference in results between the two measures
is that they capture different types of earnings-management behavior, as evidenced
by the relatively low correlation between them (r = 0.09). The low correlation
probably reflects the different stimuli (benchmarks) captured by the two measures,
and to which management responds — that is, the desire to report earnings that are
at least at last year’s level, and the desire to meet analysts’ expectations. It could be
that the capacity or willingness of the audit committee to ensure earnings quality
depends on the impact that earnings management will have. Finally, the audit com-
mittee and board variables studied here appear to capture highly overlapping
effects. Thus, although individual variable coefficients may be insignificant, each
of the estimated models employing board and audit committee variables jointly
explains the likelihood of earnings management at a statistically significant level.
7
Further analysis
In this section, I describe a number of robustness and validity checks that consider
overall results from annual regressions, tests of trends in the results, finer analysis
TABLE 4 (Continued)
Loss dummy −−1.563
†
−0.504
(−4.32) (−1.78)
Industry with high litigation risk + 0.011 0.657
†
(0.08) (3.07)

Ln(equity capitalization) + 0.108 0.329
†
(1.57) (3.32)
Institutional holdings +−0.001 0.013†
(−0.33) (2.74)
Market-to-book ratio ? 0.039 0.041
(0.68) (0.55)
Year dummies included Yes Yes
Sample size 1,621 1,621
Model significance Pearson
χ
2
Pearson
χ
2
1,624.5
†
1,087.3
†
Notes:
All variables are defined in the appendix.
t-statistics are in parentheses. The models are estimated using the GLIMMIX macro in SAS.
*
Significant at the 0.05 level.
†
Significant at the 0.01 level.
Explanatory variables
Expected
sign
Small

earnings
increases
Avoiding
negative
earnings
surprises
1112 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
on director ownership, correlated observations for each firm across years, and
potential confounding due to omitted variables.
Cumulative effects of annual regressions
To check robustness of the results, I also estimate the model in seven annual logis-
tic and ordinary least squares (OLS) regressions and compute summary statistics
to measure the overall effect of annual results. Specifically, I estimate Z1 and Z2
statistics capturing variable significance based on annual t-statistics from least
squares regressions as in Clarkson, Li, and Richardson 2004 and, alternatively,
replacing t-statistics in the Z1 and Z2 formulas with Wald
χ
2
statistics from logistic
regressions, square-rooted and adjusted for the parameter sign. The two approaches
yield very similar results. Results from logistic regressions are presented in Table 5.
Results explaining small earnings increases are highly consistent with the
results reported in Table 4 from the pooled regressions. Specifically, active business
executives and, more weakly, insiders are related to lower earnings quality, while
meeting frequency and mean committee memberships per committee member are
related to higher earnings quality. Results explaining negative earnings surprises
are modestly similar to results from Table 4: equity incentives are associated with
higher earnings quality in both the pooled and annual models, while business
executives and experienced committee members are only associated with higher

earnings quality in the annual tests. (Nevertheless, the coefficient sign is also nega-
tive in the pooled tests.) Together, the results from the annual regressions in Table
5 are broadly consistent with earlier pooled results from Table 4, and in some cases
are somewhat stronger.
Persistence of the results
To address the issue of persistence of the variable coefficients over time, I ran the
logistic regression pooled across firms and years interacting in sequence each of
the independent variables with seven yearly dummies. The model was thus esti-
mated once for each independent variable and its interactions with time, while
each of the other variables was kept in the model in its first-order form. I then
tested whether each independent variable coefficient remained the same over time
by examining the equality of the seven interactive term coefficients. The resulting
Wald
χ
2
statistic (Table 5) indicates how different the coefficients are across time.
In general, there appear to be significant differences in the relationship between
governance and earnings quality across time, with the exception of the committee
insiders and equity ownership variables. A low signal-to-noise ratio in annual
regressions makes it very difficult to detect reliable trends in time and is the likely
cause of the variation in variable significance across years. In line with this explana-
tion, differences in results across years are observed even for the control variables
that were found to be consistently significant by prior research. Overall results on
the control variables using Z1 and Z2 tests are consistent with prior work.
Audit Committees, Boards, and the Quality of Reported Earnings 1113
CAR Vol. 22 No. 4 (Winter 2005)
TABLE 5
Trend analysis of the relationship between audit committees, boards, and earnings quality measures
Intercept −1.84 −8.12
*

2.39 −3.90 −1.01 0.51 −3.76 0.33 0.22 —
−4.13 0.17 −6.03
†
−5.83
†
−3.23 −5.27
†
−4.97 −1.24 −1.61 —
Percentage of committee insiders 2.09 −0.91 2.22 3.90
†
0.57 −1.91 1.29 + 1.71
†
1.37 11.04
0.35 −0.13 −5.41
†
−7.83
*
−4.09 −3.42 −5.36 −3.52 −3.04 9.06
Percentage of active business executives 0.53 1.32
†
−0.42 0.43 0.49 2.07
†
0.10 ? 2.41
†
2.08
†
23.76
*
−0.19 −0.13 −0.50 −1.55 −0.29 −0.79 −0.98 −2.34
†

−3.07
*
15.61
†
Percentage of members with other audit −1.11 0.28 −0.93 −1.26 0.70 −0.24 −0.88 −−1.23 −1.06 16.82
†
committee experience −0.33 0.43 −0.88 −3.80
*
−1.35 −0.36 0.49 −2.27
†
−1.43 8.14
Audit committee size 0.25
†
0.23 −0.12 0.05 0.03 0.08 −0.14 − 1.15 0.99 31.92
*
−0.27
†
−0.03 0.09 0.12 0.26 0.34
†
0.23 1.64 1.15 11.94
Audit committee meetings −0.10 −0.10 0.04 −0.05 −0.12 −0.49
*
−0.08 −−2.23
†
−2.26
†
27.13
*
0.12 −0.11 −0.01 0.11 0.11 0.01 0.11 0.96 1.26 13.39
†

Ln(median stock ownership of 3.94 −6.52 7.81 0.07 −7.53 6.06 −5.62 −−0.15 −0.17 3.33
committee members) −7.74 7.71 −3.93 −2.86 −50.46 −25.15 −13.78 −1.66
†
−1.98
†
8.88
Mean tenure per committee member −0.01 0.07
†
−0.05 −0.01 0.10
†
−0.04 −0.05 ? 0.36 0.23 28.76
*
0.05 −0.03 0.08
†
−0.01 0.01 0.10
†
0.07 2.14
†
1.73 10.08
Mean directorships per committee member 0.20 −0.16 0.01 −0.20 0.10 0.27 0.02 − 0.45 0.40 26.59
*
0.29 −0.19 −0.09 −0.05 −0.13 0.04 0.15 0.27 0.24 11.52
Mean committee memberships per committee −0.85 −0.63 −0.09 0.26 −1.09 −1.11 −0.01 −−1.80
†
−1.61 15.21
†
member 0.57 −0.06 −0.97 0.74 −0.33 0.06 −0.52 0.19 0.22 7.27
(The table is continued on the next page.)
Explanatory variables 1994 1995 1996 1997 1998 1999 2000 Sign Z1 Z2 Wald
χ

2
1114 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
TABLE 5 (Continued)
Ln(inside ownership) −0.04 0.33 −0.16 0.07 −0.12 −0.20 −0.21 −−0.60 −0.37 26.11
*
0.21 −0.29 0.09 −0.12 0.66
*
0.12 0.15 1.23 0.91 9.84
Percentage board outsiders −0.32
†
−0.18 −0.02 0.09 −0.09 −0.05 0.31 −−1.46 −0.79 29.49
*
0.33 −0.05 −0.01 −0.06 0.05 −0.23 −0.26 −0.54 −0.43 16.68
†
Board size 0.22
†
0.11 −0.08 0.05 0.07 0.06 −0.06 + 1.41 1.19 32.85
*
−0.30
†
0.06 0.04 0.13 0.06 0.08 0.24 1.15 0.90 16.73
†
Sample size 243 252 248 244 237 215 182
Notes:
The top value in each cell refers to small earnings increases and the bottom value to negative surprise avoidance. The Z1 and Z2 tests examine the
overall effect of each variable based on Wald
χ
2
statistics from logistic regressions, square-rooted and sign-adjusted. The Wald

χ
2
test examines
the persistence of the individual variable coefficients over time. Z1 and Z2 tests are one-tailed except when there is no clear directional
expectation ex ante. Each model includes controls for losses, litigation risk, equity capitalization, institutional ownership, and the market-to-
book ratio.
*
p < 0.01.
†
p < 0.05.
Explanatory variables 1994 1995 1996 1997 1998 1999 2000 Sign Z1 Z2 Wald
χ
2
Audit Committees, Boards, and the Quality of Reported Earnings 1115
CAR Vol. 22 No. 4 (Winter 2005)
Robustness of the results on director ownership
One interesting finding of my analysis from Tables 4 and 5 is the negative relation
between median equity ownership of audit committee members and the likelihood
of earnings management. I probe further into this finding in two ways. First, given
that committee insiders normally own a greater fraction of equity in the firm than
outsiders, I repeat the tests on negative surprise avoidance presented in Table 4
after excluding firm-years with insider participation in the audit committee. I find
that for the sample of all-independent committees, the director ownership variable
becomes even more significant in deterring earnings management. Second, I find
this result to be driven by large outside equity holders serving on the audit committee.
Correlation of observations for each firm across years
To further account for the correlation that exists between observations for the same
firm across years, I reestimated the two models in Table 4 using the generalized
estimating equation method. This method employs a White-like variance estimator
that accounts for the correlation among observations for the same firm (Horton and

Lipsitz 1999). The results from this model are qualitatively similar to those
reported in Table 4. For small earnings increases, committee insiders and business
executives are associated with lower earnings quality. The likelihood of negative
surprise avoidance decreases with committee member equity ownership and
increases, weakly, with committee member tenure on the board.
Potential confounds due to omitted variables
To shed further light on the role of audit committees and corporate boards in
enhancing financial reporting quality, in additional tests I focus on the relation
between one-year changes in the governance variables with earnings quality. The
focus on variable changes accounts for the possibility that board and audit committee
characteristics are endogenously determined, depending on factors that are not con-
trolled for in this study and that may confound the results, such as a firm’s managerial
labor, corporate control, and other factor and product markets. The main insight of
the focus on variable changes is that unobserved omitted variables are likely to
vary widely across firms, but exhibit little variation for a single firm over time.
Two pooled logistic regression models are estimated testing the relation
between changes in audit committees, boards, and control variables to the likeli-
hood of small earnings increases and negative surprise avoidance respectively. The
SIC-based litigation risk variable that remains constant for each firm across years
and the loss dummy are included in their level form. Given that a first-difference
approach is used, changes for the first sample year are not available, and the total
sample comprises 1,330 observations. The results are presented in Table 6.
In both models, the chi-squared statistic is significant, suggesting that each
model is important in explaining the likelihood of earnings manipulation. In the
case of small earnings increases, consistent with the notion that director incentives
improve director monitoring quality, increases in median stock ownership of audit
committee members reduce the likelihood of small earnings increases. Similarly,
1116 Contemporary Accounting Research
CAR Vol. 22 No. 4 (Winter 2005)
more mean directorships per committee member, and thus a more valuable reputa-

tion capital by committee members, also reduce the likelihood of reporting small
earnings increases.
Furthermore, increases in mean tenure per committee member are associated
with a greater incidence of negative earnings avoidance. This finding suggests that
directors may become entrenched in their positions by staying on the board for a
long period, potentially acting to protect management rather than shareholders.
TABLE 6
Logistic regressions of the impact of annual changes in audit committees and the board of
directors on the likelihood of small earnings increases and negative surprise avoidance
Intercept 0.22 −2.62
*
(0.15) (20.86)
Percentage of committee insiders +−0.46 −0.15
(2.06) (0.21)
Percentage of active business executives ? −0.14 0.02
(0.58) (0.02)
Percentage of members with other audit −−0.03 0.24
committee experience (0.03) (1.96)
Audit committee size − 0.01 0.43
†
(0.09) (4.39)
Audit committee meetings − 0.01 0.45
(0.03) (0.89)
Ln(median stock ownership of committee −−0.32
†
0.17
members) (3.53) (0.98)
Mean tenure per committee member ? 0.14 0.43
†
(0.43) (5.62)

Mean directorships per committee member −−0.39
†
0.01
(5.27) (0.06)
Mean committee memberships per committee −−0.27 0.12
member (0.09) (0.37)
Ln(inside ownership) −−0.07 −0.20
(0.17) (1.56)
Percentage board outsiders − 0.001 −0.31
†
(0.31) (3.60)
Board size + 0.01 0.12
(0.02) (0.54)
(The table is continued on the next page.)
Explanatory variables
Expected
sign
Small
earnings
increases
Negative
surprise
avoidance
Audit Committees, Boards, and the Quality of Reported Earnings 1117
CAR Vol. 22 No. 4 (Winter 2005)
Increases in the percentage of board outsiders reduce the likelihood of negative
earnings avoidance, consistent with the broad notion that more independent boards
help to reduce the incidence of earnings management.
It should be noted that the power of this change analysis is limited because of
the relative “stickiness” of the governance variables (their limited intertemporal

variation), which is likely to explain some of the differences in results between the
levels and changes tests. Nevertheless, taken together, the results from Tables 4, 5,
and 6 are broadly consistent with the notion that the levels of, and changes in,
some audit committee and board characteristics reduce the likelihood of poor earn-
ings quality.
6. Summary and conclusion
In this paper, I address how the structure and activity of audit committees and the
structure of corporate boards relate to the quality of earnings information produced
by firms. First, I find that audit committee and board structures changed noticeably
between 1994 and 2000. Corporate boards became smaller and more independent.
Audit committees became more independent and more active. More business
TABLE 6 (Continued)
Loss dummy −−2.20
*
−1.30
*
(22.34) (14.39)
Litigation risk + 0.34 0.89
*
(2.63) (21.25)
Ln(equity capitalization) + 0.46 0.101
*
(3.52) (16.85)
Institutional holdings + 0.01 0.08
(0.44) (0.21)
Market-to-book ratio ? −0.13
†
−0.44
(28.39) (3.42)
Sample size 1,330 1,330

McFadden’s pseudo-R
2
7.2% 6.5%
Model significance —
χ
2
100.60
*
100.05
*
Notes:
Wald
χ
2
statistics are in parentheses. All variables are defined in the appendix. The
independent variables are computed as the change between the current and prior year
except the litigation risk and loss dummies.
*
Significant at the 0.01 level.
†
Significant at the 0.05 level.
Explanatory variables
Expected
sign
Small
earnings
increases
Negative
surprise
avoidance