davidson et al - 2005 - internal governance structures and earnings management

Accounting and Finance 45 (2005) 241–267
Internal governance structures and earnings management
Ryan Davidson
a
, Jenny Goodwin-Stewart
b
, Pamela Kent
a
a
UQ Business School, University of Queensland, St Lucia, 4072, Australia
b
School of Accountancy, Queensland University of Technology, Brisbane, 4001, Australia
Abstract
This paper investigates the role of a firm’s internal governance structure in constrain-
ing earningsmanagement. It is hypothesized that the practice of earningsmanagement
is systematically related to the strength of internal corporate governance mechanisms,
including the board of directors, the audit committee, the internal audit function and
the choice of external auditor. Based on a broad cross-sectional sample of 434 listed
Australian firms, for the financial year ending in 2000, a majority of non-executive
directors on the board and on the audit committee are found to be significantly asso-
ciated with a lower likelihood of earnings management, as measured by the absolute
level of discretionary accruals. The voluntary establishment of an internal audit func-
tion and the choice of auditor are not significantly related to a reduction in the level
of discretionary accruals. Our additional analysis, using small increases in earnings
as a measure of earnings management, also found a negative association between this
measure and the existence of an audit committee.
Key words: Audit committee; Corporate governance; Earnings management; Internal
audit function
JEL classification: M40; M41
doi: 10.1111/j.1467-629x.2004.00132.x
1. Introduction

Recent cases of inappropriate accounting practices, both in Australia and overseas,
have focused attention on the need for strong corporate governance mechanisms.
The authors acknowledge with thanks the helpful comments of two anonymous reviewers
and the Associate Editor, Professor Stephen Taylor. We also thank Christine Jubb, Ping-
Sheng Koh and participants at the 2003 Annual Conference of the Accounting and Finance
Association of Australia and New Zealand, held in Brisbane, Australia. Financial support from
the UQ Business School/KPMG Centre for Business Forensics and Queensland University of
Technology is also gratefully acknowledged.
Received 14 August 2003; accepted 29 April 2004 by Stephen Taylor (Associate Editor).
C

AFAANZ, 2005. Published by Blackwell Publishing.
242 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
Strong governance involves balancing corporate performance with an appropriate
level of monitoring (Cadbury, 1997). In the present paper, we explore the relationship
between governance mechanisms and earnings management by firms in Australia
and, hence, our focus is on the monitoring role of governance. The mechanisms
we examine are an independent board of directors (Shleifer and Vishny, 1997), an
independent board chairperson, an effective audit committee (Menon and Williams,
1994), the use of internal audit (Clikeman, 2003) and the choice of external auditor
(Becker et al., 1998; Francis et al., 1999).
Prior research has investigated the role of governance mechanisms in reducing
fraudulent financial reporting (Beasley, 1996; Dechow et al., 1996; Jiambalvo, 1996).
These studies have established a negative relationship between effective governance
mechanisms and financial reporting decisions that are in breach of Generally Ac-
cepted Accounting Principles (GAAP). However, a relatively new area of research
is the association between corporate governance and earnings management. Peasnell
et al. (2000) document that earnings management is negatively associated with the
independence of the board of directors, while other studies find significant relation-
ships between audit committee characteristics and earnings management (Chtourou

et al., 2001; Xie et al., 2001; Klein, 2002a).
The examination of the association between internal governance structures and
the practice of earnings management in Australian firms is motivated by several
factors. With the exception of Peasnell et al. (2000), which uses UK data, existing
research is predominantly US based. Therefore, we explore whether the internal
governance–earnings management relationship holds in an institutional environment
where corporate governance is less regulated and choice of governance mechanisms
is voluntary (Von Nessen, 2003). In Australia, at the time of the present study (2000),
listed companies were not required to have an audit committee or an internal audit
function. Furthermore, corporate regulators favour a principles-based approach to
governance rather than a rules-based approach (ASX, 2003). Although a similar ap-
proach exists in the UK, Peasnell et al. (2000) examine only the relationships between
earnings management and the proportion of outside directors on the board and the
existence of an audit committee. We extend this research by exploring the effect of
additional audit committee variables such as size and frequency of meetings as well
as the independence of members. We also extend board independence to examine
whether the separation of the chief executive and board chair roles is associated with
earnings management. A further contribution is the inclusion of internal audit as a
governance mechanism that is likely to be associated with a reduction in the level of
earnings management. While there has been increasing emphasis on the role played
in governance by internal audit, no prior earnings management studies have included
this variable. Australia is an ideal setting to examine this issue as evidence sug-
gests that many listed companies in Australia do not have an internal audit function.
Goodwin and Kent (2003) report that the use of internal audit is associated with the
size of the company and its commitment to strong corporate governance and risk
management.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 243

Our principal tests, using absolute discretionary accruals to measure earnings man-
agement, suggest that a lower level of earnings management is associated with the
presence of non-executive directors on the board. We also find a negative associa-
tion between earnings management and audit committees comprising a majority of
non-executives, but no relationship between earnings management and committees
comprised solely of non-executives. Our results do not support a relationship between
earnings management and the use of internal audit or the choice of a Big 5 auditor.
Additional testing, using small positive changes in earnings as an indication of earn-
ings management, suggests that audit committees are associated with this measure of
earnings management. These results have important practical implications because of
the heightened interest in corporate governance matters from governments, regulators
and standard setters.
The remainder of the paper is divided into four sections. Section 2 provides the
theoretical background for the study and develops the hypotheses. Section 3 outlines
the research method used to test the hypotheses. It also discusses the measurement
of earnings management through the estimation of discretionary accruals. Section 4
reports the present study’s results. Section 5 concludes by discussing the implications
of the researchfindings, highlighting potential limitations and consideringfuture areas
for research.
2. Theoretical background and hypotheses
2.1. Earnings management
The preparation and disclosure of true and fair financial information is central
to corporate governance, as it provides stakeholders with a foundation to exercise
their rights, in order to protect their interests (OECD, 1999). However, earnings
management, defined as: ‘the practice of distorting the true financial performance of
(a) company’ (SEC, 1999, p. 3), effectively weakens this monitoring mechanism as
it might conceal poor underlying performance.
The published literature has developed and empirically tested a variety of moti-
vations for earnings management to occur (Fields et al., 2001). These fall broadly
within the categories of agency costs, information asymmetries and externalities af-

fecting non-contracting parties. However, we are primarily concerned with the extent
to which certain corporate governance attributes limit the opportunity to manage
earnings, rather than specific incentives for earnings management to occur. Although
we attempt to control for two widely documented motives for earnings management;
namely, avoiding breaching debt covenants (Defond and Jiambalvo, 1991, 1994)
and avoiding political costs (Watts and Zimmerman, 1978; Jones, 1991; Jiambalvo,
1996), our approach is to examine a broad cross-section of firms rather than identify-
ing a specific subset with strong incentives to engage in earnings management. Such
subsets of firms are often context-specific (e.g. recent managerial change, hostile
takeover or new capital raising) and these contexts are likely to be endogenous to the
internal governance mechanisms we examine.
C

AFAANZ, 2005
244 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
2.2. Internal governance structure
The internal governance structure of a firm consists of the functions and processes
established to oversee and influence the actions of the firm’s management. The role
of these mechanisms in relation to financial reporting is to ensure compliance with
mandated reporting requirements and to maintain the credibility of a firm’s financial
statements (Dechow et al., 1995). The mechanisms that we examine in the present
study are the board of directors, the audit committee, the internal audit function and
the choice of external auditor.
1
2.2.1. Board of directors
Fama and Jensen (1983a,b) recognize the board as the most important control
mechanism available because it forms the apex of a firm’s internal governance struc-
ture. In terms of monitoring financial discretion, an effective board of directors should
ascertain the validity of the accounting choices made by management and the financial
implications of such decisions (NYSE, 2002).

From an agencyperspective,the ability of the board to act as an effective monitoring
mechanism is dependent upon its independence from management (Beasley, 1996;
Dechow et al., 1996). Board independence refers to the extent to which a board
is comprised of non-executive directors who have no relationship with the firm
beyond the role of director.
2
A non-executive director is defined as a director who
is not employed in the company’s business activities and whose role is to provide
an outsider’s contribution and oversight to the board of directors (Hanrahan et al.,
2001). A non-executive director who is entirely independent from management is
expected to offer shareholders the greatest protection in monitoring management
(Baysinger and Butler, 1985). Fama and Jensen (1983a,b) posit that the superior
monitoring ability of non-executives can be attributed to the incentive to maintain
their reputations in the external labour market.
The published literature is supported by Australian and international corporate
governance guidelines, which recognize the importance of the monitoring role of
non-executive directors (AIMA, 1997, 1995; OECD, 1999; NYSE, 2002; ASX,
2003; Standards Australia International, 2003; Bosch Committee, 1995; Cadbury
Committee, 1992). These guides suggest that best practice with respect to board
1
Another mechanism is the firm’s internal control system, but companies in Australia are
not required to report on the strength of controls. We assume that external and internal au-
ditors would ensure that controls are adequate. Furthermore, earnings management by senior
management generally overrides controls that are in place.
2
According to the BRC (1999), audit committee members are independent if: (i) they, their
spouses or children do not currently work or have not worked at the organization or its affiliates
within the past 5 years; (ii) they have not received compensation from the organization or its
affiliates for work other than board service; (iii) they are not partners, shareholders or officers
of a business with which the organization has significant business.

C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 245
composition is, at least, a majority of non-executive or independent directors.
3
This
is supported by research evidence such as Beasley (1996), who finds that the presence
of independent directors on the board reduces the likelihood of financial statement
fraud, and Dechow et al. (1996), who report that firms with a greater proportion
of non-executive directors are less likely to be subject to Securities and Exchange
Commission (SEC) enforcement actions for violating US GAAP. Based on these
findings, both Peasnell et al. (2000) and Chtourou et al. (2001) predict that board
independence is also likely to be associated with a reduction in earnings man-
agement. While Peasnell et al. (2000) find empirical support for their prediction
with respect to UK firms, Chtourou et al. (2001) fail to find an association be-
tween earnings management and board independence for a sample of US firms.
Despite these conflicting results, we hypothesize a negative association between
board independence and earnings management. Therefore, we test the following
hypothesis:
H
1a
: Earnings management is negatively associated with the independence of the
board of directors.
Another important characteristic of boards is whether there is a separation of the
roles of the chairperson and the Chief Executive Officer (CEO). While Arthur and
Taylor (1993) point out that the underlying economic determinants of separating these
roles are not well understood, corporate governance guidelines assume that a board’s
ability to perform a monitoring role is weakened when the CEO is also the chairper-
son of the board (e.g. Cadbury Committee, 1992; ASX, 2003; Standards Australia

International, 2003). The appointment of the CEO to the position of chair can lead to
a concentration of power (Beasley, 1996) and possible conflicts of interest, resulting
in a reduction in the level of monitoring. This leads to the following hypothesis:
H
1b
: Earnings management is negatively associated with the separation of the roles
of CEO and board chair.
2.2.2. Audit committee
In order to more efficiently perform their duties, boards of directors might delegate
responsibilities to board committees. In relation to monitoring the financial discretion
of management, it is the audit committee that is likely to provide shareholders with
the greatest protection in maintaining the credibility of a firm’s financial statements.
This is because of the specialized monitoring of financial reporting and audit activities
provided by the audit committee.
3
There has been a global increase in the demand for non-executives on the board because of
the requirement for audit committee members to be independent. However, it is acknowledged
that executive directors, with their in-depth knowledge of the business, also play an important
role. Hence, the key is to find the appropriate balance with regard to board composition (Klein,
2002a,b).
C

AFAANZ, 2005
246 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
In Australia, the Government’s Corporate Law Economic Reform Program (Au-
dit Reform and Corporate Disclosure) Act 2004 (CLERP 9) (Commonwealth of
Australia, 2004) proposes mandatory audit committees for the Top 500 listed compa-
nies. Furthermore, the Australian Stock Exchange (ASX) amended its listing rules in
2003 to require any company that was included in the Standard and Poor’s 500/ASX
All Ordinaries Index at the beginning of its financial year to have an audit commit-

tee during that year. In addition, in March 2003, the ASX Corporate Governance
Council released a best practice guide that recommends that all companies have an
audit committee (ASX, 2003). However, in the year 2000, there was no requirement
mandating audit committees. The only rule in place was that a listed company with
no audit committee should disclose the reasons for this in a corporate governance
statement in the company’s annual report.
Prior published literature indicates that the effectiveness of an audit committee
is dependent, in part, on the extent to which the committee is independent, its fre-
quency of meetings and its size.
4
It is contended that audit committees are unable
to function effectively when members are also executives of the firm (Lynn, 1996;
BRC, 1999). Both the published literature and governance reports suggest that au-
dit committees should consist exclusively of non-executive or independent directors
(e.g. Menon and Williams, 1994; BRC, 1999; ASX, 2003).
5
This is supported by
research that demonstrates a relationship between audit committee independence and
a higher degree of active oversight and a lower incidence of financial statement fraud
(Jiambalvo, 1996; McMullen and Raghundan, 1996; Wright, 1996). In contrast, how-
ever, Klein (2002a) reports a negative relation between earnings management and a
majority of independent directors on the audit committee, but finds no meaningful
relationship between earnings management and an audit committee comprised solely
of independent directors.
To effectively monitor the financial discretion of management, the audit commit-
tee is expected to review the financial reporting process, as well as to facilitate a
flow of information among the board of directors, the internal and external auditors,
and management (McMullen and Raghundan, 1996). However, both Cohen et al.
(2000) and Gibbins et al. (2001) report that external auditors are sceptical of the role
that audit committees play in reducing conflicts between auditors and management.

Hence, to effectively discharge their responsibilities, audit committees need to be
4
It is recognized that other characteristics are also important indicators of an audit committee’s
effectiveness. These include the financial literacy or expertise of the committee members (Kirk
Panel, 1994; Goodwin and Yeo, 2001; Goodwin, 2003; ASX, 2003), the existence of an audit
committee charter and the number of meetings held with the external auditor (BRC, 1999).
However, as current disclosure requirements do not mandate such information, these variables
cannot be tested using publicly available information.
5
The Corporate Governance Council recommends that all members of the committee should
be non-executive directors, with a majority (including the chair) being independent. However, it
acknowledges that international practice is for all members to be independent and it encourages
companies to move towards such composition within the next 3 years (ASX, 2003).
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 247
active (Collier, 1993; Hughes, 1999) and to be of sufficient size (Cadbury Committee,
1992; CIMA, 2000). Audit committee activity has been operationalized through the
number of committee meetings held during the financial year (Chtourou et al., 2001;
Xie et al., 2001), with the expectation that the more often a committee meets, the
more likely it is to carry out its responsibilities. Studies have found that the fre-
quency of audit committee meetings is negatively associated with both earnings
management, as measured by discretionary current accruals (Xie et al., 2001), and
the likelihood of enforcement action by the SEC (McMullen and Raghundan, 1996).
With regard to size, several corporate governance reports have proposed that the
committee should consist of at least three members (BRC, 1999; NYSE, 2002; ASX,
2003).
The existence of an effective audit committee provides a firm with an added layer
of governance, which is expected to constrain earnings management behaviour. This

leads to the following hypothesis.
H
2
: Earnings management is negatively associated with the presence of an effective
audit committee.
2.2.3. Internal audit function
In addition to the audit committee, firms can voluntarily establish an internal audit
function to supplement their existing internal governance framework. If established,
this function provides firms with an assurance and consulting service, which can im-
prove the effectiveness of their risk management, control, and governance processes
(IIA, 1999). An internal audit function is also expected to facilitate the operation
and effective functioning of the audit committee, as the goals of the audit function
are closely aligned with the financial reporting oversight responsibilities of the audit
committee (Scarbrough et al., 1998; Goodwin and Yeo, 2001; Goodwin, 2003). The
formation of an internal audit function is endorsed by governance reports (NYSE,
2002) and prior literature (Collier, 1993; Goodwin and Kent, 2003) as a means of
improving internal governance processes.
Although traditionally internal audit has focused more on controls and operational
risks, there has been increasing emphasis in the professional literature on the need
to also focus on earnings management and inappropriate financial reporting (Eighme
and Cashell, 2002; Martin et al., 2002; Rezaee, 2002; Clikeman, 2003; Hala, 2003).
Sherron Watkins, former Enron vice president, believes that internal auditors should
look for warning signs such as undue pressure from senior management to meet
earnings targets and compensation arrangements that might encourage employees
to manipulate earnings in order to receive financial rewards (Hala, 2003). Clikeman
(2003) argues that internal auditors should not only be actively involved in detecting
earnings management, but that they should take a proactive approach to educating
managers and directors about the dangers of the practice. Eighme and Cashell (2002)
regard the role of internal audit in detecting earnings management as being a com-
plementary one to that of external audit. They believe that both should be actively

C

AFAANZ, 2005
248 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
involved in the detection of inappropriate earnings management, thereby providing
two unrelated opinions to the audit committee.
These arguments suggest that the presence of an internal audit function should be
associated with a lower level of earnings management. Accordingly, the following
hypothesis is proposed.
H
3
: Earnings management is negatively associated with the presence of an internal
audit function.
2.2.4. External audit
The choice of a firm’s auditor is another internal governance mechanism that is
likely to be associated with earnings management. Research evidence suggests that
the large audit firms are perceived to perform higher quality audits than smaller audit
firms (DeAngelo, 1981). While examples such as Enron in the USA and HIH in
Australia might suggest otherwise, the large firms are considered to be more effective
monitors of the financial reporting process compared to smaller firms (Francis et al.,
1999; Francis and Krishnan, 1999; Kim et al., 2003). Krishnan (2003) argues that,
not only do the large audit firms have more resources and expertise to detect earnings
management, but they also have a greater incentive to protect their reputation because
of their larger client base. Past studies demonstrate that clients of Big 5 auditors report
lower levels of earnings management than clients of non-Big 5 auditors (Becker
et al., 1998; Francis et al., 1999).
6
Therefore, we expect that firms that choose a
Big 5 auditor are less likely to engage in earnings management. This gives rise to the
following hypothesis:

H4: Earnings management is negatively associated with the use of a Big 5 auditor.
3. Research design
The present study involves a cross-sectional analysis of 434 firms listed on the ASX
for the financial year ending in 2000. To test our hypotheses, we use two primary
models which regress absolute discretionary accruals on a set of governance and
control variables. The two models differ only in their measure of audit committee
independence. We also conduct additional tests, using alternative measures of both
the dependent variable and a number of independent variables.
3.1. Sample selection
Our preliminary sample of 568 firms comprised companies for which annual
reports were available either on the Connect4 database, on company websites or in
hard copy. Financial information and information pertaining to boards of directors
and audit committee characteristics were obtained from disclosures made in annual
6
For the reporting period ending in 2000, there were five main audit firms (the Big 5).
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 249
reports. However, annual report disclosure concerning the use of internal audit is
not mandatory. Therefore, this information was obtained by one of three methods.
The first method involved consulting the University of Queensland/KPMG Centre for
Business Forensics database.
7
The second method involved examining annual report
disclosures.
8
The final method involved directly contacting firms, which were either
not included on the database or whose annual reports made no mention of an internal
audit function.

To arrive at the final sample, exclusions were made on the basis of industry
classification and insufficient governance or financial information. For the purpose
of industry classification, the Global Industry Classification Standard (GICS) was
adopted.
9
Firms were classified according to the 59 GICS industries and, consistent
with prior research, industries with less than 8 firms were eliminated. Firms in the
financial sector were also excluded because of their unique working capital structures
(Klein, 2002a) and the added layer of governance imposed through regulation (Barn-
hart et al., 1994). Those firms with missing financial or governance information were
also excluded, as were 4 firms with extreme values for discretionary accruals.
10
After
these exclusions were made, the sample for the study was limited to 434 firms in 24
industries.
11
Table 1 presents a breakdown of the sample according to GICS sector,
showing the broad cross-section of firms included within the sample.
3.2. Measurement of the variables
3.2.1. Earnings management
The published literature has developed several tests of earnings management, in-
cluding the assessment of accounting policy changes (Healy, 1985; Sweeney, 1994),
specific accounting transactions (McNichols and Wilson, 1988), discretionary accru-
als (Jones, 1991) and small profits or small changes in earnings (Holland and Ram-
say, 2003). The present study uses discretionary accruals as the primary measure of
7
This database comprises information on 464 firms that responded to a survey sent to all firms
listed on the ASX in 2000. The survey included a question concerning the use of internal audit.
Approximately 65 per cent of the final sample was generated from this source.
8

Firms were recorded as having an internal audit function when the words ‘internal audit’ were
contained within the annual report. Approximately 20 per cent of the final sample provided
this information in their annual reports.
9
The GICS classification structure consists of 10 broad economic sectors aggregated from
23 industry groups, 59 industries and 122 sub-industries. The present study adopts the GICS
structure as it has become the new industry classification scheme of the ASX since 1 July
2002.
10
These firms had discretionary accruals divided by lagged assets of 1.5 or more and tests
revealed that they could be considered to be outliers.
11
From the original 568 firms, 48 were in industries with less than 8 firms, 58 were in the
financial sector and 24 had missing data, giving 438 firms before excluding the four outliers.
C

AFAANZ, 2005
250 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
Table 1
Analysis of sample by GICS sectors and industries
Sector Industry Industry Sample Sample ASX
a
(number) code (%) (%)
Energy Oil & gas (2) 101010 23 5.30 4.47
Materials Chemicals (3) 151010 8 32.26 24.76
Construction materials (4) 151020 8
Metals & mining (6) 151040 116
Paper & forest products (7) 151050 8
Industrials Building products (9) 201020 9 12.44 12.58
Construction & engineering (10) 201030 14

Industrial conglomerates (12) 201050 15
Commercial services & supplies (15) 202010 16
Consumer Auto components (21) 251010 8 17.97 12.25
discretionary Hotels, restaurants & leisure (26) 253010 23
Media (27) 254010 31
Specialty retail (31) 255040 16
Consumer Beverages (33) 302010 15 8.29 5.22
staples Food products (34) 302020 21
Health care Health care equipment & supplies (38) 351010 9 5.76 7.28
Health care providers & 351020 8
services (39)
Pharmaceuticals (40) 352010 8
Financials Real estate (45) 404010 25 5.76 18.05
Information Internet software & services 451010 8 6.45 11.17
technology Software electronic 451030 12
Equipment & instruments 451030 8
Telecommunication Diversified telecommunication 501010 17 3.92 3.06
Services Services (54)
Utilities Electric utilities (56) 551010 8 1.84 1.16
Total 434 100 100
a
Based on 1218 listed entities, as of 31 March 2002.
ASX, Australian Stock Exchange; GICS, Global Industry Classification Standard.
earnings management while we also use small profits and small changes in earnings
in our additional analysis.
Earnings management research has been dominated by studies that have followed
the general discretionary accruals framework proposed by McNichols and Wilson
(1988). This framework partitions accruals into non-discretionary and discretionary
components on the assumption that a high level of discretionary accruals (DAC)
suggests that a firm is engaging in earnings management. The most frequently used

method to decompose accruals is the modified-Jones model (Dechow et al., 1995).
This model assumes that the non-discretionary component of total accruals (NDAC)
is a function of the change in revenues adjusted for the change in receivables and the
level of property, plant and equipment, which drive working capital requirements and
depreciation charges, respectively.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 251
We use the cross-sectional version of the modified-Jones model (DeFond and
Jiambalvo, 1994; Becker et al ., 1998; Bartov et al., 2000). Under this model, the level
of discretionary accruals for a particular firm is calculated as the difference between
the firm’s total accruals and its non-discretionary accruals (NDAC), as estimated with
equation (1):
NDAC
ij t
= [ˆα
j
[1/A
ij t −1
] +
ˆ
β
1j
[REV
ij t
− REC
ij t
/A
ij t −1

]
+
ˆ
β
2j
[PPE
ij t
/A
ij t −1
]
(1)
where ˆα
j
,
ˆ
β
1j
and
ˆ
β
2j
are industry-specific coefficients estimated from the following
cross-sectional regression:
12
TAC
ij t
/A
ij t −1
= α
j

[1/A
ij t −1
] + β
1j
[REV
ij t
/A
ij t −1
]
+ β
2j
[PPE
ij t
/A
ij t −1
] + ε
ij t
(2)
where:
TAC
ijt
= total accruals for firm i in industry j in year t,
REV
ijt
= change in revenue for firm i in industry j between year t − 1 and t,
PPE
ijt
= gross property, plant and equipment for firm i in year t,
A
ijt−1

= total assets for firm i in industry j at the end of the previous year,
REC
ijt
= the change in receivables for firm i in industry j between year t − 1 and t.
The
ˆ
β
1
coefficient (change in revenues) is predicted to be positive, as changes in
revenues are expected to be positively related to changes in working capital accounts.
The expected sign on
ˆ
β
2
(property, plant and equipment) is negative, as the level
of fixed assets is expected to drive depreciation expenses and deferred taxes (Klein,
2002a).
13
Having estimated non-discretionary accruals (NDAC) from equation (1)
above, the amount of discretionary accruals (DAC) for firm i in industry j for year t
is calculated as the residual value from equation (3):
DAC
ij t
= TAC
ij t
− NDAC
ijt
. (3)
We use a cash-flow approach to estimate total accruals as this is considered superior
to the balance sheet approach (Hribar and Collins, 2002). This approach involves de-

ducting the cash flow from operations obtained from the statement of cash flows from
the amount of net income (before extraordinary items) from the income statement.
12
The adjustment for changes in receivables (i.e. modified-Jones model) is only applied to the
expectations model. To estimate the industry specific regression coefficients in equation (2),
the original Jones model is used (Dechow et al., 1995; Bartov et al., 2000).
13
All variables in the accruals expectations model (equation (2)) are scaled by lagged assets
to reduce heteroscedasticity, as it is assumed that lagged assets are positively associated with
the variance of the disturbance term (Jones, 1991).
C

AFAANZ, 2005
252 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
3.2.2. Corporate governance variables
In our primary models, board independence is measured by two dummy variables,
the first taking a value of 1 if the board of directors is comprised of a majority of
non-executive directors and 0 otherwise (BDIND), and the second variable taking a
value of 1 if the roles of the chairperson and CEO are separated and 0 otherwise
(INDCHAIR). We acknowledge that some non-executive directors are not entirely
independent from management as they might have a relationship with the firm in
addition to their role as a director. Therefore, we attempt to refine this measure
by eliminating those non-executive directors who are reported to have engaged in
transactions with the company.
14
Because inadequate disclosure by some compa-
nies results in a smaller sample, we test this measure on the subset of compa-
nies for which this information is available and report the results in our additional
analysis.
The existence of an audit committee is identified by a dummy variable with a value

of 1 if the firm has an audit committee operating during the year and 0 otherwise
(AC). As noted, our models differ in their measure of audit committee independence
(ACIND). In model 1, ACIND is a dummy variable taking the value of 1 when
the committee is comprised entirely of non-executive directors and 0 otherwise. In
model 2, ACIND takes the value of 1 when the committee is comprised of a majority
of non-executive directors and 0 otherwise.
Other indicators of audit committee effectiveness, activity and size, are measured
by the number of committee meetings held during the year (ACMEET), and the
number of directors assigned to the audit committee (ACSIZE), respectively. To test
the internal audit function hypothesis, a dummy variable is employed, taking a value
of 1 if the firm has its own internal audit function or it outsources its internal audit
activities, and 0 otherwise (IAF). Similarly, a dummy variable is used to test the
external audit hypothesis, with a value of 1 when the firm uses a Big 5 auditor and 0
otherwise (BIG5).
3.2.3. Control variables
We control for the effect of possible confounding factors (Bartov et al., 2000)
by including in our models variables that prior studies have found to be associated
with earnings management or governance variables. To control for the existence
of concentrated shareholdings that might improve monitoring (Agrawal and Knoe-
ber, 1996; Peasnell et al., 2000), we include a variable defined as the percentage
of shares held by the largest substantial shareholder (SUBSH).
15
Leverage (LEV),
14
This information is reported in the related party transactions footnote in the financial state-
ments.
15
The Corporations Act 2001 defines a substantial shareholder as one who has attained 5 per
cent or more of the total votes attached to the voting shares in the company.
C


AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 253
measured as the ratio of total liabilities to total assets, captures the incentives to
practice earnings management when close to debt covenant violations (Beasley and
Salterio, 2001; Klein, 2002a). The absolute change in earnings has been found to be
positively associated with earnings management (Klein, 2002a) and we measure this
by the absolute change in net income between the current and prior periods scaled
by total assets (ABSCH). We include the log of total assets (SIZE) to control for
the effect of size as this has been found to be negatively associated with earnings
management and positively associated with audit committee and board independence
and the use of internal audit (Bartov et al ., 2000; Klein, 2002a; Goodwin and Kent,
2003). Following Klein (2002a,b), absolute current earnings (ABSNI), measured by
the absolute value of net income before extraordinary items scaled by total assets, is
also included;
16
as is the market to book ratio (MKT), measured by market capital-
ization divided by the book value of shareholders’ equity. Klein (2002b) finds this
latter variable to be related to board and audit committee independence. Finally, to
alleviate the concern that the modified Jones model provides biased estimates of dis-
cretionary accruals when firms experience extreme earnings performance (Dechow
et al., 1995), we include a control variable for extreme performance (EXTP). This
variable takes a value of 1 if the firm is within either the top or bottom 10 per cent of
the sample for performance (measured by net income divided by total assets), and 0
otherwise.
3.3. Regression models
The following regression equation is adopted to test the hypotheses, with models
1 and 2 differing, as previously noted, only in their measure of audit committee
independence.
DAC = α + β

1
BDIND + β
2
INDCHAIR + β
3
AC + β
4
ACIND
+ β
5
ACMEET + β
6
ACSIZE + β
7
IAF + β
8
BIG5 + β
9
SUBSH
+ β
10
LEV + β
11
ABSCH + β
12
SIZE + β
13
ABSNI
+ β
14

MKT + β
15
EXTP + ε
(4)
where DAC is the absolute value of discretionary accruals, as measured by the
cross-sectional modified-Jones model and the remaining variables are as previously
defined.
16
Klein (2002a) also tests her model using signed earnings because both Kasznik (1999) and
Kothari et al. (2001) find earnings management is related to firm performance. We repeat our
test using signed net income and obtain qualitatively similar results.
C

AFAANZ, 2005
254 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
4. Results
4.1. Descriptive statistics
Table 2 provides the descriptive statistics for the variables in the model. Panel A
shows the financial variables used in the calculation of discretionary accruals, while
Panel B shows the descriptive statistics for the continuous variables in the regression
model. Panel C reports details of the dummy regression variables.
Panel A indicates that the average firm has reported total assets of $941.5m, net
income before extraordinary items of $49.8m and cash flow from operations of
$83.3m. In relation to the internal governance structures observed across the sample
firms, Panel B of the table indicates that the mean number of audit committee meetings
and the number of directors on the committee is approximately 2.5. Panel C shows
that 77 per cent of firms have a majority of non-executive directors on the board. The
same percentage of firms also maintain a division between the roles of the CEO and
chairperson. Some 83 per cent of firms report having an audit committee established,
with 64 per cent of those committees being comprised entirely of non-executive

directors and 86 per cent being comprised of a majority of non-executives. Panel C
also shows that 75 per cent of firms use a Big 5 audit firm while an internal audit
function is only present in 35 per cent of firms.
To assess the modified Jones model’s ability to discriminate between discretionary
and non-discretionary accruals, Table 3 provides a summary of the statistical prop-
erties of the model’s coefficients. The
ˆ
β
1
coefficient (change in revenues) is, on
average, positive as expected, and the coefficient on
ˆ
β
2
(property, plant and equip-
ment) is negative, as expected. Therefore, it appears that the model is well specified
and has produced plausible estimates for partitioning total accruals into their discre-
tionary and non-discretionary components (Bernard and Skinner, 1996). However,
the
ˆ
β
1
coefficient ranges from −0.753 to 0.576, with a standard deviation of 0.255,
and a majority of the observations are negative, despite the positive mean value. The
ˆ
β
2
coefficient is similarly dispersed, ranging from −0.715 to 0.912, and a standard
deviation of 0.294.
17

Table 4 reports the correlations between the variables in the regression model.
18
Total assets are correlated with a number of other variables. The audit committee
variables are also highly correlated. The highest correlation for the independent
variables is between the existence of an audit committee (AC) and the size of the
audit committee (ACSIZE), with a coefficient of 0.760. Multicollinearity is, therefore,
a possible concern and this is considered under additional analysis.
17
The large dispersions in the model’s coefficients highlight the imprecision of the estimates,
and as suggested by the published literature, the potential for the modified Jones model to
misclassify discretionary accruals (Bernard and Skinner, 1996).
18
SPSS calculates the exact correlation regardless of whether the variables are dummy or
continuous.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 255
Table 2
Descriptive statistics
Panel A: Financial variables
Variable Mean Standard Minimum Median Maximum
deviation
Total assets ($’000) 941 490 4278 816 630 90 930 65 585 000
Net income (before
extraordinary items
($’000)
49 798 269 579 −257 600 2896 4 043 000
Cash flow from operations
($’000)

83 298 455 193 −168 900 3600 6547
Total accruals ($’000)
a
0.078 1.643 −2.389 −0.029 28.476
Discretionary accruals
a
−0.070 0.224 −1.031 −0.066 1.061
Panel B: Continuous regression variables
Variable Mean Standard Minimum Median Maximum
deviation
Absolute discretionary
accruals (DAC)
a
0.156 0.176 0 0.094 1.061
Audit committee meetings
(ACMEET)
2.500 1.999 0 2 14
Audit committee size
(ACSIZE)
2.560 1.488 0 3 7
Substantial shareholding
(SUBSH)
0.264 0.196 0 0.196 0.970
Financial leverage (LEV)
b
0.448 0.315 0.002 0.466 4.278
Absolute change in net
income (ABSCH)
b
0.308 1.510 0 0.050 23.550

Log of total assets (SIZE) 11.508 2.055 6.446 11.417 17.999
Absolute net income
(ABSNI)
b
0.114 0.197 0.000 0.0610 2.820
Market to book ratio (MKT) 3.820 16.830 −17.160 1.485 287.980
Panel C: Dummy regression variables
Firms Percentage
Board of directors
Comprised of a majority of non-executive directors (BDIND) 334 77
Independent chairperson (INDCHAIR) 334 77
Audit committee
Established (AC) 360 83
Comprised entirely of non-executives (ACIND: model 1) 230 64
Comprised of a majority of non-executives (ACIND: model 2) 310 86
Internal audit function (IAF) 152 35
Big 5 auditor 326 75
Extreme earnings performance (EXTP)7918
a
Scaled by lagged total assets.
b
Scaled by total assets.
C

AFAANZ, 2005
256 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
Table 3
Descriptive statistics for estimated regression coefficients (n = 24)
TAC
ij t

/A
ij t−1
= α
j
[1/A
ij t−1
] + β
1j
[REV
ij t
/A
ij t−1
] + β
2j
[PPE
ij t
/A
ij t−1
] + ε
ij t
Cash-flow Mean Standard Minimum Median Maximum Percentage
approach
a
deviation positive
ˆα coefficient 512.644 9034.791 −28981.249 149.037 31910.344 58.33
t-statistic 2.394 9.090 −2.580 0.180 44.377
ˆ
β
1
coefficient 0.0316 0.255 −0.753 −0.008 0.576 45.83

t-statistic 0.138 1.785 −3.424 −0.116 4.238
ˆ
β
2
coefficient −0.079 0.294 −0.715 −0.060 0.912 20.83
t-statistic −1.040 1.249 −3.334 −1.102 1.016
R
2
(%) 36.46 31.19 0 28.36 99.95
Adjusted R
2
(%) 20.95 30.92 0 0 96.38
a
Cash-flow approach: TAC
t
= EBXI
t
− CFO
t
.Where:EBXI
t
= earnings before extraordinary items (net
of taxes) for period t and CFO
t
= cash flow from operations for period t. Discretionary accruals (DAC)
are estimated by subtracting the predicted level of non-discretionary accruals (NDAC) from total accruals,
as calculated under the cash-flow approach (standardized by lagged total assets). TAC
ijt
= total accruals
for firm i in industry j in year t; REV

ijt
= change in revenue for firm i in industry j between year t − 1
and t; PPE
ijt
= gross property, plant and equipment for firm i in year t; A
ijt −1
= total assets for firm i in
industry j at the end of the previous year; n = number of industries.
4.2. Multivariate analysis
Table 5 reports the results of the two regression models. Both models are significant,
with an adjusted R
2
of 0.237 and 0.239, respectively.
Recall that the ability of the board of directors to monitor the financial discretion
of management is expected to be a function of its independence, measured by board
and chairperson independence. The coefficient on the first measure (BOARDIND)is
negative and significant at p = 0.017 in model 1 and p = 0.024 in model 2. This
provides support for Hypothesis 1a that earnings management is negatively associated
with board independence. However, Hypothesis 1b is not supported as there is no
evidence of a negative relationship between an independent board chair (INDCHAIR)
and earnings management in either model. This result might be attributable to the
limited additional oversight provided by a non-executive chairperson when the board
itself is predominantly independent from management.
The second hypothesis relates to the audit committee and its role in detecting
and preventing earnings management. In model 1, we do not find any association
between the level of discretionary accruals and the existence of an audit committee
(AC), the committee’s independence (ACIND) or its effectiveness (ACMEET and
ACSIZE). Model 2, however, finds a negative association (p = 0.077) between a
reduction in earnings management and a majority of non-executiveson the committee.
These results are consistent with those of Klein (2002a) who finds that a majority

of independent directors on the audit committee is related to a reduction in earnings
management, but that an entirely independent committee has no meaningful relation
with earnings management.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 257
Table 4
Pearson correlation matrix of independent variables (p-value, two-tailed)
∗
BOARDIND INDCHAIR AC ACIND ACIND ACMEET ACSIZE IAF BIG5 SUBSH LEV ABSCH SIZE ABSNI MKT
model 1 model 2
BOARDIND 1.00
(−)
INDCHAIR 0.378 1.00
(0.000) (−)
AC 0.176 0.159 1.00
(0.000) (0.001) (−)
ACIND model 1 0.284 0.332 0.472 1.00
(0.000) (0.000) (0.000) (−)
ACIND model 2 0.342 0.350 0.685 0.688 1.00
(0.000) (0.000) (0.000) (0.000) (−)
ACMEET 0.144 0.175 0.555 0.412 0.439 1.00
(0.003) (0.000) (0.000) (0.000) (0.000) (−)
ACSIZE 0.220 0.192 0.760 0.253 0.612 0.494 1.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (−)
IAF 0.118 0.087 0.277 0.283 0.312 0.296 0.313 1.00
(0.014) (0.068) (0.000) (0.000) (0.000) (0.000) (.000) (−)
BIG5 0.138 0.093 0.228 0.166 0.243 0.216 0.243 0.223 1.00
(0.002) (0.052) (0.000) (0.001) (0.000) (0.000) (0.000) (0.000) (−)

SUBSH −0.083 −0.078 0.153 −0.048 0.061 0.123 0.161 0.114 0.044 1.00
(0.084) (0.104) (0.002) (0.315) (0.204) (0.011) (0.001) (0.017) (0.357) (−)
C

AFAANZ, 2005
258 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
Table 4 (cont’d)
BOARDIND INDCHAIR AC ACIND ACIND ACMEET ACSIZE IAF BIG5 SUBSH LEV ABSCH SIZE ABSNI MKT
model 1 model 2
LEV 0.023 −0.111 0.189 0.058 0.077 0.223 0.262 0.100 0.098 0.106 1.00
(0.629) (0.020) (0.000) (0.223) (0.106) (0.000) (0.000) (0.037) (0.041) (0.027) (−)
ABSCH −0.128 −0.105 −0.161 −0.078 0.103 −0.085 −0.109 −0.098 −0.055 −0.050 0.316 1.00
(0.007) (0.028) (0.001) (0.102) (0.030) (0.077) (0.023) (0.041) (0.250) 0.294 0.000 (−)
SIZE 0.159 0.173 0.451 0.376 0.439 0.524 0.531 0.461 0.390 0.117 0.278 −0.078 1.00
(0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.015) (0.000) (0.103) (−)
ABSNI −0.061 −0.172 −0.167 −0.142 −0.198 −0.153 −0.188 −0.119 −0.173 −0.044 0.143 0.121 −0.349 1.00
(0.202) (0.000) (0.000) (0.003) (0.000) (0.001) (0.000) (0.013) (0.000) 0.360 0.003 (0.011) 0.000 (−)
MKT 0.039 −0.097 −0.073 −0.057 −0.093 −0.065 −0.191 −0.040 −0.025 0.003 0.135 0.016 −0.121 −0.015 1.00
(0.419) (0.044) (0.128) (0.237) (0.050) (0.178) (0.059) (0.404) 0.598 0.945 0.005 (0.747) 0.011 0.762 (−)
EXTP −0.010 −0.078 −0.077 −0.114 −0.145 −0.111 −0.126 −0.095 −0.085 0.008 0.123 0.151 −0.328 0.500 0.135
(0.827) (0.103) (0.110) (0.017) (0.002) (0.020) (0.009) (0.046) (0.075) (0.869) (0.010) (0.002) (0.000) 0.000 0.005
Pearson correlations are adjusted automatically by SPSS when variables are dichotomous.
BOARDIND (board independence) = a dummy variable with a value of 1 if the board is comprised of a majority of non-executive directors, and 0 otherwise.
INDCHAIR (chairperson independence) = a dummy variable with a value of 1 if the roles of the chairperson and Chief Executive Officer are separated, and 0
otherwise. AC (audit committee) = a dummy variable with a value of 1 if the firm has an audit committee and 0 otherwise. ACIND (model 1) (audit committee
independence) = a dummy variable with a value of 1 if the audit committee is comprised solely of non-executive directors, and 0 otherwise. ACIND (model
2) (audit committee independence) = a dummy variable with a value of 1 if the audit committee is comprised of a majority of non-executive directors, and 0
otherwise. ACMEET (audit committee meetings) = the number of committee meetings held during the year. ACSIZE (audit committee size) = the number of
directors assigned to the audit committee. IAF (internal audit function) = a dummy variable with a value of 1 if the firm has an internal audit function operating
during the financial year, and 0 otherwise. BIG5 (Big 5 auditor) = a dummy variable with a value of 1 if the firm has a Big 5 audit partner, and 0 otherwise.

SUBSH (substantial shareholder) = the percentage of shares held by the largest ‘substantial shareholder’.LEV(financial leverage) = total liabilities divided
by total assets. ABSCH (absolute change in net income) = the absolute change in net income between t and t − 1 divided by total assets. SIZE (firm size) =
natural log of total assets. ABSNI (absolute net income) = the absolute net income before extraordinary items divided by total assets. MKT (market to book
ratio) = the market value of shareholders’ equity divided by the book value of equity. EXTPERF (extreme earnings performance) = a dummy variable with a
value of 1 if the firm is within the top 10 per cent or bottom 10 per cent of the sample for the performance measure (net income divided by total assets), and 0 otherwise.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 259
Table 5
Regression results
DAC = α + β
1
BDIND + β
2
INDCHAIR + β
3
AC + β
4
ACIND + β
5
ACMEET + β
6
ACSIZE + β
7
IAF +
β
8
BIG5 + β
9

SUBSH + β
10
LEV + β
11
ABSCH + β
12
SIZE + β
13
ABSNI + β
14
MKT + β
15
EXTP + ε
Variable Prediction Model 1 Model 2
Coefficient t-statistic p-value
∗
Coefficient t-statistic p-value
a
Intercept 0.168 2.853 0.005 0.170 2.915 0.004
Internal governance variables
BOARDIND 1a (−) −0.042 −2.115 0.017 −0.039 −1.972 0.024
INDCHAIR 1b (−)0.028 1.408 0.160 0.030 1.497 0.135
AC 2(−)0.055 1.495 0.136 0.062 1.685 0.093
ACIND 2(−) −0.019 −0.992 0.161 −0.036 −1.428 0.077
ACMEET 2(−) −0.005 −1.000 0.159 −0.005 −1.136 0.128
ACSIZE 2(−)0.009 1.077 0.282 0.013 1.622 0.106
IAF 3(−) −0.017 −0.982 0.163 −0.017 −0.972 0.331
BIG5 4(−) −0.006 −0.338 0.368 0.008 0.431 0.666
Control variables
SUBSH (−) −0.041 −1.078 0.141 −0.039 −1.014 0.311

ABSCH (+)0.040 6.931 0.000 0.040 6.976 0.000
LEV (+)0.096 2.843 0.002 0.095 2.789 0.003
SIZE (−) −0.011 −1.872 0.031 −0.011 −1.979 0.024
ABSNI (+)0.224 4.851 0.000 0.223 4.834 0.000
MKT (?) 0.001 1.791 0.074 0.001 1.761 0.079
EXTPERF (?) 0.017 0.739 0.461 0.016 0.687 0.493
Adjusted R20.237 0.239
F statistic 9.892 9.987
p-values (0.000) (0.000)
a
p-values are one-tailed when direction is as predicted, otherwise two-tailed.
BOARDIND (board independence) = a dummy variable with a value of 1 if the board is comprised
of a majority of non-executive directors, and 0 otherwise. INDCHAIR (chairperson independence) =
a dummy variable with a value of 1 if the roles of the chairperson and Chief Executive Officer are
separated, and 0 otherwise. AC (audit committee) = a dummy variable with a value of 1 if the firm has
an audit committee and 0 otherwise. ACIND (model 1) (audit committee independence) = a dummy
variable with a value of 1 if the audit committee is comprised solely of non-executive directors, and 0
otherwise. ACIND (model 2) (audit committee independence) = a dummy variable with a value of 1 if
the audit committee is comprised of a majority of non-executive directors, and 0 otherwise. ACMEET
(audit committee meetings) = the number of committee meetings held during the year. ACSIZE (audit
committee size) = the number of directors assigned to the audit committee. IAF (internal audit function) =
a dummy variable with a value of 1 if the firm has an internal audit function operating during the financial
year, and 0 otherwise. BIG5 (Big 5 auditor) = a dummy variable with a value of 1 if the firm has a Big
5 audit partner, and 0 otherwise. SUBSH (substantial shareholder) = the percentage of shares held by
the largest ‘substantial shareholder’.LEV(financial leverage) = total liabilities divided by total assets.
ABSCH (absolute change in net income) = the absolute change in net income between t and t − 1
divided by total assets. SIZE (firm size) = natural log of total assets. ABSNI (absolute net income) = the
absolute net income before extraordinary items divided by total assets. MKT (market to book ratio) =
the market value of shareholders’ equity divided by the book value of equity. EXTPERF (extreme
earnings performance) = a dummy variable with a value of 1 if the firm is within the top 10 per cent or

bottom 10 per cent of the sample for the performance measure (net income divided by total assets), and 0
otherwise.
C

AFAANZ, 2005
260 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
We do notfind an association between earnings managementand the existence of an
internal audit function (IAF) or the use of a Big 5 auditor in either model. Although
both coefficients are negative as predicted, neither is statistically significant and,
therefore, we do not find support for Hypothesis 3 or Hypothesis 4.
Of the seven control variables included in the models, five are statistically sig-
nificant. Earnings management is strongly associated in the predicted direction with
the absolute change in net income (ABSCH), absolute net income (ABSNI)
19
and
leverage (LEV). It is also negatively associated with firm size (SIZE) and positively
associated with market to book ratio of equity (MKT). However, we do not find
an association between the proportion of shares held by a substantial shareholder
(SUBSH) or extreme performance (EXTP).
4.3. Additional analysis
4.3.1. Additional absolute discretionary accruals analysis
Several additional tests are performed to ascertain the credibility of the initial anal-
ysis. The first set of tests, comprising three iterations, repeats the regression models
with alternative definitions of board and audit committee independence. Iteration 1
uses the proportion of non-executive directors on the board as the measure of in-
dependence. The coefficient is negative and significant (p < 0.100) in both models,
supporting the relationship between board independence and earnings management.
Iteration 2 examines audit committee independence in the same manner as iteration
1, with substantively similar results to those reported. Iteration 3 attempts to refine
the measure of independence by removing those non-executive directors with related

party transactions. As a result of inadequate disclosures, our sample is reduced to 420
firms. Our results are qualitatively similar to those reported for model 1, with board
independence significant at p = 0.050. However, the measure of audit committee
independence for model 2 (a majority of independent directors on the committee) is
not significant in this iteration (p = 0.131).
We also examine the sensitivity of audit committee meetings (ACMEET) and audit
committee size (ACSIZE) to specific cut-offs. First, we substitute ACMEET with a
dummy variable taking the value of 1 if the audit committee met at least twice during
the financial year, and 0 otherwise. We then repeat the analysis with three meetings
per year as the cut-off. The coefficients are consistent with the original models. We
also apply a similar dummy variable to ACSIZE, taking a value of 1 if the audit
committee is comprised of at least three directors and 0 otherwise. However, the
results are not sensitive to this cut-off point.
Another set of tests addresses concerns relating to the high correlation coefficients
between the audit committee variables. First, we re-perform the regression analyses
19
Qualitatively similar results are achieved when absolute change in net income and absolute
net income are scaled by beginning total assets.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 261
without the audit committee (AC) and audit committee meetings (ACMEET)vari-
ables. The revised models produce results qualitatively similar to the initial findings.
We also perform a factor analysis on the audit committee variables, replacing these
variables with the factor score in the regression models. The factor score is not sig-
nificant while the results for the remaining variables are substantively similar to the
primary models. These additional tests suggest that multicollinearity is not the basis
for the non-significant results reported in the primary analysis.
We also test for the possibility of interactions between audit committee indepen-

dence, audit committee size and the number of meetings by adding interaction vari-
ables to our primary models. However, none of the interaction effects are significant
and the models’ overall predictions remain substantively similar.
Table 4 shows that firm size (SIZE) is correlated with most of the other variables.
Specifically, consistent with Goodwin and Kent (2003), the correlation coefficient
between firm size and internal audit function (IAF)is0.461(p-value 0.000). This
correlation might confound the relationship between internal audit and earnings
management. To address this issue, and to avoid specification error by removing the
control for firm size, we re-test the regression models with only the sample firms
within the Top 500 ASX listed firms (by market capitalization). It is likely that
there is a threshold at which the perceived benefits of establishing an internal audit
function outweigh the costs of implementation. If firm size is an important factor
in determining this threshold, then the Top 500 classification might control for size,
without influencing the relationship between the internal audit function and earnings
management. The results of the regressions are consistent with the original models.
Therefore, a statistically significant relationship cannot be established between the
presence of an internal audit function and earnings management.
Final tests of robustness pertain to industry classification. To address the concern
raised in Wells (2002) that the estimation of discretionary accruals for ‘Gold’ firms
is potentially biased, because of the nature of their operations, the main regression
models are re-estimated in two ways. The first involves including a dummy variable,
taking the value of one if the firm is classified in the GICS subindustry ‘Gold’, and
zero otherwise. The second approach involves eliminating all Gold firms from the
sample. The findings from both approaches are qualitatively similar to the original
results.
4.3.2. Additional tests of earnings management
Our analysis focuses on absolute discretionary accruals as a proxy for earn-
ings management. This measure is non-directional, capturing both upwards and
downwards earnings management. Therefore, we conduct additional tests that fo-
cus only on upwards earnings management. First, we re-run our prior tests using

signed discretionary accruals as the dependent variable. Although we obtain qual-
itatively similar results for the control variables, none of the corporate governance
variables are statistically significant.
C

AFAANZ, 2005
262 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
Next, we explore the possibility that firms reporting small increases in earnings or
small profits might have achieved their results by engaging in earnings management.
We plot histograms of earnings and changes in earnings, following Holland and Ram-
say (2003). Although we do not find any discontinuity with regard to small profits
or losses,
20
we do find discontinuity in the distribution of small changes in earnings.
The results of a logistic regression with the dependent variable set at 1 for increases
in earnings (scaled by net assets) of less than 0.02 and 0 otherwise are reported in
Table 6. This table shows that firm size is positively associated while extreme per-
formance is negatively associated with small increases in earnings. Of interest to
the present study is the negative relation between small changes in earnings and
the existence of an audit committee. However, neither board independence nor audit
committee independence is associated with small increases in earnings. Although
these results should be interpreted with caution, they might suggest that earnings
management not captured by the discretionary accruals model is constrained by the
presence of an audit committee.
5. Conclusion
The significant negative relationship between earnings management and the pres-
ence of a board comprised of a majority of non-executive directors provides support
for Hypothesis 1a.
21
The findings do not support Hypothesis 1b concerning an asso-

ciation between an independent chairperson and earnings management. With respect
to Hypothesis 2, which investigates the relation between the existence of an audit
committee and specific audit committee characteristics and earnings management,
none of the proposed variables are supported by model 1, where audit committee in-
dependence is measured as all committee members being non-executives. However,
model 2 does find support for an association between an audit committee comprising
a majority of non-executives and a reduction in earnings management.
22
Our addi-
tional analysis, using small increases in earnings as a possible measure of earnings
management, finds a relation between the existence of an audit committee and the
dependent variable. Neither Hypothesis 3 relating to the internal audit function nor
Hypothesis 4 relating to the choice of a Big 5 auditor are supported in our primary
model or any of our additional testing.
The results of the present study have implications for regulators who are con-
cerned to minimize opportunities for earnings management and to improve financial
reporting quality. There are also implications for companies seeking to strengthen
20
Consequentially, logistic regressions coded 1 for firms reporting very small profits and 0
otherwise, not surprisingly fail to produce significant results at any of the accepted cut-off
points for small profits (see Holland and Ramsay, 2003).
21
This finding is consistent with the results of Peasnell et al. (2000), using data from the UK
and Klein (2002a) using US data.
22
This result is also consistent with Klein (2002a).
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 263

Table 6
Logit regression results (dependent variable = small increases in earnings)
Variable Prediction Coefficient Wald statistic p-value
a
Intercept 4.041 0.000
Internal governance variables
BOARDIND 1a (−) −0.154 0.156 0.346
INDCHAIR 1b (−) −0.238 0.383 0.268
AC 2(−) −1.528 4.555 0.016
ACIND 2(−)0.398 1.092 0.296
ACMEET 2(−) −0.016 0.037 0.424
ACSIZE 2(−)0.156 1.007 0.316
IAF 3(−) −0.065 0.040 0.421
BIG5 4(−) −0.398 1.119 0.145
Control Variables
SUBSH (−)0.735 1.091 0.296
LEV (+)0.743 1.130 0.144
SIZE (−)0.345 11.316 0.001
MKT (?) −0.043 0.552 0.457
EXTPERF (?) −1.637 4.605 0.032
Pseudo R
2
0.175
χ-square 47.085
p-value (0.000)
a
p-values are one-tailed when direction is as predicted, otherwise two-tailed.
BOARDIND (board independence) = a dummy variable with a value of 1 if the board is comprised
of a majority of non-executive directors, and 0 otherwise. INDCHAIR (chairperson independence) =
a dummy variable with a value of 1 if the roles of the chairperson and Chief Executive Officer are

separated, and 0 otherwise. AC (audit committee) = a dummy variable with a value of 1 if the firm has
an audit committee and 0 otherwise. ACIND (model 1) (audit committee independence) = a dummy
variable with a value of 1 if the audit committee is comprised solely of non-executive directors, and 0
otherwise. ACIND (model 2) (audit committee independence) = a dummy variable with a value of 1 if
the audit committee is comprised of a majority of non-executive directors, and 0 otherwise. ACMEET
(audit committee meetings) = the number of committee meetings held during the year. ACSIZE (audit
committee size) = the number of directors assigned to the audit committee. IAF (internal audit function)
= a dummy variable with a value of 1 if the firm has an internal audit function operating during the
financial year, and 0 otherwise. BIG5 (Big 5 auditor) = a dummy variable with a value of 1 if the firm
has a Big 5 audit partner, and 0 otherwise. SUBSH (substantial shareholder) = the percentage of shares
held by the largest ‘substantial shareholder’.LEV(financial leverage) = total liabilities divided by total
assets. ABSCH (absolute change in net income) = the absolute change in net income between t and t − 1
divided by total assets. SIZE (firm size) = natural log of total assets. MKT (market to book ratio) = the
market value of shareholders’ equity divided by the book value of equity. EXTPERF (extreme earnings
performance) = a dummy variable with a value of 1 if the firm is within the top 10 per cent or bottom
10 per cent of the sample for the performance measure (net income divided by total assets), and 0 otherwise.
their corporate governance with respect to financial reporting. In particular, there
would appear to be scope for strengthening the effectiveness of audit committees and
making greater use of internal audit as mechanisms to reduce earnings management.
The results and implications of the present study are subject to several limitations,
most of which suggest the need for further research. Problems associated with the use
C

AFAANZ, 2005
264 R. Davidson et al. / Accounting and Finance 45 (2005) 241–267
of discretionary accruals as a measure of earnings management are well documented
(Klein, 2002a). The fact that our alternative measure, small increases in earnings,
gives differing results might indicate that our measures are not good proxies for
earnings management. A second limitation relates to the measures employed to test
the strength of the internal governance mechanisms. A more refined classification of

director independence might provide a more informative measure of board monitoring
ability. Similarly, the present study is limited by the lack of suitable proxies to
operationalize audit committee effectiveness. The proxies adopted in the present
study, particularly audit committee size and the frequency of committee meetings,
might not be powerful enough measures of audit committee effectiveness. Other
characteristics such as the financial expertise of members, the existence of an audit
committee charter or the number of meetings with the external auditor might be
better measures of effectiveness. The internal audit measure also has limitations as we
focus only on whether a company uses internal audit. A more refined measure could
examine the status of internal audit within the firm, the role that the function plays
in corporate governance and its interaction with the audit committee. While more
detailed internal audit information would need to be obtained by survey, refinements
to the other governance measures are likely to become more readily available as a
result of CLERP 9 and the ASX best practice recommendations.
References
Agrawal, A., and C. R. Knoeber, 1996, Firm performance and mechanisms to control agency
problems between managers and shareholders, The Journal of Financial and Quantitative
Analysis 31, 377–397.
Arthur, N., and S. Taylor, 1993, Chairman, chief executive, or both? Company Director 9,
20–21.
AIMA (Australian Investment Managers’ Association), 1995, Corporate Governance: A Guide
for Investment Managers and a Statement of Recommended Corporate Practice (AIMA,
Sydney).
AIMA (Australian Investment Managers’ Association), 1997, Corporate Governance State-
ments by Major ASX Listed Companies (AIMA, Sydney).
ASX (Australian Stock Exchange), 2003, Corporate Governance Council, Principles of Good
Corporate Governance and Best Practice Recommendations (ASX, Sydney).
Barnhart, S. W., W. Marr, and S. Rosenstein, 1994, Firm performance and board composition:
some new evidence, Managerial and Decision Economics 15, 329–340.
Bartov, E., F. A. Gul, and J. S. L. Tsui, 2000, Discretionary-accruals models and audit qualifi-

cations, Journal of Accounting and Economics 30, 421–452.
Baysinger, B., and H. Butler, 1985, Corporate governance and the board of directors: perfor-
mance effects of changes in board composition, Journal of Law, Economics, and Organiza-
tion 1, 101–124.
Beasley, M. S., 1996, An empirical analysis of the relation between the board of director
composition and financial statement fraud, The Accounting Review 71, 443–465.
Beasley, M. S., and S. Salterio, 2001, The relationship between board characteristics and
voluntary improvements in audit committee compensation and experience, Contemporary
Accounting Research 18, 539–570.
C

AFAANZ, 2005
R. Davidson et al. / Accounting and Finance 45 (2005) 241–267 265
Becker, C., M. DeFond, J. Jiambalvo, and K. R. Subramanyam, 1998, The effect of audit on
the quality of earnings management, Contemporary Accounting Research 15, 1–24.
Bernard, V. L., and D. J. Skinner, 1996, What motivates managers’ choice of discretionary
accruals? Journal of Accounting and Economics 22, 313–325.
BRC (Blue Ribbon Committee on Improving the Effectiveness of Corporate Audit Commit-
tees), 1999, Report and Recommendations of the Blue Ribbon Committee on Improving the
Effectiveness of Corporate Audit Committees (NYSE, New York).
Bosch Committee (Working Group of the Australian Institute of Company Directors and Others
(Henry Bosch AO, chair)), 1995, Corporate Practices and Conduct, 3rd edn (FT Pitman
Publishing, Melbourne).
Cadbury Committee (committee on the financial aspects of corporate governance (Sir Adrian
Cadbury, chair)), 1992, Report (Gee and Company Ltd, London).
Cadbury, A., 1997, Board Focus: the Governance Debate (Egon Zehnder International).
CIMA (Chartered Institute of Management Accountants), 2000, Corporate Governance: His-
tory, Practice and Future (CIMA, London).
Chtourou, S. M., J., B
´

edard, and L. Courteau, 2001, Corporate governance and earnings
management, Working paper (University of Laval, Quebec).
Clikeman, P. M., 2003, Where auditors fear to tread, International Auditor 60, 75–79.
Cohen, J., G. Krishnamoorthy, and A. Wright, 2002, Corporate governance and the audit
process, Contemporary Accounting Research 19, 573–592.
Collier, P. A., 1993, Audit committees in major U.K. companies, Managerial Auditing Journal
8, 25–30.
Commonwealth of Australia, 2004, Corporate Law Economic Reform Program (CLERP 9)
(Audit Reform and Corporate Disclosure) Act 2004 (Australian Government Publishing
Service, Canberra).
DeAngelo, L. E., 1981, Auditor independence, ‘low-balling’ and disclosure regulation, Journal
of Accounting and Economics 3, 113–127.
Dechow, P. M., R. G. Sloan, and A. P. Sweeney, 1995, Detecting earnings management, The
Accounting Review 70, 193–225.
Dechow, P. M., R. G. Sloan, and A. P. Sweeney, 1996, Causes and consequences of earn-
ings manipulation: an analysis of firms subject to enforcement by the SEC, Contemporary
Accounting Research 13, 1–36.
DeFond, M. L., and J. Jiambalvo, 1991, Incidence and circumstances of accounting errors, The
Accounting Review 66, 643–655.
DeFond, M. L., and J. Jiambalvo, 1994, Debt covenant violation and manipulation of accruals,
Journal of Accounting and Economics 17, 145–176.
Eighme, J. E., and J. D. Cashell, 2002, Internal auditors’ roles in overcoming the financial
reporting crisis, International Auditing 17, 3–10.
Fama, E. F., and M. Jensen, 1983a, Agency problems and residual claims, Journal of Law and
Economics 26, 327–349.
Fama, E. F., and M. Jensen, 1983b, Separation of ownership and control, Journal of Law and
Economics 26, 301–325.
Fields, T. D., T. Z. Lys, and L. Vincent, 2001, Empirical research on accounting choice, Journal
of Accounting and Economics 31, 255–307.
Francis, J. R., and J. Krishnan, Accounting accruals and auditor reporting conservatism,

Contemporary Accounting Research 16, 135–165.
Francis, J. R., E. L. Maydew, and H. C. Sparks, 1999, The role of big 6 auditors in the credible
reporting of accruals, Auditing: a Journal of Practice and Theory 18, 17–34.
Gibbins, M., S. Salterio, and A. Webb, 2001, Evidence about auditor-client management
negotiation concerning client’s financial reporting, Journal of Accounting Research 39,
535–563.
C

AFAANZ, 2005