Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

Contents lists available at SciVerse ScienceDirect

Advances in Accounting, incorporating Advances in

International Accounting
journal homepage: www.elsevier.com/locate/adiac

Audit committee nancial expertise and properties of analyst earnings forecasts

John L. Abernathy a,, Don Herrmann a, 1, Tony Kang a, 2, Gopal V. Krishnan b, 3
a
b

Spears School of Business, Oklahoma State University, Stillwater, OK 74078, United States
Kogod School of Business, American University, Washington, DC 20016, United States

a r t i c l e

i n f o

Keywords:
Audit committees
Financial expertise
Analyst earnings forecasts

a b s t r a c t
An important role of nancial accounting information is to aid nancial statement users in forming expectations about the rm's future earnings. Prior research nds that accounting nancial expertise of the audit
committee is associated with higher nancial reporting quality. We extend this literature by examining the
association between audit committee nancial expertise and analysts' ability to anticipate future earnings.
We nd a signicant association between accounting nancial expertise on the audit committee and analyst
earnings forecasts that are more accurate and less dispersed. In contrast, we do not nd a signicant association between non-accounting nancial expertise (i.e., supervisory expertise) and forecast accuracy or forecast dispersion. These ndings contribute to our understanding of the benets of accounting expertise in
audit committees by demonstrating an association between accounting nancial expertise and improvements in analyst earnings forecasts.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Prior research provides evidence that having accounting nancial
experts on the audit committee is associated with higher nancial
reporting quality (e.g., DeFond, Hann, & Hu, 2005; Dhaliwal, Naiker, &
Navissi, 2010; Krishnan & Visvanathan, 2008). We extend this research
to examine whether improved nancial reporting quality from such expertise is associated with the ability to anticipate future earnings. The
Financial Accounting Standard Board (FASB) (1978) notes in the Statement of Financial Accounting Concepts No. 1 that nancial reporting
should provide information that is useful to present and potential investors and creditors and other users in making rational investment, credit
and similar decisions. While prior studies suggest that audit committee
expertise is associated with improved nancial reporting quality, there
is little empirical evidence on the association between audit committee's
nancial expertise and decisions of nancial statement users. We

We thank Wayne Thomas and seminar participants at Chulalongkorn University,

Oklahoma State University, Singapore Management University and the 2012 American Accounting Association Annual meeting for helpful comments and suggestions. Dr. Krishnan
gratefully appreciates the nancial assistance from the Joseph R. Perella and Amy M. Perella
Professorship.
Corresponding author. Tel.: +1 405 744 6349; fax: +1 405 744 1680.
E-mail addresses: john.abernathy@okstate.edu (J.L. Abernathy), don@okstate.edu
(D. Herrmann), tony.kang@okstate.edu (T. Kang), krishnan@american.edu (G.V. Krishnan).
1
Tel.: +1 405 744 8602; fax: +1 405 744 1680.
2
Tel.: +1 405 744 8631; fax: +1 405 744 1680.
3
Tel.: +1 202 885 6460; fax: +1 202 885 1992.
0882-6110/$ see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.adiac.2012.12.001

examine an important aspect of users' investment decisions, i.e., their

ability to anticipate future earnings. Specically, we investigate the
association between audit committee nancial expertise and nancial
analysts' ability to predict future earnings.
We focus on analyst earnings forecasts for several reasons. First,
Schipper (1991) suggests that analyst behavior can provide insight into
the activities and beliefs of investors that cannot be observed directly.
Analysts, as sophisticated users of nancial reporting information, provide direct evidence about whether users incorporate improvements in
nancial reporting into their decision-making process. Second, as
Kothari (2001) states, almost all models of valuation either directly or
indirectly use earnings forecasts. Thus, given analysts' role as a key
provider of information to the capital markets, empirical evidence on
whether audit committee nancial expertise relates to analysts' forecasting ability is useful to investors in rm valuation. Kecskes, Michaely,
and Womack (2010) nd evidence that analysts' earnings-based recommendation changes are more informative than discount rate-based
recommendations, supporting the notion that the value of analysts' recommendations is primarily linked to how well they can discern a rm's
future earnings. Finally, while prior research has documented a link between accounting expertise on the audit committee and nancial
reporting quality, prior research has not explored a direct link with
users of nancial statement information. We address this gap in the literature in examining the association between audit committee nancial
expertise and analysts' ability to anticipate future earnings.
The SarbanesOxley Act of 2002 required the Securities Exchange
Commission (SEC) to issue rules mandating that the audit committee
of every public company have a designated nancial expert; and that
the name of that nancial expert be disclosed (SarbanesOxley Act of
2002). The SEC suggests that having at least one nancial expert on

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

the audit committee should improve the quality of information available to investors. Prior research supports this notion by showing that
the nancial expertise of the audit committee is signicantly associated
with a lower incidence of nancial statement restatement (Abbott,
Parker, & Peters, 2004), a reduced likelihood of material weaknesses in
internal control reported during an auditor change (Krishnan, 2005), a
reduction in fraud (Farber, 2005), and lower expected rates of return
on pension plan assets (Comprix, Guo, Zhang, & Zhou, 2012).
The SEC initially proposed a stringent denition of nancial expert,
which dened individuals as nancial experts only if they had education
and experience in accounting or auditing (i.e. as a certied public accountant, auditor, chief nancial ofcer, nancial controller or accounting ofcer). In response to criticism that this denition was overly restrictive,
the SEC adopted a broader denition of audit committee nancial expert.
Specically, an audit committee member could be deemed a nancial expert if the member has had work experience in accounting or auditing, as
well as work experience in nance positions or as a chief executive ofcer (CEO) or company president. Therefore, nancial expertise may include expertise in accounting, nance, or in supervising the preparation
of nancial statements (supervisory expertise). However, many argue
that the current denition of nancial expertise may be too broad and
lack the ability to ensure high nancial reporting quality.
Consistent with this, prior research nds that the presence of accounting nancial expertise (but not non-accounting nancial expertise)
on the audit committee is associated with certain nancial reporting
characteristics such as greater accounting conservatism (Krishnan &
Visvanathan, 2008), higher quality accruals (Dhaliwal et al., 2010), and
lower earnings management (Bdard, Chtourou, & Courteau, 2004;
Carcello, Hollingsworth, & Neal, 2006). The accounting nancial expertise of the audit committee is also associated with a reduction in suspicious auditor switches (Archambeault & DeZoort, 2001) and higher
rm credit ratings (Ashbaugh-Skaife, Collins, & LaFond, 2006). Prior research also suggests that investors care about the accounting nancial
expertise of audit committee members. For example, DeFond et al.
(2005) nd that companies appointing audit committee members with
accounting nancial expertise experience signicant positive abnormal
market returns, while no market reaction is observed upon the appointment of those with non-accounting nancial expertise.
Financial analysts use accounting information to form expectations of
future earnings (e.g., Abarbanell & Bushee, 1997). Furthermore, survey
evidence suggests that audit committee nancial expertise matters to nancial analysts. Dickins, Hillison, and Platau (2009) survey nancial analysts and nd that analysts are more condent in the nancial statements
when the Audit Committee Financial Expert (ACFE) has accounting-based
nancial expertise. However, there is little evidence on how nancial analyst earnings forecasts vary with audit committee nancial expertise.
Thus, if audit committee accounting expertise increases both the quality
of nancial information which nancial analysts use to formulate their
forecasts and analyst condence in the nancial information provided,
we expect the properties of analysts' earnings forecasts to improve with
audit committee accounting expertise.
To address our research question, we examine the associations between audit committee nancial expertise and analyst earnings forecast
properties (i.e., accuracy and dispersion). Financial analysts are viewed
as sophisticated nancial statement users and their earnings forecasts
are commonly used as a proxy for the market's expectation of earnings,
which is a critical element in rm valuation. We nd that the accounting
nancial expertise of the audit committee is signicantly associated with
greater analyst forecast accuracy and lower forecast dispersion. In contrast, examining the broad denition of nancial expertise adopted by
the SEC, we nd that non-accounting nancial expertise is not signicantly associated with either improved analyst forecast accuracy or
lower analyst forecast dispersion.
This study contributes to the literature in the following ways. First,
building on prior studies that examine the relation between audit committee expertise and nancial reporting quality, we examine whether

audit committee expertise is associated with improved analyst forecasts.

Our ndings extend the literature by showing a link between accounting
expertise on the audit committee and forecasts of future earnings.
Our results also contribute to the debate on the denition of nancial
expertise on an audit committee. Our evidence suggests that accounting
nancial expertise (but not non-accounting nancial expertise) enhances
nancial analysts' ability to anticipate future earnings. In line with prior
research (Archambeault & DeZoort, 2001; Ashbaugh-Skaife et al.,
2006; Bdard et al., 2004; Carcello et al., 2006; Dhaliwal et al.,
2010; Krishnan & Visvanathan, 2008) these results suggest that accounting specic nancial expertise on the audit committee is especially benecial.
The paper proceeds as follows. Section 2 reviews prior literature and
develops the hypotheses relating audit committee nancial expertise to
the properties of analysts' forecasts. Section 3 describes the research
methodology. Section 4 presents the empirical results and Section 5
concludes the paper.
2. Prior research and hypotheses development
2.1. Audit committees and nancial expertise
Prior studies using the broad denition of nancial expertise have
provided mixed evidence about an association between nancial expertise and nancial reporting quality. Abbott et al. (2004), and Agrawal
and Chadha (2005) nd that nancial expertise of the audit committee
(under a broad denition) is negatively related to the occurrence of restatement. Farber (2005) also employs the broad denition of nancial
expertise and nds a signicantly lower occurrence of nancial fraud
in rms with nancial expertise on the audit committee. However,
Anderson, Mansi, and Reeb (2004) employ the broad denition of
nancial expertise and nd no association between audit committee
nancial expertise and cost of debt. Additionally, anecdotal evidence
suggests that nancial expertise obtained through experience as a
CEO or president does not ensure an adequate understanding of accounting matters for an audit committee member (Livingston, 2003).
Later studies adopt a narrower denition of nancial expertise, similar to the denition initially proposed by the SEC. This denition differentiates between accounting and non-accounting nancial expertise.
Such research has provided more consistent associations between accounting nancial expertise on the audit committee and nancial
reporting quality. For instance, Krishnan and Visvanathan (2008) nd
that rms with accounting nancial experts on the audit committee
are associated with more conservative nancial reporting. Dhaliwal et
al. (2010) nds a signicant positive relation between accounting expertise on audit committees and accruals quality. Additionally, research
has examined whether accounting nancial expertise on the audit committee affects stock prices. Both Davidson, Xie, and Xu (2004) and
DeFond et al. (2005) nd that the market rewards companies for the appointment of accounting nancial experts, but shows no reaction for the
appointment of audit committee members with corporate nancial
management expertise. These studies clearly suggest that the market
discriminates between accounting and non-accounting nancial expertise on the audit committee.
2.2. Hypotheses development
An effective audit committee can enhance the credibility and reliability of the nancial statements provided to users. The SEC and the
Blue Ribbon Committee (1999) suggest that the primary responsibilities of the audit committee include assessing accounting policies,
evaluating accounting judgment, appointing and overseeing external
auditors, and appraising the quality of the rm's nancial reports.
Carcello et al. (2006) indicate that, while almost all companies disclose whether an audit committee nancial expert serves on the
audit committee, the majority of these designated nancial experts

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

do not have a strong accounting background. 4 This raises the question

whether the benets of nancial expertise differ between designated
nancial experts with accounting and non-accounting backgrounds.
Prior research indicates that the accounting nancial expertise of
the audit committee is associated with higher levels of accounting
conservatism (Krishnan & Visvanathan, 2008), higher accruals quality
(Dhaliwal et al., 2010), and a lower probability of material internal
control weaknesses (Hoitash, Hoitash, & Bedard, 2009). Thus, to the
extent that accounting nancial expertise on the audit committee
leads to higher quality nancial reporting, accounting nancial expertise may further benet the expectations of future earnings developed by nancial analysts.
Prior research suggests that analysts assimilate and process publicly
available nancial information such as past earnings and prices to predict future earnings (e.g., Abarbanell & Bushee, 1997; Schipper, 1991).
Therefore, improved nancial reporting quality should be associated
with improvements in nancial analysts forecast properties. For instance, large rms, in general, provide better nancial reporting quality
than smaller rms. Lang and Lundholm (1996) nd that earnings forecasts are more accurate, less dispersed, and exhibit less volatile forecast
revisions for large rms. As another example, common law countries, in
contrast to code law countries, tend to provide better nancial reporting
quality. Barniv, Thomas, and Myring (2005) provide evidence that analysts with superior ability and resources can better differentiate themselves with superior forecasts in common law countries in comparison
to analysts in civil law countries. Finally, higher quality auditing is also
associated with better nancial reporting quality. Behn, Choi, and
Kang (2008) investigate whether audit quality, measured as Big 4 auditors and auditors with more industry expertise, is associated with
analyst forecast properties. They nd that analyst earnings forecast accuracy is higher and forecast dispersion is lower for rms audited by
Big 4 auditors and non-Big 4 auditors with industry audit expertise.
Furthermore, Dickins et al. (2009) document that nancial analysts
have more condence in nancial statements when the disclosed audit
committee nancial expert's source of expertise is accounting-based
rather than supervisory-based. Therefore, if analysts consider the nancial information of rms whose audit committee includes an accounting
nancial expert more credible, analysts are more likely to rely on the nancial information in formulating earnings forecasts, resulting in more
accurate forecasts. More importantly, based on prior research that audit
committee accounting expertise is associated with higher reporting
quality, we expect that the accounting nancial expertise of the audit
committee to be positively associated with analysts' earnings forecast
accuracy.5
Our rst hypothesis in alternative form is as follows:
H1. The accounting nancial expertise of the audit committee is positively associated with analysts' earnings forecast accuracy.
Prior research indicates that analysts' forecast dispersion reects uncertainty about the rm's information environment (e.g., Behn et al.,
2008, Payne & Robb, 2000). Imhoff and Lobo (1992) suggest that forecast dispersion is a proxy for uncertainty about earnings before they
are announced. Furthermore, Herrmann and Thomas (2005) propose
that greater forecast dispersion indicates less agreement among analysts. They suggest that analysts with more precise information regarding future earnings are more likely to be in agreement, and thus forecast
dispersion decreases.
4
In our sample, approximately 10% of the audit committee members have accounting nancial expertise, while 60% have non-accounting nancial expertise.
5
Note that a positive association between accounting nancial expertise and earnings forecast accuracy is by no means certain. Improvements in nancial reporting
quality associated with accounting expertise on the audit committee may not be significant enough to improve analyst earnings forecasts, or improvements in nancial
reporting quality may focus on areas unrelated to improvements in analyst earnings
forecasts.

If accounting nancial expertise improves nancial reporting quality, we expect the dispersion of analysts' earnings forecasts to decrease.
Unlike forecast accuracy, which is a function of both realized current period earnings and forecasted earnings, the dispersion measure does not
depend on realized earnings, the quality of which is also likely to be a
function of audit committee expertise. In this sense, forecast dispersion
complements forecast accuracy as a measure of improvement in the nancial analysts' information environment. Our second hypothesis in alternative form is as follows:
H2. The accounting nancial expertise of the audit committee is negatively associated with analysts' earnings forecast dispersion.
3. Methodology
3.1. Sample selection
Sample selection begins with rms that are included in the S&P 500
with data available for 20002008. The sample is conned to rms in
the S&P 500 to increase data availability for members of the board of directors. We exclude rms in the nancial services industries (Standard
Industrial Classication [SIC] codes 60006999) as earnings characteristics for these rms are likely to differ. The nal sample consists of 2484
rmyear observations. For each rm, we hand collected data on the
qualications and experience of the members of the audit committee
from proxy statements, 10-K reports, company websites, and other publicly available sources.
3.2. Financial expertise
To measure nancial expertise, we assign audit committee members
into one of three categories of nancial expertise. First, audit committee
members are categorized as accounting nancial experts if they have experience as a certied public accountant, auditor, chief nancial ofcer,
controller, or chief accounting ofcer, consistent with the original denition of nancial expertise proposed by the SEC. Second, audit committee members are classied as non-accounting nancial experts if they
have experience as chief executive ofcer or president of a for-prot
company. Third, those audit committee members who are neither accounting nancial experts nor non-accounting nancial experts are categorized as nonnancial experts.
Consistent with prior research (e.g. Hoitash et al., 2009; Krishnan
& Visvanathan, 2008), we measure nancial expertise of the audit
committee as the number of audit committee directors with
accounting (non-accounting) nancial expertise divided by the total
number of directors on the audit committee. AFIN (NAFIN) is the proportion of accounting (non-accounting) nancial experts on the audit
committee.
3.3. Forecast accuracy
To test hypothesis one, we use the following equation, which
controls for previously identied determinants of analysts' forecast
properties. Table 1 provides formal denitions for each of the variables
used in our analysis. Because multiple observations from the same rm
(but from different years) are in the sample, we use t-statistics based on
HuberWhite standard errors to correct for clustering by rm. These
standard errors are robust to heteroscedasticity and serial correlation
(Huber, 1967; Rogers, 1993; White, 1980) for all the analyses. We use
the following model to test H1:

ACCY 0 1 AFIN 2 NAFIN 3 SIZE 4 SURPRISE

5 LOSS 6 ZMIJ 7 HORIZON 8 STDROE
9 COVERAGE 10 EL 11 AUISPEC 12 SOX
industry dummies year dummies :

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

Table 1
Variable denitions.
ACCY
DISP
AFINEXD
ADD
AEXP
AFIN
NAFIN
NFE
SIZE
SURPRISE
LOSS
ZMIJ
HORIZON
STDROE
COVERAGE
EL
AUISPEC

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

SOX
AQ
ACSIZE
ACMEET
ACIND
BSIZE
NODUAL
BIND
INVMILLS

=
=
=
=
=
=
=
=
=

Accuracy in analysts' earnings forecasts, dened as the negative of the absolute difference between the forecast and actual earnings, scaled by lagged stock price
Standard deviation in analysts' earnings forecasts, scaled by lagged stock price
Indicator variable equal to 1 if there is an accounting nancial expert on the audit committee, 0 otherwise
Indicator variable equal to 1 in the current year and all subsequent years in the sample in which the rm added an accounting nancial expert, 0 otherwise.
Indicator variable equal to 1 if the rm had an accounting nancial expert on the audit committee for the entire sample period, 0 otherwise
Proportion of audit committee directors who qualify as accounting nancial experts to the total number of directors on the audit committee
Proportion of audit committee directors who qualify as non-accounting nancial experts to the total number of directors on the audit committee
Proportion of audit committee directors who qualify as nonnancial experts to the total number of directors on the audit committee
Logarithm of the market value of equity at the beginning of year t.
Current year earnings minus last year earnings scaled by stock price.
Indicator variable equal to 1 if the rm had negative earnings, 0 otherwise
Zmijewski's nancial distress score
Log of the average of the number of calendar days between mean forecast announcement date and subsequent actual earnings announcement date
Standard deviation of the return on equity over the previous ve years
Log of number of analysts following the client
Earnings level measured as earnings per share
Indicator variable equal to 1 if the rm is audited by a national industry audit specialist based on the denition of Reichelt and Wang (2010) specically, the
auditor has an annual market share greater than 30% in two-digit SIC industry in the national audit market, 0 otherwise
Indicator variable equal to1 if the scal year is after 2003, 0 otherwise
Accruals quality, based on the model of Dechow and Dichev (2002)
Log of total number of directors in the audit committee
Number of meetings by the audit committee during the year
Proportion of directors that are independent in the audit committee
Log of total number of directors on the board of directors
Indicator variable equal to 1if the CEO is not also the chairman of the board, 0 otherwise
Proportion of directors that are independent in the board of directors
Inverse Mills ratio for the endogenous choice of appointment of an accounting nancial expert to the audit committee

Forecast accuracy (ACCY) is measured by the negative of the absolute

value of forecast error scaled by stock price at time t 1 (Lang &
Lundholm, 1996), as follows:
ACCY t 1

jFORECAST t EPSt j
PRICEt1

where FORECASTt is the mean I/B/E/S consensus forecast of period t

earnings made during the period starting two months before the corresponding actual earnings announcement and ending three days before
the announcement, EPSt is the actual earnings per share before extraordinary items at time t, taken from I/B/E/S, and PRICEt1 is the stock price
at the end of period t 1. The absolute forecast error is multiplied by
(1) so that larger values correspond with greater forecast accuracy.
Firm size (SIZE), which is the market value of equity at the beginning
of year t, and number of analysts following (COVERAGE) are included
based on Lang and Lundholm (1996), who document a positive association between rm size, analyst following and forecast accuracy. Earnings surprise (SURPRISE) is also based on Lang and Lundholm (1996),
who document that larger changes in earnings are associated with
less accurate forecasts. The loss indicator variable (LOSS) is included
based on Hwang, Jan, and Basu (1996), who nd that analysts' forecasts
for loss-reporting rms are on average less accurate than forecasts
for prot-reporting rms. Zmijewski's (1984) nancial distress score
(ZMIJ) is also included as nancially distressed rms tend to have less
accurate forecasts.6
We control for forecast horizon (HORIZON) based on the ndings in
Brown (2001). Horizon is measured as the natural logarithm of the
average number of calendar days between the forecast announcement
date and the corresponding actual earnings announcement date. Forecasts announced closer to the actual earnings announcement date
(i.e., short forecast horizon) should be more accurate than forecasts
announced earlier (i.e., long forecast horizon). Earnings volatility
(STDROE) is included based on Kross, Ro, and Schroeder (1990), who
have shown that analysts' earnings forecasts are less accurate for
6
ZMIJ is calculated using the following equation: X=4.34.5X1 +5.7X2 0.004X3,
where X is the overall index (ZMIJ), X1 is net income/total assets; X2 is total debt/total assets;
and X3 is current assets/current liabilities. Higher values of ZMIJ indicate higher nancial
distress.

rms with higher long-term earnings volatility. Finally, we control for

earnings level (EL) based on Eames and Glover (2003), who report
that earnings level is positively related to forecast accuracy.
Behn et al. (2008) nd that auditor industry specialization is positively (negatively) associated with analyst forecast accuracy (dispersion). We include an indicator variable that equals one if the rm is
audited by a national industry audit specialist based on the denition
of Reichelt and Wang (2010). We measure audit industry specialist
(AUISPEC) as one if the auditor has an annual market share greater
than 30% in the two-digit SIC industry in the national audit market,
zero otherwise. Begley, Cheng, and Gao (2009) provide evidence
that analyst forecast accuracy has decreased and forecast dispersion
has increased following the passage of SOX. Therefore, we include
an indicator variable (SOX) equal to one if the rmyear observation
is subsequent to 2003.
To further control for rm-specic governance characteristics, we
also estimate Eq. (3) including six additional controls for rm-specic
governance characteristics that might impact properties of analysts'
forecasts as follows:
ACCY 0 1 AFIN 2 NAFIN 3 SIZE 4 SURPRISE
5 LOSS 6 ZMIJ 7 HORIZON 8 STDROE
9 COVERAGE 10 EL 11 AUISPEC 12 SOX
13 ACSIZE 14 ACMEET 15 ACIND 16 BSIZE
17 NODUAL 18 BIND industry dummies
year dummies :

Specically, we control for audit committee size by using the natural log of the number of members on the audit committee (ACSIZE),
the number of meetings held by the audit committee during the year
(ACMEET), and the proportion of independent directors on the audit
committee (ACIND). We make no expectation for ACSIZE because
prior research is mixed on audit committee size and audit committee
effectiveness. Prior research on the frequency of audit committee
meetings is also mixed (Farber, 2005; Hoitash et al., 2009); therefore,
we make no prediction about the coefcient of ACMEET. We expect a
positive association between ACIND and ACCY because Klein (2002)
and Abbott et al. (2004) nd that audit committee independence is
indicative of good governance.

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

We control for board size by using the natural log of the number of
directors on the board (BSIZE) and expect a negative association with
ACCY because prior research suggests larger boards are less effective.
We control for separation of the roles of CEO and chairman of the
board (NODUAL) using an indicator variable coded as one if the CEO is
not the chairman of the board, and zero otherwise. We expect a positive
association between NODUAL and ACCY because prior research indicates
that separation of the role of CEO and chairman of the board enhances
corporate governance potentially increasing forecast accuracy (Agrawal
& Chadha, 2005; Jensen, 1993). Finally, we control for the proportion of
directors who are independent (BIND), and expect a positive association
with ACCY because prior research suggests that more independent
boards are associated with lower incidence of fraud and earnings management (Beasley, 1996; Klein, 2002).

expert to the audit committee, we identify those rms that added an accounting nancial expert to the audit committee during the sample period and classify those rms as ADD rms in the year the accounting
nancial expert was added and all years following the addition of the
accounting nancial expert. To control for the effects of already having
an accounting nancial expert on the audit committee, we classify
rms that had an accounting nancial expert on the audit committee
for the entire sample period as AEXP rms. Both ADD and AEXP are indicator variables that take the value of one if the rm added an expert
during the sample period (ADD) or if the rm had an accounting expert
on the audit committee from before (AEXP), zero otherwise. We estimate similar models for forecast accuracy and forecast dispersion
replacing our proportion measures (AFIN and NAFIN) with measures
of the addition of board members with accounting nancial expertise
(ADD and AEXP).

3.4. Forecast dispersion

4. Empirical results
Forecast dispersion among analysts has commonly been used as a
measure of uncertainty about future earnings since it represents the
consensus among analysts regarding future rm prospects (see, for
instance, Imhoff & Lobo, 1992 and Barron & Stuerke, 1998). Lower dispersion implies a better information environment resulting in less
uncertainty surrounding expectations of future earnings. The dispersion
of analysts' forecasts (DISP) is dened as the standard deviation of
earnings forecasts issued by individual analysts scaled by stock price
at time t 1.
DISP t

STDFORECAST t
:
PRICEt1

To test the second hypothesis, we use the following equation:

DISP 0 1 AFIN 2 NAFIN 3 SIZE 4 SURPRISE
5 LOSS 6 ZMIJ 7 HORIZON 8 STDROE
9 COVERAGE 10 EL 11 AUISPEC 12 SOX
industry dummies year dummies :

The control variables are as discussed previously in the accuracy

model (1). We expect that larger rms (SIZE) would have a smaller
forecast dispersion, while rms with more volatile earnings streams
(SURPRISE, LOSS, and STDROE), nancially distressed (ZMIJ) rms,
and rms with longer forecast horizons (HORIZON) would have a
larger forecast dispersion.
We also estimate the above equation including the six additional
rm-specic governance characteristics that may also affect properties of analysts' forecasts as follows:
DISP 0 1 AFIN 2 NAFIN 3 SIZE 4 SURPRISE
5 LOSS 6 ZMIJ 7 HORIZON 8 STDROE
9 COVERAGE 10 EL 11 AUISPEC 12 SOX
13 ACSIZE 14 ACMEET 15 ACIND 16 BSIZE
17 NODUAL 18 BIND industry dummies
year dummies :

The expected signs for the coefcients were previously discussed in

relation to forecast accuracy. The expected signs in relation to forecast
dispersion are generally in the opposite direction as those for forecast
accuracy. This is true since forecast accuracy is dened so that larger
amounts imply greater accuracy, whereas for forecast dispersion, larger
amounts imply greater dispersion (i.e., more uncertainty).
3.5. Addition of accounting expertise to the audit committee
In the previous equations, we examined the proportion of accounting
nancial expertise on the audit committee. We extend these results by
also examining the addition of accounting nancial expertise to the
audit committee. To investigate the effects of adding an accounting

4.1. Descriptive statistics

Descriptive statistics are provided in Table 2. Mean forecast accuracy
(ACCY) is .00283. Since forecast accuracy is scaled by lagged stock
price with an average price of $35.98 per share, this indicates that the
mean difference between analysts' earnings forecasts and actual earnings is between 10 and 11 cents per share (.00283 $35.98 = 0.10182)
on an average earnings per share of $2.01 in our sample. Mean forecast
dispersion (DISP) is .00163. Forecast dispersion is also scaled by lagged
stock price implying an average standard deviation of earnings forecasts
of approximately 6 cents per share (.00163 $35.98 = .05865).7
AFIN has a mean value of 0.14074, or 14.1% of the audit committee
members are accounting nancial experts, while non-accounting
nancial experts (NAFIN) and nonnancial experts (NFE) comprise
61.4% and 24.5% of the audit committee members, respectively.
The average earnings surprise (SURPRISE), calculated as current year
earnings minus last year earnings scaled by stock price, is .0595, or
about 6% of stock price. Approximately 10% of the sample rmyears
report a loss (LOSS) for the year. The mean nancial distress score for
our sample (ZMIJ) is 3.123, which indicates that our sample rms
are nancially healthy and have very little nancial distress. The mean
forecast horizon (HORIZON) is a logged value, but the unlogged value
indicates that the average number of calendar days between the forecast and the actual earnings announcement date is 45 days. The mean
number of analysts following each rm (COVERAGE) is 13.3.
The average rm in our sample has just over four board members
on the audit committee (ACSIZE) and eleven members on the board of
directors (BSIZE). Audit committees meet (ACMEET), on average, over
seven times per year. Approximately 23% of CEOs are not also chairman of the board (NODUAL). While over 92% of audit committee
members are independent board members (ACIND), just over 74% of
the board members overall are independent board members (BIND).
4.2. Correlations
The correlation matrix (Pearson correlations) is provided in Table 3.
Forecast accuracy (ACCY) exhibits a negative correlation with forecast
dispersion (DISP). Forecast accuracy (ACCY) is positively and signicantly correlated with the addition of accounting nancial expertise
(ADD), but is not signicantly correlated with the proportion of accounting nancial expertise (AFIN) on the audit committee. Similarly,
forecast dispersion (DISP) exhibits a signicant negative correlation
with the addition of accounting nancial expertise (ADD), but is not signicantly correlated with the proportion of accounting nancial expertise (AFIN) on the audit committee. However, these simple correlations
7
Extreme observations of ACCY and DISP (those beyond the 99th percentile in the
distribution) are windsorized to alleviate concerns that results are driven by outliers.

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

Table 2
Descriptive statistics.
Variable

Mean

Median

Standard
deviation

Q1

Q3

ACCY
DISP
AFINEXD
AFIN
NAFIN
NFE
ADD
AEXP
ACSIZE
ACIND
ACMEET
NODUAL
BSIZE
BIND
SIZE
SURPRISE
LOSS
ZMIJ
HORIZON
STDROE
COVERAGE
EL
AUISPEC
SOX
AQ
INVMILLS

0.00283
0.00163
0.47101
0.14075
0.61391
0.24494
0.07882
0.42141
1.45527
0.92544
7.46940
0.23309
2.38500
0.74102
9.26997
0.05950
0.10306
3.12313
3.81715
0.19386
2.58647
2.01060
0.31200
0.53905
0.15942
0.52733

0.00090
0.00075
0.00000
0.00000
0.66667
0.25000
0.00000
0.00000
1.38629
1.00000
7.00000
0.00000
2.39790
0.76923
9.24578
0.01326
0.00000
3.12433
3.73767
0.04442
2.63906
1.78000
0.00000
1.00000
0.05600
0.53073

0.00577
0.00253
0.49926
0.17584
0.23483
0.21796
0.26951
0.49390
0.25667
0.14389
3.15544
0.42289
0.19176
0.14104
1.23725
0.23986
0.30410
0.96331
0.33564
0.58595
0.54942
1.85831
0.46340
0.49857
0.40482
0.07478

0.00257
0.00037
0.00000
0.00000
0.50000
0.00000
0.00000
0.00000
1.38629
0.85714
5.00000
0.00000
2.30259
0.66667
8.41170
0.00597
0.00000
3.73897
3.66356
0.02033
2.30259
1.10000
0.00000
0.00000
0.12310
0.47671

0.00032
0.00172
1.00000
0.25000
0.75000
0.40000
0.00000
1.00000
1.60944
1.00000
9.00000
0.00000
2.48491
0.84615
9.97883
0.03353
0.00000
2.52987
3.91202
0.10230
2.94444
2.79500
1.00000
1.00000
0.02710
0.58471

Note: See Table 1 for variable denitions.

do not consider the effects of other correlated explanatory variables.

Note that forecast accuracy has a signicant positive correlation with
SIZE, COVERAGE, and earnings level (EL) and a signicant negative correlation with SURPRISE, LOSS, ZMIJ, HORIZON, and STDROE. As expected,
forecast dispersion is correlated with these same variables in the opposite direction. In the next section, after controlling for these additional
explanatory variables, we nd that both the proportion and the addition
of accounting nancial expertise are signicantly associated with increased forecast accuracy and lower forecast dispersion.

4.3. Forecast accuracy

Table 4 shows the estimation results between forecast accuracy and
nancial expertise of the audit committee. The rst column presents the
estimation results from Eq. (1). The coefcient on accounting nancial
expertise (AFIN) is positive and signicant at the 0.01 level. This result
is consistent with the rst hypothesis (H1) that accounting expertise
of a rm's audit committee is positively associated with analysts' earnings forecast accuracy. To assess the economic signicance of this result,
a one standard deviation increase in AFIN is associated with an 10.6% increase in forecast accuracy. It is also interesting to note that the coefcient on non-accounting nancial expertise (NAFIN) is not signicant,
consistent with the notion that only the accounting nancial expertise
of the audit committee is associated with forecast accuracy.
The second column (2) presents the results including six additional
control variables for corporate governance characteristics of the rm.
These additional control variables include audit committee size, the
number of audit committee meetings held each year, the proportion
of independent directors on the audit committee, board size, an indicator variable if the CEO is not also chairman of the board, and the proportion of independent directors on the board of directors. 8 The results are
8
We note that out of the six additional corporate governance characteristics, only
the coefcients on NODUAL (at the .01 level) and ACMEET (at the .10 level) are significant. However, this is generally consistent with the ndings of Krishnan and
Visvanathan (2008), who nd that none of these variables is signicantly associated
with accounting conservatism; and Dhaliwal et al. (2010) who nd that only the independence of the audit committee is associated with accruals quality.

consistent with the addition of these six control variables for corporate
governance characteristics of the rm. The coefcient on AFIN remains
positive and signicant. This result supports the prediction of the rst
hypothesis (H1) that accounting nancial expertise on the audit committee is associated with higher forecast accuracy. The coefcient on
NAFIN remains insignicant, indicating that nancial expertise gained
in supervising the preparation of nancial statements (i.e., supervisory
expertise) is not associated with improvements in forecast accuracy.
Examining the signicant control variables in columns (1) and (2),
we nd that the coefcients on SIZE and COVERAGE are positively associated with forecast accuracy. Larger rms and those followed by
more analysts have more accurate earnings forecasts. On the other
hand, rms with greater earnings surprise (SURPRISE), reporting a
loss (LOSS), longer forecast horizon (HORIZON), and reporting in the
post SarbanesOxley period (SOX) tend to have less accurate earnings
forecasts. The coefcients on the control variables reported in column
(1) are generally consistent with the coefcients reported in column
(2) that includes the additional six control variables for corporate governance characteristics. The explanatory power of the model (adj. R 2)
is about 33%. Columns (3) and (4) repeat the analysis, including additional controls for endogeneity, described below.
4.4. Controlling for endogeneity
It is possible that individuals with accounting nancial expertise do
not randomly join rms but, rather, self-select rms based on certain
rm characteristics. Specically accounting experts may choose not to
serve on the audit committee of rms with low earnings quality to
avoid damaging their reputations, being sued, or undertaking an additional workload (Engel, 2005; Beasley et al. 2009). This suggests that
rm characteristics that affect analyst forecast properties and the presence of accounting experts may be endogenously determined, which
could bias our regression analysis (Maddala, 1983). To address this
issue, we build upon models used by Agrawal and Chadha (2005),
Krishnan and Visvanathan (2008), and Dhaliwal et al. (2010) to examine whether our results persist after controlling for endogeneity. To
achieve this objective, we rst estimate a probit regression model to determine the predicted probability of having an accounting nancial expert on the audit committee. The dependent variable is AFINEXD, which
is an indicator variable that equals one if the rm has at least one accounting expert on the audit committee and zero otherwise.
From prior literature, we identify rm size, nancial reporting quality, leverage, operating performance, corporate governance, sales growth,
board size, inside ownership, capital intensity, earnings volatility, and
rm age, as rm characteristics that can drive the presence of accounting
experts on the audit committee. We include SIZE, which is the logarithm
of the market value of equity, because larger rms may have more
knowledgeable boards; therefore, small rms may benet more from
having an accounting expert on the audit committee (Agrawal &
Chadha, 2005). We measure nancial reporting quality as accruals quality (AQ) based on the model in Dechow and Dichev (2002). We include
AQ because accounting nancial experts may be more likely to serve
on the audit committee of rms with higher nancial reporting quality.
DEBT, which is long-term debt divided by total assets, is included because
rms with higher leverage are likely to have a greater need for accounting nancial expertise (Agrawal & Chadha, 2005). We measure operating
performance by the prior three-year average return on assets (PROA).
PROA is included because directors with accounting expertise may be
picked by better-managed rms, which are less likely to have accounting
problems. GINDEX measures the strength of a rm's governance system
based on 24 corporate governance provisions (Gompers, Ishii, &
Metrick, 2003). A low GINDEX means that a rm has a strong governance
system. We expect that better governed rms will be more likely to have
an accounting nancial expert on the audit committee. SGROW is the annual percentage change in sales, and is included because high-growth
rms are likely to have greater nancial needs, which requires nancial

DISP
AFIN
NAFIN
ADD
AEXP
SIZE
SURPRISE
LOSS
ZMIJ
HORIZON
COVERAGE
STDROE
EL
AUISPEC
AQ
ACSIZE
ACIND
ACMEET
NODUAL
BIND
BSIZE
SOX
INVMILLS

ACCY

DISP

AFIN

NAFIN

ADD

AEXP

SIZE

SURPRISE

LOSS

ZMIJ

HORIZON

COVERAGE

STDROE

EL

AUISPEC

AQ

ACSIZE

ACIND

ACMEET

NODUAL

BIND

BSIZE

SOX

(0.68)
(0.02)
0.01
0.03
(0.06)
0.32
(0.48)
(036)
(0.19)
(0.11)
0.20
(0.19)
0.30
0.04
(0.00)
0.02
(0.04)
(0.10)
0.03
(0.06)
0.09
(0.04)
0.22

0.04
(0.00)
(0.03)
0.07
(0.31)
0.53
0.43
0.27
0.08
(0.15)
0.24
(0.28)
(0.03)
0.02
(0.01)
0.06
0.14
(0.01)
0.07
(0.07)
0.06
(0.24)

(0.46)
0.13
0.78
(0.06)
0.02
0.07
(0.07)
(0.01)
(0.10)
0.11
0.02
(0.07)
(0.09)
(0.15)
0.04
0.29
0.04
0.09
(0.08)
0.30
(0.31)

(0.04)
(0.35)
0.07
(0.00)
(0.02)
0.02
(0.05)
0.04
(0.00)
0.06
0.03
0.02
0.09
0.07
(0.04)
(0.02)
0.11
0.06
0.03
0.07

(0.25)
0.03
0.00
(0.03)
0.01
(0.00)
(0.00)
(0.02)
0.02
0.03
0.01
0.06
0.01
0.01
(0.02)
0.02
0.03
0.07
0.00

(0.10)
0.02
0.09
(0.04)
0.03
(0.13)
0.10
0.04
(0.05)
(0.09)
(0.04)
0.03
0.22
0.02
0.10
(0.06)
0.24
(0.27)

(0.23)
(0.23)
(0.30)
(0.16)
0.58
(0.15)
0.27
0.10
(0.15)
0.11
(0.04)
0.06
0.03
(0.04)
0.31
0.10
0.26

0.41
0.27
0.06
(0.12)
0.31
(0.30)
(0.03)
0.00
(0.02)
0.02
0.12
0.01
0.05
(0.06)
0.02
(0.20)

0.28
0.11
(0.07)
0.21
(0.39)
(0.06)
(0.01)
(0.09)
(0.02)
0.08
0.05
0.01
(0.12)
(0.08)
(0.19)

0.14
(0.23)
0.32
(0.12)
0.06
0.10
0.13
0.07
0.03
(0.12)
0.09
0.12
(0.13)
(0.09)

(0.12)
0.02
(0.06)
0.01
0.05
0.02
(0.01)
0.01
(0.00)
(0.04)
0.02
(0.08)
(0.03)

(0.12)
0.07
(0.01)
(0.07)
(0.02)
(0.07)
0.00
0.12
(0.15)
0.12
(0.08)
0.16

(0.15)
(0.07)
(0.04)
(0.03)
0.03
0.09
0.04
0.06
(0.04)
0.01
(0.30)

0.05
(0.06)
0.11
0.03
0.01
(0.11)
0.11
0.14
0.21
0.11

0.06
0.11
(0.02)
(0.02)
(0.05)
0.02
0.03
(0.01)
0.11

0.03
0.02
(0.15)
0.01
(0.03)
0.03
(0.22)
0.15

0.06
(0.07)
(0.12)
0.20
0.36
0.04
0.07

(0.01)
(0.12)
0.47
0.01
0.07
(0.01)

0.08
0.11
(0.02)
0.41
(0.87)

(0.21)
(0.01)
0.06
(0.08)

(0.06)
0.15
(0.16)

0.01
0.20

0.37

Note: See Table 1 for variable denitions.

This table presents the correlations between regression variables for the full sample. Bold correlations denote signicance at p b 0.10 (two-tailed).

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

Table 3
Pearson correlations.

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

expertise on the audit committee. We control for board size (BSIZE) in

the regression because larger boards are more likely to include a nancial
expert, other things being equal. INOWN is the percentage of stock
owned by all ofcers and directors of the corporation. We include
INOWN because management's desire to protect private control benets
may decrease its desire to appoint an accounting expert to the board.
AEMP is a measure of capital intensity, computed as total assets divided
by number of employees. Firms that are more capital intensive are likely
to have a greater need for accounting expertise on the audit committee
(Agrawal & Chadha, 2005). EVOL is earnings volatility for the past three
years. Krishnan and Visvanathan (2008) suggest that rms with greater
earnings volatility should have greater asymmetric information problems in trying to obtain external nance, which would increase the
need for accounting nancial expertise on the audit committee. We measure rm age (FIRMAGE) as the age of the rm from the date of listing in
number of years. Younger rms are likely to have a greater need for nancial expertise because they tend to have greater nancing needs.
Based on these explanatory variables, we estimate the following
probit regression for our sample rms:
AFINEXD 0 1 SIZE 2 AQ 3 DEBT 4 PROA
5 GINDEX 6 SGROW 7 BSIZE 7 INOWN
8 AEMP 9 EVOL 10 FIRMAGE :

From this equation, we calculate the Inverse Mills ratio (the ratio of
the probability density function to the cumulative distribution function), and include it as an additional explanatory variable in Eqs. (1)
and (3) examining the association between forecast accuracy and accounting nancial expertise.9 Columns (3) and (4) of Table 4 present
the ndings. The results are consistent with those reported earlier in
columns (1) and (2). The proportion of accounting nancial experts
on the audit committee is signicantly associated with forecast accuracy, whereas the proportion of non-accounting nancial experts on the
audit committee has an insignicant association with forecast accuracy.
The signicance and direction of the additional control variables also remain basically unchanged after including the Inverse Mills ratio in the
equation. The coefcient on the Inverse Mills ratio is positive and significant in column (3) but not in (4). The explanatory power of the model
(adj. R2) increases slightly from 33% to 34% with the addition of the Inverse Mills ratio as an explanatory variable in the equation. Overall, the
ndings reported in columns (3) and (4) are consistent with our rst
hypothesis (H1) and help alleviate concerns that the primary results
are driven by endogeneity.

Table 4
Forecast accuracy and nancial expertise of the audit committee.
Variable

Predicted
sign

(1)

(2)

(3)

(4)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

INTERCEPT

AFIN

NAFIN

SIZE

SURPRISE

LOSS

ZMIJ

HORIZON

STDROE

COVERAGE

EL

AUISPEC

SOX

0.0069***
(0.000)
0.0017***
(0.007)
0.0004
(0.419)
0.0008***
(0.000)
0.0076***
(0.000)
0.0028***
(0.000)
0.0001
(0.367)
0.0007**
(0.028)
0.0002
(0.232)
0.0004*
(0.086)
0.0004***
(0.003)
0.0000
(0.844)
0.0038***
(0.000)

0.0086***
(0.000)
0.0019***
(0.003)
0.0004
(0.410)
0.0007***
(0.000)
0.0076***
(0.000)
0.0028***
(0.000)
0.0001
(0.389)
0.0007**
(0.032)
0.0001
(0.365)
0.0004*
(0.062)
0.0004***
(0.004)
0.0001
(0.756)
0.0034***
(0.000)

ACSIZE

ACMEET

ACIND

BSIZE

NODUAL

BIND

0.0043*
(0.061)
0.0020***
(0.002)
0.0005
(0.233)
0.0008***
(0.000)
0.0075***
(0.000)
0.0029***
(0.000)
0.0002*
(0.097)
0.0007**
(0.036)
0.0002
(0.185)
0.0002
(0.187)
0.0004***
(0.003)
0.0000
(0.857)
0.0035***
(0.000)
0.0002
(0.644)
0.0001*
(0.082)
0.0008
(0.218)
0.0004
(0.436)
0.0007***
(0.002)
0.0008
(0.345)

INVMILLS

0.0062**
(0.035)
0.0020***
(0.003)
0.0005
(0.268)
0.0008***
(0.000)
0.0076***
(0.000)
0.0029***
(0.000)
0.0002*
(0.095)
0.0007**
(0.022)
0.0001
(0.556)
0.0003
(0.122)
0.0004***
(0.000)
0.0001
(0.776)
0.0035***
(0.000)
0.0001
(0.796)
0.0000
(0.952)
0.0008
(0.289)
0.0006
(0.322)
0.0007***
(0.002)
0.0007
(0.382)
0.0039
(0.382)
0.337
2484

Adj. R2
Sample size

0.331
2484

0.336
2484

0.0033**
(0.043)
0.332
2484

Notes: Variables are dened in Table 1. Each model includes, but does not tabulate, year
dummies. p-Values (indicated within parentheses) are computed based on HuberWhite
robust standard errors that correct for serial correlation among multiple-year observations.
*, **, *** indicate signicance at the 10%, 5%, and 1% levels, respectively. Signicances are
one-tailed tests where predicted signs are specied and two tailed tests otherwise.

4.5. Forecast dispersion

Table 5 shows the estimation results between forecast dispersion
and nancial expertise of the audit committee. AFIN, the variable of interest, has a negative and signicant coefcient across all four columns.
This result is consistent with the second hypothesis (H2) that the accounting nancial expertise of the audit committee is associated with
lower analyst forecast dispersion. Once again, it is interesting to note
the economic signicance of this nding. Specically, a one standard
deviation increase in AFIN is associated with a 6.13% decrease in forecast
dispersion. Similar to our results for forecast accuracy, NAFIN is negative
but insignicant across all four columns, indicating that only the accounting nancial expertise of the audit committee is associated with
lower forecast dispersion.
Turning to the control variables, SIZE is negatively associated with
forecast dispersion. Larger rms are associated with a lower
9
In the rst stage regression, all explanatory variables except for ISSUE and BSIZE are
signicant at the .10 level or better. The pseudo-R2 of the model is 0.066, which is modest, but higher than that reported in Agrawal and Chadha (2005). Given the low explanatory power in the rst stage model, care should be taken in reliance of the
coefcients from the second stage results.

dispersion of earnings forecasts. SURPRISE, LOSS, ZMIJ, AUISPEC, and

SOX are positively associated with forecast dispersion. Firms with
greater earnings surprise, reporting a loss, having a higher nancial
distress score, audited by industry audit specialists, and reporting in
the post SarbanesOxley period are related to a higher dispersion of
earnings forecasts. The explanatory power of the model is about 41%.
Column (2) shows the forecast dispersion results of Eq. (6) including
the six additional corporate governance characteristics of the rm. The
results are consistent with those reported in column (1). AFIN, our primary variable of interest, has a negative and signicant coefcient
supporting the second hypothesis (H2) that the accounting nancial expertise of the audit committee is associated with lower forecast
dispersion. NAFIN is negative but insignicant, indicating that only the
accounting nancial expertise of the audit committee is signicantly associated with lower analyst forecast dispersion. Columns (3) and (4) include an additional explanatory variable, the Inverse Mills ratio, to
address concerns that rm characteristics affecting forecast dispersion
and rm characteristics affecting the presence of accounting experts
on the audit committee are endogenously determined. A comparison
of the results reported in columns (3) and (4) with those reported in

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111
Table 5
Forecast dispersion and nancial expertise of the audit committee.
Variable

Table 6
Forecast accuracy and the addition of accounting expertise to the audit committee.

Predicted
sign

(1)

(2)

(3)

(4)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

INTERCEPT

AFIN

NAFIN

SIZE

SURPRISE

LOSS

ZMIJ

HORIZON

STDROE

COVERAGE

EL

AUISPEC

SOX

0.0039***
(0.000)
0.0004*
(0.072)
0.0000
(0.942)
0.0003***
(0.000)
0.0035***
(0.000)
0.0017***
(0.000)
0.0002***
(0.000)
0.0001
(0.365)
0.0002**
(0.046)
0.0002**
(0.028)
0.0001**
(0.011)
0.0001
(0.247)
0.0018***
(0.000)

0.0046***
(0.000)
0.0005**
(0.046)
0.0000
(0.954)
0.0003***
(0.000)
0.0035***
(0.000)
0.0017***
(0.000)
0.0002***
(0.000)
0.0000
(0.375)
0.0001*
(0.095)
0.0002**
(0.032)
0.0001**
(0.014)
0.0001
(0.209)
0.0017***
(0.000)

ACSIZE

ACMEET

ACIND

BSIZE

NODUAL

BIND

0.0022**
(0.011)
0.0006**
(0.028)
0.0001
(0.719)
0.0004***
(0.000)
0.0035***
(0.000)
0.0018***
(0.000)
0.0001**
(0.032)
0.0000
(0.465)
0.0002**
(0.039)
0.0003**
(0.021)
0.0001**
(0.010)
0.0001
(0.243)
0.0016***
(0.000)
0.0002
(0.185)
0.0001
(0.011)
0.0007
(0.012)
0.0003
(0.183)
0.0002**
(0.018)
0.0001
(0.831)

INVMILL

0.0000
(0.983)
0.0006**
(0.028)
0.0001
(0.631)
0.0005***
(0.000)
0.0034***
(0.000)
0.0018***
(0.000)
0.0001*
(0.063)
0.0000
(0.499)
0.0003**
(0.033)
0.0003**
(0.014)
0.0001***
(0.007)
0.0001
(0.462)
0.0017***
(0.000)
0.0003*
(0.080)
0.0001**
(0.047)
0.0007**
(0.015)
0.0001
(0.596)
0.0002**
(0.017)
0.0001
(0.680)
0.0048
(0.123)
0.411
2484

Adj. R2
Sample size

0.402
2484

0.409
2484

0.0013*
(0.091)
0.403
2484

Variable

Predicted
sign

(1)

(2)

(3)

(4)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

INTERCEPT

ADD

AEXP

SIZE

SURPRISE

LOSS

ZMIJ

HORIZON

STDROE

COVERAGE

EL

AUISPEC

SOX

0.0059***
(0.006)
0.0005**
(0.028)
0.0002
(0.320)
0.0008***
(0.000)
0.0076***
(0.000)
0.0025***
(0.000)
0.0002
(0.201)
0.0008**
(0.048)
0.0002
(0.398)
0.0002
(0.372)
0.0004***
(0.003)
0.0001
(0.600)
0.0036***
(0.000)

0.0070***
(0.005)
0.0006**
(0.022)
0.0003
(0.256)
0.0008***
(0.000)
0.0076***
(0.000)
0.0025***
(0.000)
0.0002
(0.204)
0.0008**
(0.027)
0.0002
(0.544)
0.0003
(0.318)
0.0004***
(0.003)
0.0001
(0.552)
0.0034***
(0.000)

ACSIZE

ACMEET

ACIND

BSIZE

NODUAL

BIND

0.0037
(0.166)
0.0006**
(0.016)
0.0003*
(0.097)
0.0009***
(0.000)
0.0075***
(0.000)
0.0026***
(0.000)
0.0003*
(0.062)
0.0008**
(0.033)
0.0003
(0.329)
0.0001
(0.658)
0.0004***
(0.003)
0.0001
(0.639)
0.0035***
(0.000)
0.0004
(0.432)
0.0001
(0.239)
0.0009
(0.202)
0.0003
(0.663)
0.0008***
(0.003)
0.0005
(0.611)

INVMILLS

0.0042
(0.334)
0.0006***
(0.017)
0.0003*
(0.096)
0.0008***
(0.000)
0.0075***
(0.000)
0.0026***
(0.000)
0.0003*
(0.078)
0.0008*
(0.063)
0.0002
(0.458)
0.0001
(0.648)
0.0004***
(0.003)
0.0001
(0.608)
0.0035***
(0.000)
0.0004
(0.488)
0.0000
(0.847)
0.0009
(0.199)
0.0003
(0.620)
0.0008***
(0.003)
0.0004
(0.625)
0.0011
(0.858)
0.344
2195

Adj. R2
Sample size

0.339
2195

0.344
2195

0.0020
(0.230)
0.340
2195

Notes: Variables are dened in Table 1. Each model includes, but does not tabulate, year
dummies. p-Values (indicated within parentheses) are computed based on HuberWhite
robust standard errors that correct for serial correlation among multiple-year observations.
*, **, *** indicate signicance at the 10%, 5%, and 1% levels, respectively. Signicances are
one-tailed tests where predicted signs are specied and two tailed tests otherwise.

Notes: Variables are dened in Table 1. Each model includes, but does not tabulate, year
dummies. p-Values (indicated within parentheses) are computed based on HuberWhite
robust standard errors that correct for serial correlation among multiple-year observations.
*, **, *** indicate signicance at the 10%, 5%, and 1% levels, respectively. Signicances are
one-tailed tests where predicted signs are specied and two tailed tests otherwise.

columns (1) and (2) indicate the ndings continue to hold with inclusion of controls for endogeneity.

endogeneity. The coefcient on ADD is positive and signicant in all

four columns. This provides additional evidence in support of H1, that
not only the proportion, but also the addition of accounting expertise
to the audit committee is positively associated with forecast accuracy.
The coefcient on AEXP is also positive, but insignicant.10
Table 7 provides the estimation results between forecast dispersion and the addition of accounting expertise to the audit committee.
The coefcient on ADD is negative and signicant in all four columns.
In further support of H2, the addition of accounting expertise to the
audit committee is associated with a reduction in forecast dispersion.
Similar to the results reported in Table 6, the coefcient on AEXP is in
the same direction as ADD, but insignicant. Overall, we provide evidence of a link between accounting nancial expertise and analyst
earnings forecasts. These ndings hold for accounting nancial expertise measured in two ways: (1) as a proportion of members serving

4.6. Addition of accounting expertise to the audit committee

In Tables 4 and 5, we examined the proportion of accounting expertise on the audit committee and its relation to analyst earnings
forecasts. In this section, we extend these ndings by examining the
addition of accounting nancial expertise to the audit committee.
We classify rms as ADD rms in the year the accounting nancial expert was added and all years following the addition of the accounting
nancial expert. We classify rms that had an accounting nancial expert on the audit committee for the entire sample period as AEXP
rms.
Table 6 presents the estimation results between forecast accuracy
and the addition of accounting expertise to the audit committee. Column (1) presents our main equation, column (2) includes six additional
control variables for corporate governance characteristics of the rm,
and columns (3) and (4) include the Inverse Mills ratio to control for

10
This may be attributable to the small number of rms disclosing an accounting nancial expert over our entire sample period beginning in 2001.

10

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111

Table 7
Forecast dispersion and the addition of accounting expertise to the audit committee.
Variable

Predicted
sign

(1)

(2)

(3)

(4)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

Coefcient
(p-value)

INTERCEPT

ADD

AEXP

SIZE

SURPRISE

LOSS

ZMIJ

HORIZON

STDROE

COVERAGE

EL

AUISPEC

SOX

0.0035***
(0.000)
0.0002*
(0.099)
0.0001
(0.358)
0.0003***
(0.000)
0.0034***
(0.000)
0.0017***
(0.000)
0.0002***
(0.002)
0.0001
(0.173)
0.0002*
(0.093)
0.0003***
(0.019)
0.0001***
(0.006)
0.0001***
(0.085)
0.0018***
(0.000)

0.0039***
(0.000)
0.0002*
(0.082)
0.0001
(0.290)
0.0003***
(0.000)
0.0034***
(0.000)
0.0017***
(0.000)
0.0002***
(0.002)
0.0001
(0.205)
0.0002*
(0.078)
0.0003**
(0.023)
0.0001***
(0.007)
0.0002*
(0.075)
0.0018***
(0.000)

ACSIZE

ACMEET

ACIND

BSIZE

NODUAL

BIND

0.0019*
(0.053)
0.0002**
(0.050)
0.0001
(0.170)
0.0004***
(0.000)
0.0034***
(0.000)
0.0018***
(0.000)
0.0001**
(0.029)
0.0001
(0.277)
0.0002**
(0.038)
0.0003***
(0.008)
0.0001***
(0.006)
0.0001*
(0.089)
0.0017***
(0.000)
0.0003
(0.149)
0.0000**
(0.035)
0.0009***
(0.006)
0.0002
(0.429)
0.0003***
(0.008)
0.0001***
(0.799)

INVMILLS

0.0009
(0.670)
0.0002*
(0.055)
0.0001***
(0.210)
0.0005***
(0.000)
0.0033***
(0.000)
0.0018***
(0.000)
0.0001*
(0.094)
0.0001
(0.621)
0.0004**
(0.025)
0.0003***
(0.005)
0.0001***
(0.004)
0.0001
(0.246)
0.0018***
(0.000)
0.0004*
(0.052)
0.0002**
(0.042)
0.0008***
(0.007)
0.0000
(0.980)
0.0003***
(0.008)
0.0000
(0.961)
0.0058*
(0.079)
0.421
2195

Adj. R2
Sample size

0.412
2195

0.419
2195

0.0008
(0.282)
0.413
2195

Notes: Variables are dened in Table 1. Each model includes, but does not tabulate, year
dummies. p-Values (indicated within parentheses) are computed based on HuberWhite
robust standard errors that correct for serial correlation among multiple-year observations.
*, **, *** indicate signicance at the 10%, 5%, and 1% levels, respectively. Signicances are
one-tailed tests where predicted signs are specied and two tailed tests otherwise.

on the audit committee and (2) as the addition of accounting nancial

expertise to the audit committee.
4.7. Robustness tests
Because prior research has shown that the accounting nancial expertise of the audit committee is positively associated with nancial
reporting quality, we perform a robustness test to directly control for nancial reporting quality using an accruals quality measure (AQ) developed by Dechow and Dichev (2002). 11 We are interested in whether
nancial reporting quality represents an intervening variable between
audit committee nancial expertise and analyst forecast properties.
We re-estimate our previous analyses and include AQ in the models as
an additional explanatory variable to explore whether the coefcients
on audit committee nancial expertise as well as their signicance
11
Dechow and Dichev (2002) model normal accruals as a function of past, present,
and future cash ows, that is, NACCt = (OCFt 1, OCFt, OCFt + 1). Specically:
WC= + 1CFOt 1 + 2CFOt +3CFOt + 1 + t, where (t) or absolute t proxies for accrual quality as an unsigned measure of extent of accrual errors.

level are reduced. If we include AQ in the forecast accuracy model originally reported in Table 4, column 1, the coefcient on AFIN drops from
0.0017 to 0.0016, but is still signicant (t=2.45). The coefcient on AQ
is signicant and positive (t= 2.04). If we include AQ in the forecast dispersion model originally reported in Table 5, column 1, AFIN goes from
0.0004 to 0.00039, and it is still signicant (t=1.39). The coefcient on AQ is positive, but not signicant in the dispersion model.
Taken together, these results are consistent with the idea of nancial
reporting quality being related to analyst forecast properties, and suggest that audit committee nancial expertise represents a dimension
of reporting quality not fully captured in the conventional reporting
quality measure.
5. Conclusion
While prior research provides evidence that having an accounting
expert on an audit committee improves nancial reporting quality,
little is known as to whether the higher reporting quality due to accounting expertise translates into tangible economic benets to nancial statement users. Our results show that nancial analyst
earnings forecast properties (i.e., more accurate and less dispersed
forecasts) are associated with the accounting expertise of a rm's
audit committee. Furthermore, we do not nd a signicant association between non-accounting nancial expertise and properties of
analyst forecasts.
These ndings are of direct importance to investors in the capital
markets since analyst earnings forecasts are primary inputs in equity
valuation. Our ndings further contribute to the growing literature on
audit committee's expertise by documenting that accounting expertise (and not non-accounting expertise) is associated with greater analysts' forecast accuracy and lower forecast dispersion. The results
also have important implications for regulators, corporate boards,
and others in dening requirements for nancial expertise. Our ndings suggest that adopting a narrower denition of a nancial expert
as originally proposed by the SEC is likely to enhance the audit committees' effectiveness. Our ndings are also relevant to regulators in
other countries who are considering steps to enhance the effectiveness of audit committees.
Finally, we note that our ndings are subject to limitations. First, we
document an association rather than causation between accounting expertise on the audit committee and attributes of analysts' earnings forecasts. While we provide evidence of a link between accounting expertise
and analyst earnings forecasts, we cannot state conclusively that accounting expertise on the audit committee is directly attributable to improvements in forecast accuracy and forecast dispersion of analysts.
Second, while we take steps to alleviate concerns regarding endogeneity,
we cannot rule out the possibility that our results might be inuenced by
omitted correlated variables. We further attempt to mitigate this concern by controlling for several observable governance and other characteristics in our model.
Despite these limitations, this paper extends the literature regarding nancial expertise on the audit committee. While prior literature
provides evidence of link between accounting nancial expertise and
nancial reporting quality, we extend the literature in providing evidence of an association between accounting nancial expertise and
the use of nancial reporting information in the form of improved analyst earnings forecasts.
References
Abarbanell, J. S., & Bushee, B. J. (1997). Fundamental analysis, future earnings, and
stock prices. Journal of Accounting Research, 35(Spring), 124.
Abbott, L., Parker, S., & Peters, G. (2004). Audit committee characteristics and restatements. Auditing: A Journal of Practice & Theory, 23(1), 6987.
Agrawal, A., & Chadha, S. (2005). Corporate governance and accounting scandals. Journal of Law and Economics, 48, 371406.
Anderson, R. C., Mansi, S. A., & Reeb, D. M. (2004). Board characteristics, accounting report
integrity, and the cost of debt. Journal of Accounting and Economics, 37, 315342.

J.L. Abernathy et al. / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 111
Archambeault, D., & DeZoort, F. T. (2001). Auditor opinion shopping and the audit
committee: An analysis of suspicious auditor switches. International Journal of Auditing,
5(1), 3352.
Ashbaugh-Skaife, H., Collins, D. W., & LaFond, R. (2006). The effects of corporate governance on rms' credit ratings. Journal of Accounting and Economics, 42(1/2), 203227.
Barniv, R., Thomas, W. B., & Myring, M. (2005). The association between the legal and
nancial reporting environments and forecast performance of individual analysts.
Contemporary Accounting Research, 22(4), 727758.
Barron, O. E., & Stuerke, P. S. (1998). Dispersion in analysts' earnings forecasts as a
measure of uncertainty. Journal of Accounting, Auditing, and Finance, 13(Summer),
245270.
Beasley, M. (1996). An empirical analysis of the relation between the board of director
composition and nancial fraud. The Accounting Review, 71(4), 443465.
Beasley, M. S., Carcello, J. V., Hermanson, D. R., & Neal, T. (2009). The audit committee
oversight process. Contemporary Accounting Research, 26(1), 65122.
Bdard, J., Chtourou, S. M., & Courteau, L. (2004). The effect of audit committee expertise, independence, and activity on aggressive earnings management. Auditing: A
Journal of Practice & Theory, 23(2), 1335.
Begley, J., Cheng, Q., & Gao, Y. (2009). Changes in analysts' information environment following the SarbanesOxley Act and the global settlement. Working paper. University
of British Columbia.
Behn, B., Choi, J. -H., & Kang, T. (2008). Audit quality and properties of analyst earnings
forecasts. The Accounting Review, 83(2), 327349.
Blue Ribbon Committee (1999). Report and recommendations of the Blue Ribbon Committee on improving the effectiveness of corporate audit committees. New York
Stock Exchange and National Association of Securities Dealers.
Brown, L. D. (2001). How important is past analyst forecast accuracy? Financial Analyst
Journal(November/December), 4449.
Carcello, J., Hollingsworth, C., & Neal, T. (2006). Audit committee nancial experts: A
closer look using rm designations. Accounting Horizons, 20(4), 351373.
Comprix, J., Guo, J., Zhang, Y., & Zhou, N. (2012). Accounting expertise on audit committees and expected rates of return on pension assets: Post 132r evidence. Working paper. Syracuse University.
Davidson, W. N., III, Xie, B., & Xu, W. (2004). Market reaction to voluntary announcements of audit committee appointments: The effect of nancial expertise. Journal
of Accounting and Public Policy, 23(July), 279293.
Dechow, P., & Dichev, I. D. (2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77(Supplement), 3559.
DeFond, M. L., Hann, R. N., & Hu, X. (2005). Does the market value nancial expertise on
audit committees of boards of directors? Journal of Accounting Research, 43(2),
153193.
Dhaliwal, D., Naiker, V., & Navissi, F. (2010). The association between accruals quality
and the characteristics of accounting experts and mix of expertise on audit committees. Contemporary Accounting Research, 27(Fall), 787827.
Dickins, D., Hillison, W., & Platau, S. (2009). Do nancial statement users care about differences in board members' source of nancial expertise? Views of nancial analysts. The Journal of Applied Business and Economics, 9(2), 2136.
Eames, M., & Glover, S. (2003). Earnings predictability and the directing of analysts'
earnings forecast errors. The Accounting Review, 78(July), 707724.
Engel, E. (2005). Discussion of Does the market value nancial expertise on audit committees of boards of directors? Journal of Accounting Research, 43(2), 195204.
Farber, D. (2005). Restoring trust after fraud: Does corporate governance matter? The
Accounting Review, 80(2), 539561.

11

Financial Accounting Standards Board (FASB) (1978). Objectives of nancial reporting

by business enterprises. Statement of Financial Accounting Concepts No. 1. Norwalk,
CT: Financial Accounting Standards Board (FASB).
Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate governance and equity prices.
Quarterly Journal of Economics, 118(1), 107155.
Herrmann, D., & Thomas, W. B. (2005). Rounding of analyst forecasts. The Accounting
Review, 80(July), 805823.
Hoitash, U., Hoitash, R., & Bedard, J. C. (2009). Corporate governance and internal control over nancial reporting: A comparison of regulatory regimes. The Accounting
Review, 84(3), 839867.
Huber, P. J. (1967). The behavior of maximum likelihood estimates under nonstandard
conditions. Proceedings of the fth Berkeley symposium on mathematical statistics
and probability, 1. (pp. 221223)Berkeley, CA: University of California Press.
Hwang, L., Jan, C., & Basu, S. (1996). Loss rms and analysts' earnings forecast errors.
Journal of Financial Statement Analysis, 1(Winter), 1831.
Imhoff, E. A., & Lobo, G. J. (1992). The effect of ex ante earnings uncertainty on earnings
response coefcients. The Accounting Review, 67(April), 427439.
Jensen, M. (1993). The modern industrial revolution, exit, and the failure of internal
control systems. Journal of Finance, 48(3), 831880.
Kecskes, A., Michaely, R., & Womack, K. (2010). What drives the value of analysts'
recommendations: Earnings estimates or discount rate estimates? Working paper. Virginia Tech Available at SSRN: http://ssrn.com/abstract=1478451
Klein, A. (2002). Audit committee, board of director characteristics, and earnings management. Journal of Accounting and Economics, 33, 375400.
Kothari, S. P. (2001). Capital markets research in accounting. Journal of Accounting and
Economics, 31(13), 105231.
Krishnan, J. (2005). Audit committee quality and internal control: An empirical analysis. The Accounting Review, 80(2), 649675.
Krishnan, G. V., & Visvanathan, G. (2008). Does the SOX denition of an accounting expert matter? The association between audit committee directors' accounting expertise and accounting conservatism. Contemporary Accounting Research, 25(3),
827857.
Kross, W., Ro, B. T., & Schroeder, D. (1990). Earnings expectations: the analysts' information advantage. The Accounting Review, 65(April), 461476.
Lang, M., & Lundholm, R. (1996). Corporate disclosure policy and analyst behavior. The
Accounting Review, 71(October), 467492.
Livingston, P. (2003). Financial experts on audit committees. Financial Executive, 19, 67.
Maddala, G. S. (1983). Limited-dependent and qualitative variables in econometrics.
Cambridge, UK: Cambridge University Press.
Payne, J., & Robb, S. (2000). Earnings management: The effect of ex ante earnings uncertainty. Journal of Accounting, Auditing & Finance, 15(Fall), 371392.
Reichelt, K., & Wang, D. (2010). National and ofce specic measure of auditor industry
expertise and effects on audit quality. Journal of Accounting Research, 48(3), 647686.
Rogers, W. H. (1993). Regression standard errors in clustered samples. Stata Technical
Bulletin, 13, 1923.
SarbanesOxley Act of 2002 (2002). (SOX) Public Law 107204. 107th Congress, 2d Session, July 24, 2002 Available on Congress.gov
Schipper, K. (1991). Commentary on analysts' forecasts. Accounting horizons(December),
105121.
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a
direct test for heteroskedasticity. Econometrica, 48(May), 817838.
Zmijewski, M. E. (1984). Methodological issues related to the estimation of nancial
distress prediction models. Journal of Accounting Research, 22(Supplement), 5982.