J. Account.

Public Policy xxx (xxxx) xxx

Contents lists available at ScienceDirect

J. Account. Public Policy

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

Short Communication

Do seasoned industry specialists provide higher

audit quality? A re-examination q
John Daniel Eshleman ⇑, Peng Guo
Rutgers, The State University of New Jersey, School of Business – Camden, United States

a r t i c l e i n f o a b s t r a c t

Article history: The purpose of this paper is to re-examine the recent findings in Gaver and Utke (2019)
Available online xxxx (GU) who find that seasoned industry specialist auditors provide higher audit quality.
We first illustrate how the magnitude of residuals from the accruals model can vary signif-
JEL classifications: icantly by industry, thus highlighting the importance of including industry fixed effects
M41 when the dependent variable is discretionary accruals. We next replicate GU’s findings.
M42 We first attempt to replicate prior literature after including industry fixed effects in the
Keywords:
audit quality (discretionary accruals) models. This is an important control, as the relevant
Audit quality benchmark for a firm with an industry specialist auditor is a peer firm in the same industry
Industry specialization with a non-specialist auditor. We find that after including industry fixed effects, there is no
Accruals models association between seasoned industry specialist auditors and discretionary accruals. We
Industry effects also find that the association between industry specialization and discretionary accruals
is very sensitive to the way in which the researcher calculates specialization. Our findings
are informative for shareholders of public companies who vote on auditor ratification.
Ó 2020 Elsevier Inc. All rights reserved.

1. Introduction

Audit firms often discuss the importance of industry expertise in achieving high quality audits. For example, Ernst &
Young states that ‘‘We believe that having our professionals develop a deep understanding of industry-specific issues
improves the quality of our audits” (E&Y, 2017). Understanding whether and how industry specialization impacts audit qual-
ity is an important issue for public firms choosing among auditors and for auditors concerned with the quality of their audits.
Because of its importance in public company audits, industry specialization has received much attention from research-
ers. Although industry expertise is unobservable at the firm, office, or auditor level, researchers have used an audit firm’s
market share within an industry as a measure of industry specialization. Initial research in this area finds that industry spe-
cialist audit firms and industry specialist audit offices are associated with higher earnings quality, and therefore higher audit
quality (Balsam et al., 2003; Reichelt and Wang, 2010).
Minutti-Meza (2013) points out that defining industry expertise using market share proxies results in vast differences in
client characteristics between specialists and non-specialists. This creates an econometric issue because auditors with larger

q
We thank Keval Amin, Qian Feng, Francis Kim, Miguel Minutti-Meza, Leah Muriel, Zhifeng Yang, and all workshop participants at the 2019 American
Accounting Association Annual meeting, Stony Brook University, and Rutgers School of Business – Camden for helpful comments. Any errors that remain are
our own.
⇑ Corresponding author.
E-mail addresses: dan.eshleman@rutgers.edu (J.D. Eshleman), pg480@rcamden.rutgers.edu (P. Guo).

https://doi.org/10.1016/j.jaccpubpol.2020.106770
0278-4254/Ó 2020 Elsevier Inc. All rights reserved.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
2 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

market shares tend to have larger clients, which have higher earnings quality. Minutti-Meza (2013) finds that when using
matching techniques or client fixed effects to remove the confounding effect of firm size, there is no association between
market share based measures of industry specialization and audit quality.
In a recent paper, Gaver and Utke (2019) (hereafter, GU) challenge the findings of Minutti-Meza (2013). The authors argue
that newly ‘‘unseasoned” industry specialists do not provide audit quality similar to those of ‘‘seasoned” industry specialists
who have been specialists in an industry for a number of years. Using the absolute value of discretionary accruals, positive
discretionary accruals, and book-tax differences as measures of audit quality, GU show that seasoned industry specialists
provide higher audit quality compared to both non-specialists and unseasoned specialists. They find that these results are
not sensitive to using matching techniques employed by Minutti-Meza (2013). The authors conclude that the seasoning pro-
cess takes up to three years, at which point the auditor industry leader becomes a seasoned expert and provides audit quality
similar to other experts.
An important element missing from both Minutti-Meza (2013) and GU is control for industry effects in the audit quality
models. Minutti-Meza (2013, footnote 20) state that ‘‘. . .the discretionary accruals model does not include industry fixed
effects because this audit-quality proxy is estimated by industry.” Similarly, Gaver and Utke (2019, footnote 16) state that
‘‘. . .we do not include industry fixed effects in the accruals models because accruals are estimated by industry.” However,
when testing whether industry specialist auditors provide higher audit quality, the relevant benchmark is other firms in
the same industry, not all firms in all industries. Therefore, conceptually, it is important to control for the effect of being
in a certain industry on audit quality.1 Furthermore, there are econometric issues with not including industry fixed effects
in accruals models. Hribar and Nichols (2007) point out that the absolute value of discretionary accruals in a given industry
is an increasing function of the variance of the underlying error term in the 1st stage model. In other words, the worse the accru-
als model fits a particular industry, the larger the magnitude of discretionary accruals in that industry. It is therefore quite
important to control for the effect of industry when the dependent variable is discretionary accruals. With this in mind, the pur-
pose of this paper is to challenge the findings of GU by re-examining the relationship between seasoned industry specialization
and audit quality. We do so by including industry fixed effects in the audit quality models.
Our sample is comprised of 28,906 firm-year observations with Big 4 auditors during the 2003–2015 time period. We first
illustrate that the magnitude of discretionary accruals is decreasing in the goodness of fit of the accruals model for a partic-
ular industry, as Hribar and Nichols (2007) predict. We then replicate the findings of GU, finding a significantly positive asso-
ciation between seasoned industry specialization and audit quality using both the magnitude of discretionary accruals and
positive discretionary accruals. We then add industry fixed effects in the audit quality models. An F-test reveals that the coef-
ficients on 2-digit SIC code industry fixed effects are jointly significant in explaining both the unsigned and signed discre-
tionary accruals. Importantly, we find that the coefficient on GU’s seasoned industry specialist variable becomes
statistically insignificant when we include industry fixed effects. We also analyze the relationship between seasoned indus-
try specialists and positive discretionary accruals. We find that adding industry fixed effects to the model results in the coef-
ficient on the seasoned industry specialist variable becoming insignificant. Collectively, these findings highlight the
importance of controlling for industry effects in tests of earnings management.
In subsequent analysis we examine other research design choices. In calculating the industry market shares to compute
industry specialization measures, Reichelt and Wang (2010) use audit fees, Minutti-Meza (2013) use client assets, and GU
use client sales. This makes it very difficult to reconcile the mixed evidence. A recent paper by Audoussett-Coulier et al.
(2016) state that ‘‘Our findings suggest that audit fee-based measures should probably be prioritized by researchers and that
previous empirical findings based on other measurement variables need to be re-examined” –(Audoussett-Coulier et al.,
2016, 158). We find that the association between seasoned industry specialists and audit quality measured using discre-
tionary accruals is not robust to using audit fees to calculate industry shares. We also find that the association between sea-
soned industry specialists and audit quality is much weaker if we use client assets as to calculate market shares.
Finally, we demonstrate that after including industry fixed effects, there is no association between auditor industry spe-
cialization at the local city level and audit quality. This finding is robust to several city-level measures of specialization,
including a portfolio-share based measure in which specialization is defined based on the percentage of the auditor’s client
portfolio operating in a given industry.
This study contributes to the literature in several ways. First, our finding that the positive association between special-
ization and audit quality disappears when including industry fixed effects highlights the importance of controlling for indus-
try effects in tests of earnings management, following Hribar and Nichols (2007).2 Unlike Minutti-Meza (2013), who argues
that any positive association between specialists and quality is driven by endogeneity, we argue that it is attributable to a lack of
controls for industry effects. Third, the evidence that GU’s findings are sensitive to the use of audit fee-based specialization mea-
sures underscores the importance of justifying the use of industry specialization measures and examining the sensitivity of find-
ings to alternative measures of specialization (Audoussett-Coulier et al., 2016). Overall, the findings in this paper serve to help
the methodology of future research on audit quality and industry specialization.

1
In other words, the test is whether clients in industry j who have specialist auditors have higher earnings quality compared with clients in industry j who
do not have specialist auditors. If we fail to control for the earnings quality differences of being in industry j, we are now comparing clients in industry j with
specialist auditors to all clients with non-specialist auditors, whether in industry j or any other industry.
2
Many audit quality studies using discretionary accruals as the measure of audit quality do include industry fixed effects (e.g., Choi et al., 2010, 2012;
Lawrence et al., 2011; Lopez and Peters, 2012; Gul et al., 2013; Bills et al., 2016).

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx 3

The findings in this paper are also relevant to public policy. The Public Company Accounting Oversight Board (PCAOB)
lists industry specialization as one of its 28 audit quality indicators (PCAOB, 2015). In addition, many large audit firms tout
their industry specializations on their websites.3 It is therefore important to understand whether and how industry specialists
provide higher quality audits. Whether specialization impacts the quality of audits is also very useful information for the public.
Investors may consider this information when voting on auditor ratification.4 In addition, our findings could impact the orga-
nizational structure of audit firms. If auditor specialists do not provide, or are perceived as not providing higher quality audits,
audit firms may respond by organizing as generalists. Therefore, understanding the relationship between specialization and
quality is important for audit firm’s strategic decisions.
The remainder of the paper is organized as follows. Section 2 contains background information and discusses issues with
the research design of industry specialization studies. Section 3 contains the research design, Section 4 contains the sample
selection and reports descriptive statistics, Section 5 reports the main empirical results, Section 6 reports additional analyses,
and Section 7 concludes.

2. Background and purpose of study

2.1. Auditor industry specialization background

Throughout the 1990s, audit firms increasingly organized and marketed themselves along industry lines (Emerson, 1993).
Consequently, auditor industry specialization has received much attention from researchers. Experimental research gener-
ally finds that industry expertise is associated with better auditor judgment. In particular, experiment participants with
greater industry knowledge have been shown to be more accurate in detecting errors and to have better risk assessment
(Solomon et al., 1999; Owhoso et al., 2002; Taylor, 2000; Low, 2004). Finally, Hammersley (2006) finds that industry special-
ist auditors are better able to identify misstatement patterns.
Archival research has also examined whether industry expertise affects audit quality. Unlike experimental research in
which the researchers can recruit participants with and without industry expertise, archival researchers do not have access
to the level of industry knowledge auditors actually possess. i.e., industry expertise is unobservable at the audit firm, audit
office, and individual auditor level. Archival researchers have therefore used an audit firm’s market share within an industry
as a measure of industry specialization (Balsam et al., 2003; Francis et al., 2005b; Reichelt and Wang, 2010). Initial research
in this area finds that industry specialist audit firms and industry specialist audit offices are associated with higher earnings
quality, and therefore higher audit quality (Balsam et al., 2003; Reichelt and Wang, 2010). There is some research using non-
U.S. settings which finds that audit partners with industry expertise are associated with higher audit quality (Chin and Chi,
2009; Chi and Chin, 2011).5

2.2. Issues with measures of industry specialization (Minutti-Meza, 2013)

However, Minutti-Meza (2013) challenges this literature. He points out that there are both conceptual and econometric
issues with measuring industry expertise based on the auditor’s market-share within an industry. Conceptually, an audit
firm with a small market share may have extensive industry knowledge. Industry knowledge could be obtained through
other means, which the market-share based measure of expertise do not capture. For example, the audit firm could audit
a small number of clients for many years, or could offer training to its individual auditors, or could provide consulting ser-
vices, or could hire industry experts from other audit firms. As well, because the market-share based measures of industry
expertise rely on Compustat, they do not account for private firms. An auditor with a large clientele of private firms will be
classified as a non-specialist if one uses the market-share based measure. Therefore, conceptually there are many reasons
why an auditor with the largest industry market share may not have the highest level of industry expertise.
Minutti-Meza (2013) also points out that using a market-share based measure of auditor industry specialization creates
econometric concerns. Because of the way the variable is constructed, specialists will have larger clients compared to non-
specialists (Minutti-Meza, 2013, 785). This is problematic because large clients tend to have higher pre-audit earnings
quality (Lawrence et al., 2011; Eshleman and Guo, 2014). Minutti-Meza (2013) then uses a propensity-score matching
technique to match specialist auditors with non-specialist auditors and re-examines the relationship between specializa-
tion and quality. Using several measures of audit quality, he finds that industry specialization has no association with audit
quality.

3
For example, see Ernst & Young’s website (https://www.ey.com/en_gl/what-we-do) or KPMG’s (https://home.kpmg/cn/en/home/careers/who-we-are.
html).
4
Investors do appear to consider auditor characteristics such as the level of nonaudit services, auditor tenure and other corporate governance characteristics
when voting on auditor ratifications (Raghunandan, 2003; Dao et al., 2008; Glass Lewis, 2011; Son et al., 2017).
5
There is also a large literature which finds audit fee premiums for industry specialist audit firms, audit offices, and individual audit partners (Ferguson et al.,
2003; Francis et al., 2005a, 2005b; Taylor, 2011; Goodwin and Wu, 2014). We do not discuss this literature as our focus is on auditor industry specialization and
audit quality.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
4 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

2.3. Seasoned specialists (Gaver and Utke, 2019)

Recently, Gaver and Utke (2019) (hereafter, GU) challenge the findings of Minutti-Meza (2013). The authors hypothesize
that not all specialists are equal. Specifically, new ‘‘unseasoned” industry specialists are different from ‘‘seasoned” industry
specialists who have been specialists for a few years. The authors then show that seasoned industry specialists (but not
unseasoned specialists) are associated with higher audit quality.

2.4. Controls for industry effects

One problem plaguing the research on industry specialization and audit quality is the absence of industry fixed effects in
the accruals models. Both Minutti-Meza (2013, footnote 20) and Gaver and Utke (2019, footnote 16) state that they do not
include industry fixed effects in the accruals models because discretionary accruals are estimated by industry.6 However,
there are both conceptual and econometric reasons for including industry fixed effects in the accruals models. Conceptually,
industry specialization studies attempt to provide evidence on whether clients with seasoned industry specialist auditors are
associated with higher audit quality relative to other clients in the same industry who are not being audited by a seasoned
industry specialist. Therefore, the relevant benchmark is other clients in the same industry. If the researchers estimate audit
quality models separately for each industry, there would be no need to control for industry effects. However, this literature
(e.g., Reichelt and Wang, 2010; Minutti-Meza, 2013; Gaver and Utke, 2019) estimates audit quality models on the entire panel
dataset. It is then important to control for the effect of belonging to a certain industry. Accruals quality measures such as dis-
cretionary accruals tend to vary by industry, based on characteristics of the industry, such as operating cycle length (Francis
et al., 2005a). In an important paper, Hribar and Nichols (2007) show that the mean value of the magnitude of discretionary
accruals is an increasing function of the variance of the error term in the first-stage discretionary accrual estimation model.
Therefore, in industries in which the accruals model does not fit the accrual generating process very well, we will tend to
see larger magnitudes of residuals from the first stage model. It is therefore very important to control for industry effects when
the dependent variable is discretionary accruals, whether the absolute value or the signed value.

2.4.1. Purpose of study

The purpose of this study is to therefore re-examine the relationship between auditor industry specialization and audit
quality. We first replicate GU’s findings and then test the sensitivity of their findings to different research design choices. We
are primarily interested in whether the positive association between seasoned industry specialists and audit quality is sen-
sitive to controlling for industry effects. In additional tests, we also examine the sensitivity of GU’s findings to using different
measures of industry specialization. In doing so, we use GU’s first two measure of audit quality: the absolute value of dis-
cretionary accruals and signed discretionary accruals.7

3. Research design

We first replicate GU’s results by estimating the following linear regression:

AQ it ¼ a0 þ a1 SEASONEDit þ a2 UNSEASONEDit þ a3 LOGMKT it þ a4 LEV it þ a5 ROALit þ a6 ROAit þ a7 LOSSit

þ a8 CFOit þ a9 BTMit þ a10 ABS AC LAGit þ a11 GROWTHit þ a12 ALTMANit þ a13 STDEARNit þ a14 TENUREit
þ Year Fixed Effects þ eit ð1Þ

where AQ is an audit quality proxy. Subscripts i and t denote firm and year, respectively. Following GU we use the absolute
value of discretionary accruals (ADA) as our primary measure of audit quality and use positive discretionary accruals (DA) as
an additional measure of audit quality. When we use positive discretionary accruals as our measure of audit quality, we limit
the sample to firm-years with positive discretionary accruals.
Industry specialists are defined as auditors with the largest market share within a 2-digit SIC code industry and whose
market share is more than 10 percent greater than its next largest competitor during the year. We calculate industry market
shares using client sales. Specialization is measured at the national level. One key variable introduced by GU is UNSEASONED,
which equals 1 if the auditor is in its first year of being classified as a specialist in a given industry, 0 otherwise. The variable
of interest is SEASONED, which equals 1 if the auditor is classified as an industry specialist and UNSEASONED is coded zero,
otherwise SEASONED equals 0. In other words, SEASONED equals 1 only for audit firms which have been specialists for at least
one year.8

6
Reichelt and Wang (2010) also do not include industry fixed effects in their discretionary accruals model.
7
GU also use a third measure of audit quality: book-tax differences. We do not use this measure, as it is not commonly used as a measure of audit quality. An
examination of all audit quality studies published in the top 5 accounting journals (The Accounting Review, Journal of Accounting and Economics, Journal of
Accounting Research, Contemporary Accounting Research, and Review of Accounting Studies) during the 2013–2018 time period reveals that out of 44 total
audit quality studies, 0 used book-tax differences as a measure of audit quality.
8
For SAS code used to calculate seasoned industry specialists, see Utke (2018).

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx 5

The choice of control variables is identical to GU and Minutti-Meza (2013) and includes controls for firm size (LOGMKT),
financial leverage (LEV), current and lagged return on assets (ROA and ROAL), an indicator variable for firms reporting losses
(LOSS), cash flows from operations (CFO), the book-to-market ratio (BTM), the absolute value of lagged accruals (ABS AC LAG),
sales growth (GROWTH), Altman’s (1983) Z-score (ALTMAN), the standard deviation of earnings (STDEARN), auditor tenure
(TENURE). The model also includes year fixed effects to control for the fact that audit quality may exhibit intertemporal vari-
ation. We follow Peterson (2009) and cluster standard errors by firm.

4. Sample selection and descriptive statistics

We use GU’s sample selection process using the same sample years (2003–2015). We begin with 118,378 firm-years on
Compustat which have a CIK number to match to Audit Analytics. We delete 21,809 observations with missing Standard
Industrial Classification scheme industry information. We then delete observations which are not audited by a Big 4 auditor
in the current and prior year. After removing firms not headquartered in the U.S. and observations with missing data, we are
left with 28,906 firm-year observations (see Table 1).9
Table 2 reports descriptive statistics for our sample. The mean (median) firm has an absolute value of discretionary accru-
als of 0.0633 (0.0387), which is very similar to the distribution reported by prior research (Minutti-Meza, 2013; Gaver and
Utke, 2019). Approximately 15.49 percent of firm-years in our sample are audited by an industry specialist, with most of
these firms being audited by a seasoned industry specialist. The distribution of the remaining variables is consistent with
prior research.
Table 3 reports Pearson correlations. Consistent with GU, we observe a negative correlation between SEASONED and the
absolute value of discretionary accruals (0.04). Both SEASONED2 and SEASONED3 also exhibit a negative correlation with
the absolute value of discretionary accruals. No independent variable exhibits a very high correlation with our variable of inter-
est (SEASONED). Therefore, multicollinearity is unlikely to cloud any inferences we may draw from the coefficient on SEASONED.

5. Empirical results

5.1. Illustration of the importance of controlling for industry in accruals models

To illustrate the importance of industry membership on discretionary accruals, we report the results from estimating the
discretionary accruals model for four different industries. We report results using two industries which we expect to have
heterogeneous accrual generating processes: (i.) SIC code 99 – Miscellaneous and (ii.) SIC code 73 – Business Services. We
then present results using two industries which we expect to have homogenous accrual generating processes (i.) SIC code
42 –Trucking and Warehousing, and (ii.) SIC code 26 –Paper and Allied Products. We estimate the accruals models by
industry-year. Therefore, the results we report are average coefficient estimates, average t-statistics, and average adjusted
R2 s, averaged over all years available on Compustat with available data (1987–2017).10
Table 4 reports the results. Panel A reports results for the two industries in which we expect the accruals models to be a
poor fit. Firms in SIC code 99 –miscellaneous are expected to have heterogeneous accrual generating processes, which should
reduce the power of the accruals models. This is exactly what we find. The model has weak explanatory power, with an
adjusted R2 of only 3 percent. We observe a similar finding for SIC code 73 – business services, which has an adjusted R2
of only 11.2 percent, suggesting that the variables included in the accruals model can explain only 11.2 percent of the vari-
ation in accruals. Table 4 –Panel B reports the results for the two industries in which we expect the accruals models to be a
good fit. For SIC code 42 – Trucking and Warehousing, we note that the accruals model has high explanatory power, with an
adjusted R2 of 42.8 percent. Similarly, SIC code 26 – Paper and Allied Products has an adjusted R2 of 29.6 percent.
Comparing Panels A and Panels B, we note that, consistent with Hribar and Nichols’ (2007) observation, the standard
deviation of the residual from the accruals model is much higher for the industries in which the accruals model is a poor
fit. For the poor fitting industries, the standard deviation of the residuals ranges from 0.190 to 0.235. In contrast, for the good
fit industries in Panel B, the standard deviation of the residual ranges from 0.073 to 0.077, which is approximately a third of
the variation observed in the poor fitting industries in Panel A. The result of this is that the mean absolute value of the resid-
ual from the accruals models is much higher in poor fitting industries. For example, the mean (median) absolute value of the
residual for SIC code 99 is 0.151 (0.082) compared to just 0.046 (0.028) for SIC code 26.
Panel C of Table 4 reports the results of estimating the audit quality model (Eq. (1)) using the absolute value of discre-
tionary accruals as the dependent variable after splitting the sample into two groups –highly heterogeneous industries
and low heterogeneous industries. Highly heterogeneous industries are those industries in which the standard deviation
of the residual from the 1st stage accruals model is above the median compared to other industries. Those industries in
which the standard deviation is below the median value are considered low heterogeneous industries.11 The results suggest

9
We thank Steve Utke for providing SAS code for sample selection and the computation of seasoned industry specialization measures.
10
We begin in 1987 because we require cash flows from operations and the statement of cash flows was not available before this time.
11
We estimate the accruals models by industry. Because some industries have more firms than others, this results in an unequal number of observations in
the two subsamples.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
6 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

Table 1
Sample selection.

Firm-years on Compustat without missing CIK number 118,378

Less: firm-years with missing SIC (21,809)
Less: firm-years not audited by a Big 4 auditor in the current year (40,680)
Less: firm-years not audited by a Big 4 auditor in the prior year (2,995)
Less: firm-years not headquartered in the U.S. (9,646)
Less: firm-years with missing data necessary to estimate models (14,342)
Final Sample 28,906

Table 2
Descriptive statistics.

Variable N Mean Median Std Dev Min. Max.

ADA 28,906 0.0633 0.0387 0.0840 0.000 1.061
DA 28,906 0.0021 0.0050 0.1051 0.944 1.061
SPECIALIST 28,906 0.1549 0.0000 0.3618 0.000 1.000
SEASONED 28,906 0.1324 0.0000 0.3389 0.000 1.000
UNSEASONED 28,906 0.0225 0.0000 0.1484 0.000 1.000
SEASONED2 28,906 0.1168 0.0000 0.3212 0.000 1.000
UNSEASONED2 28,906 0.0381 0.0000 0.1914 0.000 1.000
SEASONED3 28,906 0.1028 0.0000 0.3037 0.000 1.000
UNSEASONED3 28,906 0.0521 0.0000 0.2222 0.000 1.000
AU CHANGE 28,906 0.0210 0.0000 0.1434 0.000 1.000
LOGMKT 28,906 6.7283 6.7357 1.9060 1.910 11.378
LEV 28,906 0.2001 0.1535 0.2145 0.000 0.988
ROAL 28,906 0.0087 0.0368 0.1866 2.937 0.287
ROA 28,906 0.0244 0.0515 0.1684 0.878 0.310
LOSS 28,906 0.2999 0.0000 0.4582 0.000 1.000
CFO 28,906 0.0622 0.0866 0.1581 0.691 0.352
BTM 28,906 0.4558 0.4179 0.5999 3.144 2.466
ABS AC LAG 28,906 0.0894 0.0629 0.0933 0.002 0.565
GROWTH 28,906 0.1392 0.0756 0.4149 0.699 2.776
ALTMAN 28,906 7.3480 8.0998 4.0934 0.000 14.181
STDEARN 28,906 122.0068 25.4369 301.8595 0.688 2,079.030
TENURE 28,906 0.7243 1.0000 0.4469 0.000 1.000

This table reports the mean, median, standard deviation, minimum, and maximum value for each variable used in the analysis. Continuous independent
variables are winsorized at the 1st and 99th percentile. Refer to Appendix A for variable definitions.

that any significant association between SEASONED and audit quality is limited to the small subsample of observations in low
heterogeneity industries. i.e., the result is limited to a sample in which the accruals model fits the industry better. This high-
lights the importance of considering the goodness of fit of the accruals models.

5.2. Replication of GU and add controls for industry

In this section we replicate GU’s main findings and present results by modifying their model to include industry fixed
effects.
Table 5 – Panel A, Column (1) presents the results of replicating GU’s findings of a negative association between seasoned
industry specialist auditors and the absolute value of discretionary accruals.12 In our regression, the coefficient on SEASONED
is 0.003 with a t-statistic of 1.942. However, when we add controls for industry fixed effects in column (2), the coefficient on
SEASONED is reduced to 0.000 with a t-stat. of 0.154. We perform an F-test to test the joint significance of the coefficients on
the industry fixed effects. The results, reported near the bottom of Panel A, suggest that industry fixed effects are jointly signif-
icant in explaining the absolute value of discretionary accruals (F-stat. = 11.61; p-val. = 0.000). It is worth noting that the sig-
nificance of the industry fixed effects is comparable to that of the year fixed effects (F-stat. = 10.86; p-val. = 0.000). Therefore,
they are just as important as year fixed effects in predicting the dependent variable.
Table 5 – Panel A, Column (3) presents the results of replicating GU’s findings of a negative association between seasoned
industry specialists and positive discretionary accruals.13 However, when we add industry fixed effects to the model in column
(4), the coefficient on SEASONED is reduced to 0.001 and is not statistically significant. An F-test reveals that the coefficients on
the industry fixed effects are jointly significant in explaining positive discretionary accruals (F-stat. = 15.21; p-val. = 0.000).

12
Gaver and Utke (2019) present these results in their Table 8, Model (1). When comparing results, it is important to note that GU present one-tailed p-values,
whereas we present t-statistics.
13
As in GU, we use one-tailed tests for the SEASONED variable.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx 7

Table 3
Correlation matrix.

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
(1) ADA 1.00
(2) DA 0.19 1.00
(3) SPECIALIST 0.04 0.01 1.00
(4) SEASONED 0.04 0.01 0.91 1.00
(5) UNSEASONED 0.02 0.00 0.35 0.06 1.00
(6) SEASONED2 0.04 0.01 0.85 0.93 0.06 1.00
(7) UNSEASONED2 0.02 0.00 0.46 0.16 0.76 0.07 1.00
(8) SEASONED3 0.03 0.01 0.79 0.87 0.05 0.93 0.07 1.00
(9) UNSEASONED3 0.02 0.00 0.55 0.30 0.65 0.11 0.85 0.08 1.00
(10) AU_CHANGE 0.00 0.00 0.01 0.01 0.01 0.01 0.00 0.01 0.00 1.00
(11) LOGMKT 0.23 0.03 0.09 0.08 0.03 0.08 0.03 0.08 0.04 0.02 1.00
(12) LEV 0.03 0.01 0.04 0.04 0.01 0.04 0.01 0.04 0.01 0.02 0.13 1.00

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
(13) ROAL 0.28 0.03 0.03 0.03 0.02 0.02 0.03 0.02 0.02 0.00 0.38 0.00
(14) ROA 0.19 0.14 0.02 0.02 0.02 0.01 0.02 0.01 0.02 0.00 0.35 0.03
(15) LOSS 0.24 0.21 0.03 0.03 0.02 0.02 0.03 0.02 0.02 0.02 0.42 0.03
(16) CFO 0.20 0.16 0.02 0.01 0.02 0.01 0.03 0.01 0.02 0.00 0.38 0.01
(17) BTM 0.09 0.08 0.01 0.01 0.00 0.01 0.00 0.01 0.00 0.00 0.08 0.20
(18) ABS_AC_LAG 0.26 0.09 0.04 0.04 0.02 0.04 0.02 0.04 0.02 0.01 0.22 0.00
(19) GROWTH 0.09 0.04 0.02 0.02 0.01 0.02 0.01 0.02 0.00 0.00 0.00 0.02
(20) ALTMAN 0.09 0.00 0.08 0.08 0.01 0.08 0.01 0.08 0.01 0.02 0.03 0.52
(21) STDEARN 0.04 0.03 0.06 0.06 0.00 0.06 0.00 0.06 0.01 0.00 0.42 0.10
(22) TENURE 0.07 0.03 0.03 0.02 0.01 0.02 0.01 0.02 0.01 0.21 0.14 0.01
Variable (13) (14) (15) (16) (17) (18) (19) (20) (21) (22)
(13) ROAL 1.00
(14) ROA 0.87 1.00
(15) LOSS 0.53 0.56 1.00
(16) CFO 0.74 0.82 0.54 1.00
(17) BTM 0.11 0.10 0.02 0.03 1.00
(18) ABS_AC_LAG 0.47 0.29 0.27 0.18 0.12 1.00
(19) GROWTH 0.06 0.06 0.05 0.08 0.04 0.07 1.00
(20) ALTMAN 0.10 0.05 0.01 0.07 0.14 0.08 0.15 1.00
(21) STDEARN 0.05 0.08 0.04 0.08 0.06 0.05 0.06 0.14 1.00
(22) TENURE 0.08 0.07 0.08 0.05 0.03 0.09 0.11 0.04 0.06 1.00

This table reports Pearson correlation coefficients below the diagonal. Correlations significant at the 5 percent level are bolded. Refer to Appendix A for
variable definitions.

Table 5 – Panels B and C reproduce the analyses in Panel A using alternative seasoning periods, as suggested by GU. We
are able to replicate their results when using a two-year seasoning period, but the results disappear when we add industry
fixed effects (Panel B, column (2) and (4)). We are able to replicate the negative association between SEASONED3 and the
absolute value of discretionary accruals, but this result disappears as well when we add industry fixed effects. Finally, we
are unable to replicate the negative association between SEASONED3 and positive discretionary accruals.14

6. Additional analyses

6.1. Calculating market shares for specialization

In this section we explore whether the association between industry specialization and audit quality changes when using
audit fees to calculate industry market shares when classifying specialists.15 Audoussett-Coulier et al. (2016) demonstrate
that using different industry specialization measures can result in different inferences. The authors conclude that ‘‘Our findings
suggest that audit-fee based measures should probably be prioritized by researchers and that previous empirical findings based
on other measurement variables need to be re-examined” (Audoussett-Coulier et al., 2016, 158). We therefore use audit-fee
based specialization measures, which is quite common (e.g., Francis et al., 2005a, 2005b; Reichelt and Wang, 2010; Cahan
et al., 2011; Chi et al., 2011; Fung et al., 2012; Kim et al., 2015; Bills et al., 2016; Eshleman and Lawson, 2017). SEASONEDFEE
and UNSEASONEDFEE are calculated in the same way as the original seasoned and unseasoned variables, except that we use audit
fee data to calculate the market shares for classifying specialists. We then re-estimate our models.

14
Results are robust to using Fama and French 48 industry classification fixed effects and calculating industry specialization based on market shares
calculated using Fama and French 48 industries (untabulated).
15
Audoussett-Coulier et al. (2016, 140) note that client sales are a surrogate for audit fees, and are typically used in international studies or studies with a
long time-series of data, for which audit fee data are not available.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
8 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

Table 4
Estimating discretionary accruals for different industries.

Panel A -Industries in which the accruals model is a poor fit

SIC Code 99 - Miscellaneous SIC Code 73 - Business Services
Total N 1,334 Total N 19,021
Average N per year 43.032 Average N per year 613.581
AdjustedR2 0.030 AdjustedR2 0.112
Std. Dev. of residual 0.235 Std. Dev. of residual 0.190
Mean absolute value of residual 0.151 Mean absolute value of residual 0.124
Median absolute value of residual 0.082 Median absolute value of residual 0.075
Panel B - Industries in which the accruals model is a good fit
SIC Code 42 - Trucking and Warehousing SIC Code 26 - Paper and Allied Products
Total N 1,193 Total N 1,814
Average N per year 38.484 Average N per year 58.516
AdjustedR2 0.428 AdjustedR2 0.296
Std. Dev. of residual 0.073 Std. Dev. of residual 0.077
Mean absolute value of residual 0.049 Mean absolute value of residual 0.046
Median absolute value of residual 0.036 Median absolute value of residual 0.028
Panel C – Audit Quality Regressions after Splitting Sample based on Std. Dev. of Residual
Low Heterogeneous High Heterogeneous Industries
Industries
Dependent Variable = ADA ADA
Variable Coef. [t-stat.] Coef. [t-stat.]
SEASONED 0.002 [1.459] 0.000 [0.158]
Controls Included Included
Year Fixed Effects Included Included
N 6585 22,321
Adjusted R2 0.158 0.135

*, **, and *** denote statistical significance at the 10, 5, and 1 percent levels, respectively, using two-tailed tests.
Panels A and B report the results of estimating the accruals model used by GU for four different industries. The dependent variable in Panels A and B is total
accruals (ACC it ), defined as income before extraordinary items minus operating cash flows, all scaled by lagged assets. Regressions are estimated separately
for each year, resulting in 31 separate regressions for each industry. Therefore, the adjusted R2 reported are the average values for all 31 regressions. The
table also reports the standard deviation of the residuals, as well as the mean and median values of the absolute value of the residuals.
Panel C reports the results of estimating regressions after splitting the sample into high and low heterogeneous industries. High heterogeneous industries
are those in which the standard deviation of the residual from the accruals model is above the median compared to other industries, which are low
heterogeneous industries. Refer to Appendix A for variable definitions.

Table 6 reports the results. The first four columns report results using a one-year seasoning period, columns (5) through
(8) reports results using a two-year seasoning period, and columns (9) through (12) reports results using a three-year sea-
soning period. Interestingly, no matter how long of a seasoning period we use, we find that we cannot replicate GU’s main
finding of a negative association between seasoned industry specialization and the absolute value of discretionary accruals.
The coefficient on the fee-based measures of seasoned industry specialization is insignificant in Column (1), regardless of the
length of the seasoning period. Adding industry fixed effects in column (2) does not change our inferences. Column (3)
reveals that we are unable to replicate GU’s finding of a negative association between seasoned specialists and positive dis-
cretionary accruals. In column (4) we add industry fixed effects and find a marginally significant but positive coefficient on
SEASONEDFEE , suggesting that seasoned industry specialists are associated with lower audit quality. Taken together, it
appears that the association between seasoned industry specialists and audit quality is sensitive to defining specialization
using client sales rather than audit fees.
Finally, we assess whether GU’s findings are sensitive to using client assets to calculate industry market shares, as in
Minutti-Meza (2013). Table 7 reports the results when using client assets to calculate market shares when defining industry
specialists. In the first four columns, we use a 1-year seasoning period and are able to replicate GU’s main findings in col-
umns (1) and (3). The results disappear when adding industry fixed effects in columns (2) and (4). However, when using
a 2 or 3 year seasoning period in the subsequent columns, we note that we are unable to replicate the negative association
between seasoned industry specialists and the absolute value of discretionary accruals, even without adding industry fixed
effects. The remaining results are similar to those reported in the first four columns. Taken as a whole, the evidence in Table 7
suggests that some of GU’s main findings are sensitive to calculating industry market shares using client assets, as in
Minutti-Meza (2013).16

16
This is of particular concern because GU’s purpose is to challenge the findings of Minutti-Meza (2013). It is therefore important to consider whether the
results are robust to using an approach similar to that of Minutti-Meza (2013).

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx 9

Table 5
Replication of GU and the effect of adding industry fixed effects.

Panel A – 1 year seasoning period

(1) (2) (3) (4)
Dependent Variable = ADA ADA DA DA
Variable Predicted Sign Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.]
SEASONED – 0.003** 0.000 0.002** 0.001
[1.942] [0.154] [1.660] [0.462]
UNSEASONED ? 0.001 0.000 0.001 0.002
[0.295] [0.066] [0.241] [0.656]
Controls Included Included Included Included
Year Fixed Effects Included Included Included Included
Industry Fixed Effects No Included No Included
N 28,906 28,906 15,571 15,571
Adjusted R2 0.144 0.157 0.374 0.399
F-test of Industry FE F-stat. = 11.61. F-stat. = 15.21.
P-value = 0.000 P-value = 0.000
F-test of Year FE F-stat. = 10.86. F-stat. = 9.27.
P-value = 0.000 P-value = 0.000
Panel B – 2 year seasoning period
(1) (2) (3) (4)
Dependent Variable = ADA ADA DA DA
Variable Predicted Sign Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.]
SEASONED2 – 0.003** 0.000 0.003** 0.001
[2.052] [0.172] [1.929] [0.381]
UNSEASONED2 ? 0.001 0.000 0.000 0.002
[0.372] [0.041] [0.047] [0.612]
Controls Included Included Included Included
Year Fixed Effects Included Included Included Included
Industry Fixed Effects No Included No Included
N 28,906 28,906 15,571 15,571
Adjusted R2 0.144 0.157 0.374 0.399
F-test of Industry FE F-stat. = 12.06. F-stat. = 15.14.
P-value = 0.000 P-value = 0.000
F-test of Year FE F-stat. = 10.88. F-stat. = 9.24.
P-value = 0.000 P-value = 0.000
Panel C – 3 year seasoning period
(1) (2) (3) (4)
Dependent Variable = ADA ADA DA DA
Variable Predicted Sign Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.]
SEASONED3 – 0.003** 0.001 0.001 0.003
[1.650] [0.531] [1.026] [1.195]
UNSEASONED3 ? 0.002 0.002 0.000 0.001
[1.239] [1.127] [0.154] [0.416]
Controls Included Included Included Included
Year Fixed Effects Included Included Included Included
Industry Fixed Effects No Included No Included
N 28,906 28,906 15,571 15,571
Adjusted R2 0.144 0.157 0.374 0.399
F-test of Industry FE F-stat. = 11.73. F-stat. = 15.25.
P-value = 0.000 P-value = 0.000
F-test of Year FE F-stat. = 10.86. F-stat. = 9.25.
P-value = 0.000 P-value = 0.000

*, **, and *** denote statistical significance at the 10, 5, and 1 percent levels, respectively. We use one-tailed tests for seasoned variables and two-tailed tests
for all other variables.
This table reports estimated coefficients and t-statistics from OLS regressions. In models (1) and (2), the dependent variable is the absolute value of
discretionary accruals (ADA). In models (3) and (4), the dependent variable is the value of discretionary accruals (DA). Models (3) and (4) are estimated on
only observations with positive discretionary accruals. All regressions include control variables listed in Eq. (1) as well as year fixed effects. 2-digit SIC code
industry fixed effects are included in models (2) and (4). Standard errors are clustered by firm. Panels A, B, and C use different definitions of a seasoned and
unseasoned industry specialist (i.e., SEASONED, SEASONED2, and SEASONED3). The table also reports the F-test of the joint significance of both the year and
the industry fixed effects. Refer to Appendix A for variable definitions.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
10 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

Table 6
Using audit fees to calculate market shares for specialization.

1-Year Seasoning Period 2-Year Seasoning Period 3-Year Seasoning Period

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Variable Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.]
SEASONEDFEE 0.001 0.001 0.002 0.003** 0.002 0.000 0.002 0.003** 0.002 0.001 0.001 0.005**
[0.799] [0.495] [1.068] [1.657] [1.101] [0.073] [1.200] [1.657] [0.808] [0.590] [0.571] [1.993]
FEE
UNSEASONED 0.004 0.004 0.005* 0.006** 0.002 0.003* 0.003 0.004* 0.001 0.002 0.001 0.003*
[1.476] [1.520] [1.687] [1.993] [1.258] [1.714] [1.163] [1.917] [0.730] [1.129] [0.402] [1.676]
Controls and Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE
Industry Fixed No Yes No Yes No Yes No Yes No Yes No Yes
Effects
N 28,905 28,905 15,571 15,571 28,905 28,905 15,571 15,571 28,905 28,905 15,571 15,571
Adjusted R2 0.144 0.157 0.374 0.399 0.144 0.157 0.374 0.399 0.144 0.157 0.374 0.399

*, **, and *** denote statistical significance at the 10, 5, and 1 percent levels, respectively. We use one-tailed tests for seasoned variables and two-tailed tests
for all other variables. T-statistics are based on standard errors clustered by firm. The dependent variable in all regressions is the absolute value of
discretionary accruals, ADA. Refer to Appendix A for variable definitions.

6.2. Local industry specialization and portfolio share approach

In this section we examine whether measuring industry specialization at the audit office level is associated with audit
quality when controlling for industry fixed effects. We define a local specialist (LOCAL SPEC) as the auditor with the largest
market share of an industry within a given metropolitan statistical area (MSA). We report results using sales, audit fees, and
total assets to calculate the market shares. We also report results using a portfolio share approach to calculate industry spe-
cialization. Here, specialization is measured as the total sales of all clients of the auditor of a particular industry divided by
the total sales of all clients of the auditor (PS SPEC).17
Table 8 – Panel A reports the results of re-estimating our audit quality model using local industry specialization measures.
The first two columns reveal that, even without controlling for industry fixed effects, there is no relation between local
industry specialization and audit quality. The next two columns use audit fees to calculate market shares. Consistent with
Reichelt and Wang (2010) and Minutti-Meza (2013), we find a negative coefficient on our measure of local industry special-
ization. However, this coefficient becomes insignificant when we include industry fixed effects in the following column. Col-
umns (5) and (6) report results using total assets to calculate market shares. Similar to the sales-based local specialization
results, we find an insignificant relation between specialists and audit quality with or without industry fixed effects.
Table 8 – Panel B reports the results of re-estimating the signed discretionary accruals model using local measures of spe-
cialization. In general, the evidence suggests that these local measures exhibit a negative relationship with positive discre-
tionary accruals. However, the strength of the relationship is significantly attenuated when including industry fixed effects.
Table 8 – Panel C reports results when using a portfolio share approach to measure specialization. Columns (1) and (2)
report results using client sales to calculate the specialization measure, columns (3) and (4) use audit fees, and columns
(5) and (6) use client assets. Using any approach, the results suggest that the portfolio share based measure of industry spe-
cialization is not significantly associated with the magnitude of discretionary accruals.
Table 8 – Panel D reports results similar to Panel C except here we use signed discretionary accruals. The results are fairly
consistent with Panel C. A noteworthy exception is that when using client assets to calculate portfolio share specialization,
we find that specialists are associated with lower signed discretionary accruals, even when including industry fixed effects
(column 6).

7. Conclusion

In this paper we re-examine the relationship between auditor industry specialization and audit quality. This literature has
several research design choices which cloud inferences one can make. Our first analysis shows how the goodness of fit of the
discretionary accruals models varies by industry, which has consequences for the standard deviation of the residual, making
the inclusion of industry fixed effects very important. We next demonstrate that the prior finding of a positive association
between seasoned industry specialist auditors and audit quality is sensitive to including controls for industry. F-tests indi-
cate that the coefficients on industry fixed effects are jointly significant in explaining the dependent variable (discretionary
accruals). In fact, the industry fixed effects are equally as important as year fixed effects in explaining discretionary accruals.
We next follow the recommendations of Audoussett-Coulier et al. (2016) and re-examine Gaver and Utke’s (2019) main

17
For example, if an auditor had 10 clients, each with $1 in sales, and 7 out of the 10 clients were in the Trucking and Warehouse industry, the value of
PS SPEC would be 0.70 for engagements of Trucking and Warehouse clients. We calculate this measure at the MSA level and not the national level, as the value
of PS SPEC does not exhibit much variation at the national level.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.

Table 7

J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

Using client assets to calculate market shares for specialization.

1-Year Seasoning Period 2-Year Seasoning Period 3-Year Seasoning Period

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Variable Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.]
SEASONEDASSET 0.002** 0.000 0.003*** 0.002* 0.002* 0.001 0.004*** 0.002* 0.002 0.001 0.004*** 0.003*
[1.710] [0.259] [2.410] [1.407] [1.312] [0.491] [2.415] [1.515] [1.075] [0.624] [2.341] [1.603]
ASSET
UNSEASONED 0.003 0.001 0.004* 0.003 0.004** 0.002 0.003 0.002 0.005** 0.002 0.003* 0.002
[1.068] [0.438] [1.736] [1.176] [2.235] [0.958] [1.564] [0.934] [2.487] [1.045] [1.688] [0.883]
Controls and Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry Fixed Effects No Yes No Yes No Yes No Yes No Yes No Yes
N 28,906 28,906 15,571 15,571 28,906 28,906 15,571 15,571 28,906 28,906 15,571 15,571
Adjusted R2 0.144 0.157 0.374 0.399 0.144 0.157 0.374 0.399 0.144 0.157 0.374 0.399

*, **, and *** denote statistical significance at the 10, 5, and 1 percent levels, respectively. We use one-tailed tests for seasoned variables and two-tailed tests for all other variables.
This table reports results from estimating audit quality regressions in which the dependent variable is the absolute value of discretionary accruals, ADA. The seasoned industry specialist variables used in this table
(i.e., SEASONEDASSET and UNSEASONEDASSET ) are computed using market shares which are based on the auditor’s share of client assets from clients in a given industry divided by the total assets of all clients in that
industry on Compustat. T-statistics are based on standard errors clustered by firm. Refer to Appendix A for variable definitions.

11
12 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

Table 8
Local industry specialization measures and portfolio share based measures.

Panel A – Local Industry Specialization Measures and Unsigned Discretionary Accruals

(1) (2) (3) (4) (5) (6)
Dependent Variable ¼ ADA
Variable Coef. Coef. Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.]
LOCAL SPEC 0.001 0.000
[0.938] [0.189]
FEE
LOCAL SPEC 0.001 0.001
[0.629] [0.586]
LOCAL SPEC ASSET 0.002** 0.001
[1.917] [0.684]
Controls and Year FE Included Included Included Included Included Included
Industry FE No Included No Included No Included
N 27,590 27,590 27,590 27,590 27,590 27,590
Adjusted R2 0.143 0.157 0.143 0.157 0.143 0.157
Panel B – Local Industry Specialization Measures and Positive Discretionary Accruals
(1) (2) (3) (4) (5) (6)
Dependent Variable ¼ DA
Variable Coef. Coef. Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.]
LOCAL SPEC 0.003** 0.000
[2.264] [0.330]
FEE
LOCAL SPEC 0.004*** 0.002
[3.215] [1.506]
LOCAL SPEC ASSET 0.004*** 0.002*
[3.515] [1.644]
Controls and Year FE Included Included Included Included Included Included
Industry FE No Included No Included No Included
N 14,939 14,939 14,939 14,939 14,939 14,939
Adjusted R2 0.377 0.404 0.377 0.404 0.378 0.404
Panel C – Portfolio Share Measures and Unsigned Discretionary Accruals
(1) (2) (3) (4) (5) (6)
Dependent Variable ¼ ADA
Variable Coef. Coef. Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.]
PS SPEC 0.001 0.000
[0.546] [0.103]
PS SPEC FEE 0.000 0.000
[0.478] [0.231]
PS SPEC ASSET 0.001 0.001
[0.423] [0.222]
Controls and Year FE Included Included Included Included Included Included
Industry FE No Included No Included No Included
N 27,590 27,590 27,590 27,590 27,590 27,590
Adjusted R2 0.143 0.157 0.143 0.157 0.143 0.157
Panel D – Portfolio Share Measures and Positive Discretionary Accruals
(1) (2) (3) (4) (5) (6)
Dependent Variable ¼ DA
Variable Coef. Coef. Coef. Coef. Coef. Coef.
[t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.] [t-stat.]
PS SPEC 0.003 0.003
[1.364] [1.335]
PS SPEC FEE 0.003* 0.003
[1.364] [1.335]
PS SPEC ASSET 0.003 0.005***
[1.379] [2.582]
Controls and Year FE Included Included Included Included Included Included
Industry FE No Included No Included No Included
N 14,939 14,939 14,939 14,939 14,939 14,939
Adjusted R2 0.377 0.404 0.377 0.404 0.377 0.404

*, **, and *** denote statistical significance at the 10, 5, and 1 percent levels, respectively. We use one-tailed tests for industry specialization variables and
two-tailed tests for all other variables.
This table reports results from estimating audit quality regressions. All regressions include all control variables listed in Eq. (1) as well as year fixed effects.
2-digit SIC code industry fixed effects are included in regressions where indicated. T-statistics are based on standard errors clustered by firm. Refer to
Appendix A for variable definitions.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx 13

findings using audit fees to calculate market shares for classifying industry specialists. Using audit fee-based measures of
specialization, we find no association between seasoned industry specialist auditors and audit quality measured as the mag-
nitude of discretionary accruals or positive discretionary accruals.
This study makes an important contribution to the industry specialization literature. While experimental studies have
provided robust evidence that industry specialists make better decisions and are better able to identify misstatement pat-
terns (Solomon et al.1999; Owhoso et al., 2002; Low, 2004; Hammersley, 2006), the existing archival evidence is mixed.
While theory may predict a positive association between industry specialists and audit quality, it is important for archival
researchers to conduct careful analyses. We argue that when testing for a quality difference, the relevant benchmark for a
firm with an industry specialist auditor is all peer firms in the same industry who are audited by non-specialists. We demon-
strate that failure to control for industry fixed effects is responsible for some of the associations of prior research. Finally, we
also highlight the importance of deciding how to measure industry specialization by demonstrating that several relation-
ships documented by prior literature disappear when using audit fee-based measures of specialization. This need not imply,
however, that industry specialist auditors do not provide higher audit quality. It may be the case that archival research’s fail-
ure to find a positive relationship between specialists and audit quality is attributable to utilizing market share based mea-
sures of specialization. Or it may be the case that specialization is best captured at the audit partner level (Chin and Chi,
2009; Chi and Chin, 2011). We leave it to future archival research to find better measure of specialization at the audit firm
level. Finally, we note that our findings are relevant for the public, as shareholders of many public companies vote on auditor
ratification. Given the evidence that investors do consider auditor characteristics such as nonaudit service fees and auditor
tenure when voting on auditor ratification (Raghunandan, 2003; Dao et al., 2008), it is quite likely that they will consider
whether the auditor is a specialist. Therefore, our findings are important for investors as they make auditor ratification
decisions.

Appendix A. Variable definitions (with Compustat Mnemonics)

Variable Definition
Dependent variables
ADA The absolute value of discretionary accruals, estimated as the absolute value of the residual from
estimating the following model by 2-digit SIC code industry-year:
ACC it ¼ b0 þ b1 ð1=AT it1 Þ þ b2 CHGREV it þ b3 PPEit þ b4 ROAit1 þ eit
Where,
ACC it ¼ Total accruals, defined as income before extraordinary items minus operating cash flows,
all scaled by lagged assets.
AT it1 ¼ Total assets in period t-1.
CHGREV it ¼ The change in total revenues divided by lagged assets.
PPEit ¼ Net property, plant, and equipment divided by lagged assets.
ROAit1 ¼ net income in year t-1 divided by total assets at the beginning of year t-1.
We require at least 10 observations per industry-year to estimate the model.
DA Discretionary accruals, defined as the residual from estimating the equation in the variable
definition above this one.

Variables of interest
SPECIALIST 1 if the auditor has the highest market share in a given industry and that market share is also more
than 10 percent higher than the next largest competitor for the year, 0 otherwise. Each auditor’s
market share is measured as the sum of all of its client’s sales in an industry-year divided by the
total sales of all clients in that industry-year. Specialization is measured at the U.S. national level.
SEASONED 1 if the auditor is classified as a specialist and UNSEASONED is coded zero, otherwise SEASONED
equals 0.
UNSEASONED 1 if the auditor is in its first year of being classified as a specialist in a given industry, 0 otherwise.
SEASONED2 1 if the auditor is classified as a specialist and UNSEASONED2 is coded zero, otherwise SEASONED2
equals 0.
UNSEASONED2 1 if the auditor is in its first or second year of being classified as a specialist in a given industry, 0
otherwise.
SEASONED3 1 if the auditor is classified as a specialist and UNSEASONED3 is coded zero, otherwise SEASONED3
equals 0.
UNSEASONED3 1 if the auditor is in its first, second, or third year of being classified as a specialist in a given
industry, 0 otherwise.

(continued on next page)

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
14 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx

Appendix A (continued)

Variable Definition
SEASONEDFEE 1 if the auditor is classified as a specialist and UNSEASONEDFEE is coded zero, otherwise
SEASONEDFEE equals 0. Note that for this variable, when we calculate industry market shares, we
use audit fees instead of client sales.
UNSEASONEDFEE 1 if the auditor is in its first year of being classified as a specialist in a given industry, 0 otherwise.
Note that for this variable, when we calculate industry market shares, we use audit fees instead of
client sales.
SEASONED2FEE 1 if the auditor is classified as a specialist and UNSEASONED2FEE is coded zero, otherwise
SEASONED2FEE equals 0. Note that for this variable, when we calculate industry market shares, we
use audit fees instead of client sales.
UNSEASONED2FEE 1 if the auditor is in its first or second year of being classified as a specialist in a given industry, 0
otherwise. Note that for this variable, when we calculate industry market shares, we use audit fees
instead of client sales.
SEASONED3FEE 1 if the auditor is classified as a specialist and UNSEASONED3FEE is coded zero, otherwise
SEASONED3FEE equals 0. Note that for this variable, when we calculate industry market shares, we
use audit fees instead of client sales.
UNSEASONED3FEE 1 if the auditor is in its first, second, or third year of being classified as a specialist in a given
industry, 0 otherwise. Note that for this variable, when we calculate industry market shares, we
use audit fees instead of client sales.
SEASONEDASSET 1 if the auditor is classified as a specialist and UNSEASONEDASSET is coded zero, otherwise
SEASONEDASSET equals 0. Note that for this variable, when we calculate industry market shares, we
use client total assets instead of client sales.
UNSEASONEDASSET 1 if the auditor is in its first year of being classified as a specialist in a given industry, 0 otherwise.
Note that for this variable, when we calculate industry market shares, we use client total assets
instead of client sales.
SEASONED2ASSET 1 if the auditor is classified as a specialist and UNSEASONED2ASSET is coded zero, otherwise
SEASONED2ASSET equals 0. Note that for this variable, when we calculate industry market shares, we
use client total assets instead of client sales.
UNSEASONED2ASSET 1 if the auditor is in its first or second year of being classified as a specialist in a given industry, 0
otherwise. Note that for this variable, when we calculate industry market shares, we use client
total assets instead of client sales.
SEASONED3ASSET 1 if the auditor is classified as a specialist and UNSEASONED3ASSET is coded zero, otherwise
SEASONED3ASSET equals 0. Note that for this variable, when we calculate industry market shares, we
use client total assets instead of client sales.
UNSEASONED3ASSET 1 if the auditor is in its first, second, or third year of being classified as a specialist in a given
industry, 0 otherwise. Note that for this variable, when we calculate industry market shares, we
use client total assets instead of client sales.
LOCAL SPEC 1 if the client’s auditor has the largest market share (based on client sales) within an industry in a
given metropolitan statistical area (MSA) in a given year, 0 otherwise.
LOCAL SPEC FEE 1 if the client’s auditor has the largest market share (based on fees) within an industry in a given
MSA in a given year, 0 otherwise.
LOCAL SPEC ASSET 1 if the client’s auditor has the largest market share (based on client assets) within an industry in a
given MSA in a given year, 0 otherwise.
PS SPEC The sum of the sales of all of the auditor’s clients in the same industry as client i, divided by the
sum of all sales from all clients of the auditor.
PS SPEC FEE Calculated as described above, except using audit fees instead of sales.
PS SPEC ASSET Calculated as described above, except using client assets instead of sales or fees.

Control variables
LOGMKT it The natural log of the market value of equity (PRCC F  CSHO).
LEV it Total debt (DLTT) divided by average total assets.
ROALit Net income (NI) in year t-1 divided by average total assets in year t-1.
ROAit Net income in year t divided by average total assets.
LOSSit 1 if net income is negative in year t, 0 otherwise.
CFOit Cash flows from operations (OANCF) divided by average total assets.
BTMit The book value of equity (CEQ ) divided by the market value of equity (PRCC F  CSHO).
ABS AC LAGit The absolute value of total accruals (IB  OANCF) in year t-1 divided by average total assets in year
t-1.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770
 J.D. Eshleman, P. Guo / J. Account. Public Policy xxx (xxxx) xxx 15

Appendix A (continued)

Variable Definition
GROWTHit Sales growth (SALE) from the prior year.
ALTMANit Altman’s (1968) bankruptcy score as modified by Hillegeist et al. (2004), calculated as
1; 000eX =ð1 þ eX Þ,
Where
X ¼ 4:34  0:08  ðWCAP=AT Þ þ 0:04  ðRE=AT Þ  0:10  ðPI þ XINT  IDIT Þ=AT  0:22
ðPRCC F  CSHOÞ=LT þ 0:06  ðSALE=ATÞ
STDEARNit The standard deviation of income before extraordinary items over the prior four years (year t-4 to
year t-1).
TENUREit 1 if the client has had the same auditor for five or more years, 0 otherwise.

References

Altman, E.I., 1983. Corporate Financial Distress: A Complete Guide to Predicting, Avoiding, and Dealing with Bankruptcy. Wiley, New York.
Audoussett-Coulier, S., Jeny, A., Jiang, L., 2016. The validity of auditor industry specialization measures. Auditing: J. Practice Theory 35 (1), 139–161.
Balsam, S., Krishnan, J., Yang, J.S., 2003. Auditor industry specialization and earnings quality. Auditing: J. Practice Theory 22 (2), 71–97.
Bills, K.L., Swanquist, Q., Whited, R.L., 2016. Growing pains: audit quality and office growth. Contemp. Account. Res. 33 (1), 288–313.
Cahan, S.F., Jeter, D.C., Naiker, V., 2011. Are all industry specialist auditors the same?. Auditing: J. Practice Theory 30 (4), 191–222.
Chi, W., Lisic, L.L., Pevzner, M., 2011. Is enhanced audit quality associated with greater real earnings management?. Account. Horizons 25 (2), 315–335.
Chin, C.-L., Chi, H.-Y., 2009. Reducing restatements with increased industry expertise. Contemp. Account. Res. 26 (3), 729–765.
Chi, H.-Y., Chin, C.L., 2011. Firm versus partner measures of auditor industry expertise and effects on auditor quality. Auditing: J. Practice Theory 30 (2), 201–
229.
Choi, J.-H., Kim, C., Kim, J.-B., Zang, Y., 2010. Audit office size, audit quality, and audit pricing. Auditing: J. Practice Theory 29 (1), 73–97.
Choi, J.-H., Kim, J.-B., Qiu, A.A., Zang, Y., 2012. Geographic proximity between auditor and client: how does it impact audit quality?. Auditing: J. Practice
Theory 31 (2), 43–72.
Dao, M., Mishra, S., Raghunandan, K., 2008. Auditor tenure and shareholder ratification of the auditor. Account. Horizons 22 (3), 297–314.
Emerson, J., 1993. KPMG Peat Marwick: setting the new practice framework standard. Profess. Services Rev. 1.
Ernst & Young (E&Y), 2017. Our commitment to audit quality. Available at: https://www.ey.com/Publication/vwLUAssets/ey-audit-quality-report-2017/
$FILE/ey-audit-quality-report-2017.pdf (accessed November 7, 2018).
Eshleman, J.D., Guo, P., 2014. Do Big 4 auditors provide higher audit quality after controlling for the endogenous choice of auditor?. Auditing: J. Practice
Theory 33 (4), 197–219.
Eshleman, J.D., Lawson, B.P., 2017. Audit market structure and audit pricing. Account. Horizons 31 (1), 57–81.
Ferguson, A., Francis, J.R., Stokes, D.J., 2003. The effects of firm-wide and office-level industry expertise on audit pricing. Account. Rev. 78 (2), 429–488.
Francis, J., LaFond, R., Olsson, P., Schipper, K., 2005a. The market pricing of accruals quality. J. Account. Econ. 39, 295–327.
Francis, J.R., Reichelt, K.J., Wang, D., 2005b. The pricing of national and city-specific reputations for industry expertise in the U.S. Audit Market. Account. Rev.
80 (1), 113–136.
Fung, S.Y.K., Gul, F.A., Krishnan, J., 2012. City-level auditor industry specialization, economies of scale, and audit pricing. Account. Rev. 87 (4), 1281–1307.
Gaver, J.J., Utke, S., 2019. Audit quality and specialist tenure. Account. Rev. 94 (3), 113–147.
Glass, Lewis & Co. (Glass Lewis), 2011. Proxy Paper Guidelines 2011 Proxy Season: An Overview of the Glass Lewis Approach to International Proxy Advice
(United States). Glass, Lewis & Co., San Francisco, CA.
Goodwin, J., Wu, D., 2014. Is the effect of industry expertise on audit pricing an office-level or a partner-level phenomenon?. Rev. Acc. Stud. 19 (4), 1532–
1578.
Gul, F.A., Wu, D., Yang, Z., 2013. Do individual auditors affects audit quality? Evidence from archival data. Account. Rev. 88 (6), 1993–2023.
Hammersley, J., 2006. Pattern identification and industry-specialist auditors. Account. Rev. 81 (2), 309–336.
Hillegeist, S.A., Keating, E.K., Cram, D.P., Lundstedt, K.G., 2004. Assessing the probability of bankruptcy. Rev. Acc. Stud. 9 (1), 5–34.
Hribar, P., Nichols, D.C., 2007. The use of unsigned earnings quality measures in tests of earnings management. J. Account. Res. 45 (5), 1017–1053.
Kim, J.-B., Lee, J.J., Park, J.C., 2015. Audit quality and the market value of cash holdings: the case of office-level auditor industry specialization. Auditing: J.
Practice Theory 34 (2), 27–57.
Lawrence, A., Minutti-Meza, M., Zhang, P., 2011. Can Big 4 versus Non-Big 4 differences in audit-quality proxies be attributed to client characteristics?.
Account. Rev. 86 (1), 259–286.
Lopez, D.M., Peters, G.F., 2012. The effect of workload compression on audit quality. Auditing: J. Practice Theory 31 (4), 139–165.
Low, K.Y., 2004. The effect of industry specialization on audit risk assessments and audit planning decisions. Account. Rev., 201–209
Minutti-Meza, M., 2013. Does auditor industry specialization improve audit quality?. J. Account. Res. 51 (4), 779–817.
Owhoso, V.E., Messier, W.F., Lynch, J., 2002. Error detection by industry-specialized teams during the sequential audit review. J. Account. Res. 40, 883–900.
Peterson, M., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Rev. Financial Stud. 22, 435–480.
Public Company Accounting Oversight Board (PCAOB), 2015. Concept Release on Audit Quality Indicators. Release No. 2015-005. PCAOB, Washington, D.C.
Raghunandan, K., 2003. Nonaudit services and shareholder ratification of auditors. Auditing: J. Practice Theory 22 (1), 155–163.
Reichelt, K.J., Wang, D., 2010. National and office-specific measures of auditor industry expertise and effects on audit quality. J. Account. Res. 48 (3), 647–
686.
Solomon, I., Shields, M., Whittington, O., 1999. What do industry-specialist auditors know?. J. Account. Res. 37, 191–208.
Son, M., Song, H., Park, Y., 2017. PCAOB inspection reports and shareholder ratification of the auditor. Account. Public Interest 17 (1), 107–129.
Taylor, M., 2000. Bounded rationality, uncertainty, and competence: The effects of industry specialization on auditors’ inherent risk assessments and
confidence judgments. Contemp. Account. Res. 17 (4), 693–712.
Taylor, S.D., 2011. Does audit fee homogeneity exist? Premiums and discounts attributable to individual audit partners. Auditing: J. Practice Theory 30 (4),
249–272.
Utke, S., 2018. Calculating Auditor Industry Specialization Tenure: Code from Gaver and Utke (2018). Available at: https://papers.ssrn.com/sol3/papers.cfm?
abstract_id=3212035.

Please cite this article as: J. D. Eshleman and P. Guo, Do seasoned industry specialists provide higher audit quality? A re-examination, J.
Account. Public Policy, https://doi.org/10.1016/j.jaccpubpol.2020.106770