Professional Documents
Culture Documents
This Content Downloaded From 83.253.247.117 On Wed, 26 Apr 2023 09:05:09 UTC
This Content Downloaded From 83.253.247.117 On Wed, 26 Apr 2023 09:05:09 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
American Accounting Association is collaborating with JSTOR to digitize, preserve and extend
access to The Accounting Review
Jean C. Bedard
Bentley College
Diana Falsetta
University of Miami
ABSTRACT: This study investigates auditor-provided tax services from 2000 to 2002
during which fees paid to auditors for nonaudit services (NAS) were disclosed, but
separate disclosure of tax service fees was voluntary. We examine changes in t
market for tax NAS in 2002, as Congress debated possible prohibition of these ser-
vices. Using the Heckman MLE procedure to control for selection bias in tax fee dis
closure, we find that a strong pre-2002 positive association between tax fees an
higher than expected audit fees weakened significantly in 2002, suggesting that so
companies paying high audit fees reduced or terminated auditor-provided tax service
in 2002. Our findings also suggest that auditor-provided tax services were reduced
among new or short-tenure clients. Controlling for tax complexity, we also find som
evidence that early "returns" (i.e., tax rate reductions) to companies from auditor-
provided tax services were reduced in 2002. Finally, results of our selection model imp
that the decision to voluntarily disclose fees paid to the auditor for tax services is
positively associated with tax complexity and auditor tenure, and negatively associate
with the proportion of NAS fees to total fees. Overall, our findings imply significan
change in auditor-provided tax services prior to 2003, when separate disclosure
auditor-provided tax service fees was mandated.
The authors are grateful to Dan Dhaliwal and two anonymous reviewers for their helpful comments
John Graham for providing marginal tax rate data used in this study. We also thank Elizabeth Conn
Dunbar, Julie Hertenstein, Ganesh Krishnamoorthy, LeAnn Luna, Mario Maletta, Jim Maroney, Lillian Mi
Phillips, George Plesko, Tim Rupert, Marjorie Shelley, Pete Wilson, Arnie Wright, and participants of
forums at Australia National University, Boston College, Texas A&M University, and the University of W
as well as those attending the University of Oklahoma Accounting Research Conference and the 2005
Region and American Accounting Association Annual Meetings, for comments on earlier versio
manuscript.
1095
I. INTRODUCTION
his study examines the market for auditor-provided tax services durin
a period in which Congressional debate on possible prohibition of thes
likely contributed to fundamental changes in this market. In general, nonaud
(NAS) provision by auditors is controversial due to potential positive and neg
In particular, joint provision of audit and tax services may be beneficial due t
achieved by having a single professional services firm serve as both auditor
sultant. However, potentially damaging outcomes of joint service provision
promising audit quality and risking shareholder value.
Our study period represents a crucial time in the audit and tax services ind
scale financial frauds focused the attention of the Securities and Exchange
(SEC), and ultimately the U.S. Congress, on the issue of audit quality. One im
in the ensuing debate was whether provision of NAS, including tax services
auditor would be allowed or prohibited. Following extensive discussion rega
provision of various NAS, the final version of the Sarbanes-Oxley Act o
allowed auditor-provided tax services, but disallowed other specific NAS.
We study three basic issues during the limited period that immediately
includes the Congressional debate over SOX in 2002, which may under-repr
implementation of changes in service relationships resulting from SOX legisla
sequent regulations. However, this period does provide an initial look at firm
expected changes in auditor-provided tax services coincident with the SOX
we examine whether the association between tax fees and audit fees change
period of increased pressure on corporations to separate their audit and tax s
ers. Second, we investigate whether the association between tax fees and the
client/auditor relationship changed over the same period. This question is imp
SOX specifically includes auditor tenure as a factor relevant to auditor indep
(Section 207). While using separate tax and audit service providers shoul
association between tax fees and auditor tenure, limited prior research (Bec
finds no association between auditor tenure and tax service fees. Third, we exa
the association between tax service fees and their benefits (as measured by
decline in tax rates) changed during this period. We expect that increasing
impending regulatory change likely altered the product mix in the tax servi
2002, reducing the association between tax fees and tax savings. In studying
we address a specific type of NAS that is interesting and important due to
characteristics. First, in contrast to other NAS (e.g., systems development), ta
have a direct and immediate impact on client income and cash flows throug
reduction.2 Second, tax services may also affect public policy in terms of t
impact on tax equity. Thus, our service-specific approach complements oth
In this paper, "nonaudit services" (NAS) refer to services, other than the financial statement a
a company's auditor. "Tax fees" refers to fees paid to auditors for tax services.
2 Phillips et al. (2003) examine earnings management through deferred tax expense. They docum
tax expense is incrementally useful over total accruals and discretionary accruals in detecting ear
ment to avoid an earnings decline or loss. Their results suggest that deferred tax expense i
earnings management.
3 As noted by Firth (2002), a positive association of audit and NAS fees need not imply auditor independence
concerns if this association is due to firm-specific factors that increase the cost of both services. However,
these characteristics are controlled for and the association remains strong, auditor-client economic bonding
suggested.
Research on the association between NAS and indicators of financial reporting quality
yields inconsistent findings, with some studies (Frankel et al. 2002; Ferguson et al. 2004
Reynolds et al. 2004) suggesting that NAS provision is associated with lower earning
quality, while others studies find no association between earnings quality (or earnings re-
statements) and NAS provision (e.g., Ashbaugh et al. 2003; Chung and Kallapur 200
Raghunandan et al. 2003; Kinney et al. 2004). In contrast, Larcker and Richardson (2004)
find that higher audit and NAS fees are associated with higher earnings quality, concludin
that reputation effects constrain auditors from allowing clients to manage earnings. Thu
recent research is mixed on whether NAS compromise auditor independence.
Hla: Tax fees paid to auditors are positively associated with audit fees that are greater
than expected given client size and risk characteristics.
HIlb: The association of tax fees with audit fees will weaken in 2002.
4 However, some concern is raised by Favere-Marchesi (2006), who finds that auditors providing tax services to
clients assess a lower risk of fraud than auditors who do not.
H2b: The association of tax fees with the length of the auditor-client relationship will
change in 2002.
Hypotheses 3a and 3b consider the association between investment in tax NAS and
subsequent reductions in tax rates, and changes in this association in 2002, respectively.
Mills et al. (1998) find that investment in tax services (performed by auditors, nonauditors,
and in-house personnel) among large corporations in 1991 is positively associated with tax
complexity and negatively associated with tax rates. Consistent with this research, H3a
proposes that companies invest in tax planning to reduce future tax liabilities. Thus, higher
tax fees should be associated with declining future tax rates, as the effect of the purchased
tax strategies becomes evident. Hypothesis 3b proposes a weakening of this association in
2002, due to increasing attention by regulators and the financial press on auditor indepen-
dence, audit quality, and questionable tax positions.
H3a: Higher tax fees paid to auditors are associated with reductions in future tax rates.
H3b: The association of tax fees with reductions in future tax rates will weaken in
2002.
III. METHODS
Description of the Sample and Variable Definitions
Sample Derivation and Dependent Variable
Data for this study are obtained from Standard & Poor's Audit Fee Data (2
Compustat Annual Industrial 2003, and Graham's (1996a, 1996b) estimated ma
rates (1980-2003). During our study period, registrants were required to repo
to auditors in three categories: audit, audit-related, and other. The latter catego
tax services as well as other NAS. Many firms provided an additional fee categ
ically noting tax fees paid to the auditor. Others provided detail on the compo
"other" category in notes accompanying the proxy statement disclosure. Our
variable, lnTAXFEE, is the natural log of tax fees paid to auditors, as reporte
sources.
Firms Repor
Tax Fee Inform
Total Sample (Disclosers
Firm-Years Firms Firm-Years Fi
a This table presents the derivation of the sample used in the hypothesis test
z bThe numbers do not sum to the first column because some firms did not
S cFirms must have at least two years of available marginal tax rate informa
0o
Control Variables
One advantage of studying tax services as a specific NAS is the ability to control for
firm characteristics affecting investment in tax services, to improve the power of our hy-
pothesis tests. We use several variables to measure the tax complexity of our sample firms.
First, firms with relatively higher deferred taxes are likely to pay higher fees for advice
successful in deferring liabilities to future periods. DEFTAX is a continuous variable mea-
suring the relative amount of Deferred Taxes (Compustat item #50) to Total Tax Expense
(Compustat item #16). Second, firms involved in mergers or acquisitions are likely to need
advice on the tax implications of changes in their firm structure and limitations imposed
on effectively using the tax attributes of the merging entities or the acquired firm. MERGER
is a dichotomous variable equal to 1 if the firm engaged in a merger, acquisition, or other
corporate restructuring, and 0 otherwise. Third, firms doing business in foreign jurisdictions
likely need the assistance of tax professionals in preparing multiple jurisdiction tax returns
as well as determining appropriate timing of repatriation of earnings. FOREIGN is equal
to 1 if taxes were paid to a foreign jurisdiction (Compustat item #64), and 0 otherwise.
5 Plesko (2003) questions the Graham marginal tax rate measure because of its limited distribution characteristics.
We acknowledge this possible limitation but note that the Graham measure is used extensively in the literature
and is one of the only available sources of these estimates.
Empirical Models
The discussion above suggests the f
InTAXFEE = xo + oxUNEXPECT
+ o2UNEXPECTED AUDIT F
+ O4TENURE X YEAR 2002
+ ot6AUDITOR CHANGE X
+ "sCHANGE MTR(ETR) X
+ a" oMERGER + aoqFOREIG
+ "14EXPATRIATE + as5YEA
+ oa21nASSETS + a22LEVERA
+ o24-87INDUSTRY + u. (1)
DISCLOSE = 80 + ?1 UNEXPECTE
+ 83AUDITOR CHANGE + 8
+ 86FOREIGN + s7NOL + 8,
+ "8olnASSET + 8s1SEGMEN
+ "13LEVERAGE + 814NAS/
+ 816YEAR 2002 + u.
(2)
We view the selection model as preliminary to our primary goal of testing hypotheses
on the association of the level of tax fees with other variables. Thus, we provide only a
brief discussion of its results in the next section, prior to providing results of our outcome
model (Model (1)). As an alternative to the full sample analysis, we apply Heckman MLE
to a subsample, containing only disclosers and a smaller sample of nondisclosers that paid
the auditor for tax services. Thus, in the subsample model, selection bias is narrowed to
the disclosure decision only. This approach conforms to advice by Wooldridge (2002), who
notes complications introduced by multiple possible sources of selection bias. To accom-
plish this, we use the S&P 2003-2004 Audit Fee File to identify firms that paid tax fees
to auditors in 2002, but did not report those fees in that year.9 The subsample contains 300
nondisclosing firms that paid tax fees to their auditor in 2002, combined with 2,405 firm-
years disclosing tax fees paid. This procedure specifically addresses the issue of selection
bias regarding the decision to disclose fees rather than the combined decision to purchase
tax services from the auditor and disclose.
IV. RESULTS
Descriptive Statistics
Table 2 reports descriptive statistics for our sample. Table 2, Panel A, indicat
the mean (median) fee paid by firms reporting auditor-provided tax services appr
$787,000 ($173,000). Comparing tax fees reported by disclosing firms for each y
tabled), we find that the median fee declines significantly through the sample p
($256,880 in 2000, $191,000 in 2001, and $145,150 in 2002; Pearson Chi-square
8 The addition of relative NAS fees to the selection model accomplishes the frequently cited need to
additional variables not contained in the outcome model, in order to avoid multicollinearity. Madd
and Wooldridge (2002) note that it is not technically necessary for the selection equation to contain
variables, and caution that adding inappropriate variables to the selection equation can leave the issu
selection unaddressed. Stolzenberg and Relles (1997) note that independent variables in the two mo
often identical in the sociology literature, and show that the Heckman procedure may not be superi
even with a large sample size. Based on this caution, we estimate our models several different way
Heckman approach with the addition of relative NAS fees, which might affect disclosure but have low
with the outcome equation dependent variable; (2) the Heckman approach using only the variables re
tax complexity, auditor relationship, and controls, which we theorize are related to both selection and
(3) censored normal regression, which assumes a random selection process; and (4) OLS regression. W
inferences across the three approaches remain unchanged, control for selection bias using both Heckma
produces some changes in model coefficients, including increased significance of tax complexity. T
similarity of results across these approaches supports the validity of our findings.
9 Identification is possible because with mandatory disclosure in 2003, firms were also required to r
payments.
ao Descriptive Statistics
Firms Reporting F
ch
a
Tax Fee Reporting T
Information (Ful
cz Full Sample (Disclosers) Nondisc
Variable Description (n = 5,727) (n = 2,405) (n = 3
Panel A: Continuous Variablesa
t,
1\
TAX FEE Fee paid to auditor for tax NA 787 (2
services (thousands) 173
AUDIT FEE Fee for financial statement 848 (2,110) 1,127 (2,824) 64
audit (thousands) 285 334
TENURE Number of years financial 8.98 (9.30) 9.92 (9.45) 8.3
statement audit performe
by the current auditor
CHANGE MTR Percentage change in estimated -1.41% (12.3) -1.38% (12.4) -1
marginal tax rates in the 0%
following period (Graham)
CHANGE ETR Percentage change in effective -1.91% (5.92) -2.19% (6.00) -1
tax rates in the following -0.45% -0.94%
period
DEFTAX Current deferred tax expense 0.53 (23.4) 0.45 (18.5)
divided by total tax expense 0.14 0
SEGMENTS Number of business segments 4.63 (3.16) 4.84 (3.29) 4
4 4 4
ASSETS Total assets (in millions) 6,199 (40,425) 9,311 (57,429) 3,948
433 523 3
LEVERAGE Total liabilities + Total assets 0.56 (0.44) 0.56 (0.34) 0
0.55 0.56 0.
Pretax ROA Pretax income - Total assets 0.01 (0.76) 0.01 (0.33) 0
0.05 0.04 0.
Ch
O
O
TABLE 3
Selection Model Results: Heckman Maximum Likelihood Method
effectiveness. Results of our hypothesis tests are as follows. Hypothesis la predicts that
investment in auditor-provided tax services is positively associated with audit fees that are
greater than expected in 2000-2001, given client characteristics. The positive and significant
coefficient (p = 0.00) on UNEXPECTED AUDIT FEE supports Hla. The negative and
significant coefficient on the interaction of UNEXPECTED AUDIT FEE with YEAR 2002
(p = 0.02) supports Hlb, which suggests that the positive association between tax-related
NAS fees and audit fees weakens in 2002. The marginal effects for these variables, con-
ditional on tax fee reporting, imply a 62.19 percent increase in tax fees for a one standard
deviation increase in unexpected audit fees prior to 2002, and a 37.79 percent increase
cs
cs
O InTAXFEE = ao + a-UNEXPECTED AUDIT FEE + 2%UNEXPECTED AUDIT FEE X
X
+ "4TENURE X YEAR 2002 + ot5AUDITOR CHANGE + u6AUDITOR CH
3t~
+ "X8CHANGE MTR(ETR) X YEAR 2002 + g9DEFTAX + ao0MERGER +
ch
N.
+ o"4EXPATRIATE + ,15YEAR 2002 + al6-20AUDITOR IDs + a211nASSE
+ a24-87INDUSTRY
O A
cs
This table presents results of Heckman MLE outcome models, explaining the natural log of fees paid to auditors for
subsequent year change in marginal tax rates, tax complexity, and control variables. Column A presents results of t
tax rates as alternate measures of tax service effectiveness. Column B presents results of the subsample, which con
their auditor for tax services.
*, **, *** Indicate significance at 10 percent, 5 percent, and 1 percent, respectively, in a one-tailed test if directi
otherwise.
aVariables are defined in Table 2. Dichotomous variables for industry and for individual Big 5/4 audit firms are incl
efficiency of presentation.
b Values reported are the unstandardized regression coefficients.
Ch
~
O
X
;S
~=C.
;S
3t~
Ch
O
C,
Ch
Y
O
O
0\
" We evaluate conditional marginal effects at the mean of the distribution. We report the conditional effects (i.e.,
only for firms reporting tax fees), as the economic impact is more meaningful among firms whose tax service
payments are known.
12 These results are obtained excluding firm-years that involved auditor changes from Arthur Andersen LLP in
2002 due to the possibility that measurement error is associated with the unusual circumstances surrounding the
"flight" from Andersen. When the 2002 Arthur Andersen switches are included in the sample, the results for
the AUDITOR CHANGE and AUDITOR CHANGE X YEAR 2002 are statistically unchanged although coefficients
and p-values are slightly different. However, the 2002 switches from Arthur Andersen to other auditors exhibit
greater reduction in tax fees paid to the new auditor than do other auditor switches.
Supplementary Analyses
We perform two supplementary a
following the passage of SOX in 2002
ued to decline, or increased once it
prohibited. Either result is possible,
maintained during 2003, the final v
tax services by auditors. Using the
data for 1,124 of the 1,182 sample
The mean (median) tax fee paid by t
mean (median) tax fee reported in
(p = 0.00). This suggests that altho
factors such as the need for audit co
to depress this market.
Second, we investigate whether inve
with evidence of earnings managemen
is an association between tax fees an
test by first computing abnormal ac
ample, Francis et al. 2005).13 When
Model (1), we do not find a significa
and earnings management, either bef
on the possible effects of high tax
important to note that our results in
rates. This implies that firms buying
from the decline in tax rate, and th
in other ways. Further research shou
The Accounting R
This study examines factors associated with investment in auditor-provided tax services
during 2000-2002, a period of change in the regulatory climate for NAS and for tax servi-
ces specifically. While there has been research directed toward joint audit/NAS provision
since disclosure of U.S. audit and NAS fees in 2000, there is little evidence on tax services
specifically, despite its importance to companies, providers, and the public interest. Thi
issue is important because while the tax services are not among prohibited NAS, thi
exemption has not been adequately examined in light of the concerns over auditor inde-
pendence.14 Research findings on the association between total NAS and auditor indepen
dence are mixed, with some evidence provided by Kinney et al. (2004) that tax services
may differ from other types of NAS in this regard. Our research provides three main
findings.
First, we find a strong positive association between tax fees and unexpected audit fees
during 2000-2001 and a significant reduction in the strength of this positive association in
2002. We also find that that during 2000-2001, new clients paid significantly greater
tax fees than continuing clients; however, in 2002, shorter-tenure clients paid lower tax
fees. Overall, these findings suggest that the reduction in association between audit and
tax service fees in 2002 was due to new and relatively short-tenure clients terminating tax
services (or not initiating new services), while services provided by longer-tenure auditors
were maintained despite growing pressure by regulators to seek tax services elsewhere. Two
prior studies are important in interpreting this result. Kinney et al. (2004) find that higher
tax fees paid to auditors are associated with fewer earnings restatements. In addition, Myers
et al. (2003) find that longer auditor tenure is associated with higher earnings quality.
Combining these results with ours, the evidence suggests that service quality is the main
driver of joint audit/tax service provision, rather than attempts by clients to "buy" the
auditor's acquiescence. This finding supports the PCAOB's decision not to ban auditors
from performing all tax services for their clients, but to limit the prohibition to certain tax
services likely to be more problematic. Our supplemental comparison of 2003 tax fees paid
by the 2002 reporters in our sample shows a continued decline in this market, even after
the final version of SOX did not prohibit auditors from providing tax services to clients.
Our second main finding relates to the association between tax fees and subsequent
changes in tax rates. Our results imply that during 2000-2001, higher tax fees were asso-
ciated with greater reductions in future tax rates. However, we find some evidence that this
association weakened during 2002, in which tax fees paid were not associated with future
declines in tax rates (controlling for firm size, income, and tax complexity), implying no
or reduced "return" to those expenditures. Although we cannot directly observe the nature
of tax services provided, our interpretation of these findings is that the services provided
in 2002 were less likely to produce reduction in tax rates.
Our third main set of findings is derived from our selection model (Model (2)). The
significant selection bias parameter provides evidence that the voluntary disclosure of tax
fees during the early 2000s was not a random event. As has been found in other financial
disclosure contexts, the disclosure decision varies depending on firm-specific and auditor-
specific factors. For instance, the difference in significance of the effects of firm size and
14 At this writing the PCAOB has proposed further restrictions on the provision of tax services by auditors, which
have yet to be ratified by the SEC. See http://www.pcaobus.org/News andEvents/News/2004-12-14.asp for
further discussion of the proposed ban.
Limitations
Our findings are limited by features of our data and method. First, reporting of tax fees
paid to auditors was voluntary during our sample period, and disclosure of tax fees paid to
nonaudit providers was not required. We attempt to control for this selection bias using the
Heckman procedure. While control may not be complete, one of the significant contribu-
tions of the current study is to show that during a period of regulatory uncertainty, the
provision of tax services by auditors changed from previous patterns. Thus, use of voluntary
disclosure data is necessary in order to address our research questions, and this study
provides a foundation for further studies of tax fees under mandatory reporting. Second,
while we measure tax rate changes over a single year, tax-planning effects are likely to
occur over longer horizons. This implies that the observed relationship between change in
tax rates and current tax fees may understate the tax benefit of fees paid, especially for
effective tax rates. Finally, we observe the tax services market over a limited period, and
thus may not fully observe the implementation of changes in service relationships resulting
from SOX legislation and subsequent regulations.
REFERENCES
Abdel-khalik, R. 1990. The jointness of audit fees and demand for MAS: A self-select
Contemporary Accounting Research 6 (Spring): 295-322.
Alam, P., and A. Baez-Diaz. 2005. The joint provision of audit and non-audit services:
the U.S. banking industry. Working paper, Kent State University, Kent, OH.
Antle, R., E. A. Gordon, G. Narayanamoorthy, and L. Zhui. 2002. The joint determi
fees, NAS fees and abnormal accruals. Working paper, Yale University, New Hav
Ashbaugh, H., R. LaFond, and B. W. Mayhew. 2003. Do NAS compromise auditor ind
Further evidence. The Accounting Review 78 (July): 611-639.
Beck, P. J., T. Frecka, and I. Solomon. 1988. An empirical analysis of the relationship
involvement and auditor tenure: Implications for auditor independence. Journal o
Literature 7: 65-83.
Bell, T., W. Landsman, and D. Shackleford. 2001. Auditors' perceived business risk and audit fees:
Analysis and evidence. Journal of Accounting Research 39: 35-43.
Brandon, D., A. Crabtree, and J. Maher. 2004. Nonaudit fees, auditor independence, and bond ratings.
Auditing: A Journal of Practice & Theory 23: 2 (September): 89-103.
Carcello, J. V., D. R. Hermanson, and T. L. Neal. 2002. Disclosures in audit committee charters and
reports. Accounting Horizons 16 (4): 291-304.
Chung, H., and S. Kallapur. 2003. Client importance, nonaudit services, and abnormal accruals. The
Accounting Review 78 (October): 931-955.