Вы находитесь на странице: 1из 24

Auditor Independence andAuditor-Provided

Tax Service: Evidence from Going-Concern


Audit Opinions Prior to Bankruptcy Filings
Dahlia Robinson
SUMMARY: This study examines whether auditors provision of tax services impairs
auditor independence by focusing on auditors going-concern opinions among a sample
of bankruptcy ling rms. The evidence from the bankruptcy setting is particularly sa-
lient given that the bankruptcy of corporations such as Enron motivated several provi-
sions of the Sarbanes-Oxley Act SOX of 2002. More recently, auditors provision of tax
service to their audit clients has been the focus of new rules by the Public Company
Accounting Oversight Board PCAOB. Consistent with improved audit quality from
information spillover, the study documents a signicant positive correlation between the
level of tax services fees and the likelihood of correctly issuing a going-concern opinion
prior to the bankruptcy ling. One implication of this result is that restricting tax services
by auditors of poorly performing rms may diminish the quality of auditors reporting
decisions without leading to an improvement in auditor independence.
Keywords: auditor independence; tax service fees; going-concern qualication; infor-
mation spillover.
Data Availability: The data are available from public sources cited in the paper.
INTRODUCTION
Investor condence in the quality and integrity of publicly available corporate nancial in-
formation prepared by public companies and certied by independent auditors is considered vital
to the proper functioning of U.S. capital markets. Beginning in the late 1990s, the Securities and
Exchange Commission SEC became increasingly concerned that the growth of nonaudit services
provided to audit clients could undermine that condence. The SEC feared a potential erosion of
auditor independence caused by economic self-interests of audit rms in the provision of nonaudit
services NAS. The perceived link between the provision of NAS and degraded independence
ultimately encouraged the SEC to impose new rules restricting certain types of NAS that audit
rms could provide for their audit clients and requiring new proxy disclosures by registrants of the
magnitude of audit fees and several categories of nonaudit fees SEC 2000.
Dahlia Robinson is an Assistant Professor at the University of South Florida.
This paper benets from comments from Mark Dawkins, Lisa Gaynor, Donald Goldman, Yuchang Hwang, Jennifer Kahle,
Henock Louis, Michael Mikhail, Jackie Reck, Philip Reckers, Michael Robinson, Nathan Stuart, workshop participants at
Arizona State University, and two anonymous reviewers.
Auditing: A Journal of Practice &Theory American Accounting Association
Vol. 27, No. 2 DOI: 10.2308/aud.2008.27.2.31
November 2008
pp. 3154
Submitted: February 2007
Accepted: May 2008
Published Online: March 2009
31
With the adoption of these new SEC nal rules in early 2001, concerns about auditor inde-
pendence and nonaudit service diminished until the much-publicized bankruptcies of Enron,
WorldCom, and Global Crossing
1
catapulted these two issues to the forefront depicting the audit
rm as a collaborator in deception Reynolds et al. 2004, 29 and motivated several provisions
of the Sarbanes-Oxley Act SOX of 2002 Oates 2002; Reynolds et al. 2004. In commentaries on
these failures, the press highlighted the large numbers of bankruptcy lings in which auditors
failed to issue a going-concern qualication in the audit opinion issued immediately prior to the
lings, and often cited these cases as evidence of lack of auditor independence Hilzenrath 2001;
Weil 2001, 2003; Dietz 2002; Venuti 2004. Further, this diminished independence was often
linked to the provision of nonaudit services Weil 2003; Venuti 2004; Lindberg and Beck 2004.
2
Given this perceived link between degraded auditor independence and the provision of NAS in
bankruptcy ling rms, this study examines the association between auditor independence prox-
ied by auditors issuance of going-concern opinions and auditor-provided tax services among
bankruptcy-ling rms.
Several studies investigating the relation between auditor independence and NAS infer auditor
independence by the clients nancial reporting quality, generally proxied by the level of account-
ing accruals or whether the rm meets/beats earnings forecasts for example, Frankel et al. 2002;
Antle et al. 2002; Ashbaugh et al. 2003; Chung and Kallapur 2003; Francis and Ke 2003;
Reynolds et al. 2004; Larcker and Richardson 2004. One clear advantage of utilizing the going-
concern modied audit opinion compared to accounting accruals or meeting/beating earnings
forecasts is that the going-concern opinion is a direct outcome of the auditors decisions and
actions taken during the audit.
A number of other studies for example, Sharma and Sidhu 2001; DeFond et al. 2002; Geiger
and Rama 2003 also utilize the going-concern decision to investigate whether the provision of
NAS impairs auditor independence. However, by looking at nonaudit fees in the aggregate, these
studies seem to imply that all nonaudit services have the same impact on the audit decision.
However, Francis 2004 suggests that tax service is the one area of nonaudit service that would be
likely to benet the audit. Accordingly, in this study, I disaggregate nonaudit fees into a tax and a
nontax component.
My focus on the tax service component of nonaudit services is also motivated by recent SEC
rules and SOX provisions that allow auditor-provided tax services while banning several types of
other nonaudit services. Implicitly, this suggests that regulators and legislators believe that auditor-
provided tax services are less likely to impair independence, and may even improve the audit
through knowledge transfer. As the SEC itself acknowledges, the review of the registrants
tax returns and reserves require substantial knowledge about the audit client Sage and Sage
2005, 31.
Simunic 1984 provides a model of knowledge spillovers or cost interdependencies that
relate to client-specic spillovers in which NAS such as tax service may complement the audit
and enhance audit quality. Such knowledge spillovers occur because the audit and tax services
require elements of the same information set and/or the same professional qualications. In fact,
Francis 2004, 363 suggests that the one area of nonaudit service in which knowledge spillover
would seem plausible is the joint preparation of tax returns and the audit of tax-related accounts
in the nancial statements. Consistent with information spillover, Kinney et al. 2004 provide
1
Other large corporations ling for bankruptcy during this period include Adelphia Communications, Conseco Inc., UAL
Corp., Pacic Gas and Electric, Kmart Corp., The Finova Group, NTL, Reliance Group Holdings, and Federal Mogul.
These 12 corporations had combined assets of $381 billion at the time of ling Venuti 2004.
2
Astudy by Bloomberg News reports that in 54 percent of the 673 largest bankruptcies of public corporations since 1996,
auditors provided no warnings in annual nancial statements issued in the months before bankruptcy Dietz 2002.
32 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
some empirical evidence that nancial reporting quality measured as rms propensity to restate
previously led nancial statements is higher for rms that purchase larger amounts of tax
services from the incumbent auditor.
In the specic context of the going-concern qualication, Goldman 2006 suggests that
poorly performing rms potentially have an enhanced need for some types of tax planning. Such
tax planning strategies, for example, may help these rms maximize taxable losses resulting in tax
refundsan important issue for cash-strapped, poorly performing rms. Having the tax manager
and partner involved in corporate tax compliance and planning also provides more general scru-
tiny of the transactions of the rm. For example, a taxpayer with going-concern problems is more
likely to have loss carryforwards in some or many jurisdictions. Growing operating and capital
loss carryforwards require tax planning to prevent loss expiration. Participating in such tax plan-
ning makes the tax partner and manager better informed about the need for a valuation allowance,
which is related to going-concern risk. Thus tax compliance and planning work can potentially
lead to information spillover that improves audit quality.
3,4
Despite these potential positive effects of auditor-provided tax services on audit quality, recent
Public Company Accounting Oversight Board PCAOB initiatives targeted specically at prohib-
iting auditor-provided tax service appear to be driven by concerns that auditor independence will
be compromised by the additional fees from tax services PCAOB 2004. More specically, the
PCAOB indicated concerns that auditor-provided tax service could impair auditors independence,
and concluded that imposing more stringent ethical and independence requirements with respect to
certain tax services is necessary to restore public condence in the ethics and independence of
public company auditors PCAOB 2004. The SEC recently approved the PCAOB rules on auditor
independence, thus limiting the types of tax services independent auditors may provide to their
SEC audit clients SEC 2006.
5
Thus, the renewed legislative efforts to restrict auditor-provided
tax services suggest a need for academic research on this issue, which is the purpose of this study.
Using a sample of 209 rms ling for Chapter 11 bankruptcy, I nd no empirical evidence
that auditor independence is compromised by the provision of tax services. Specically, I nd a
positive and signicant relation between the likelihood of correctly issuing a going-concern opin-
ion in the last audit report prior to the bankruptcy ling and the level of tax services fees. I nd no
signicant correlation between the going-concern opinion and either audit fees or the nontax
component of nonaudit fees. This result holds even after controlling for unexpected fees and for
endogeneity among audit, nonaudit, tax, and nontax NAS fees. The positive relation between
auditors issuance of going-concern opinions and tax fees is inconsistent with concerns about
3
Goldman 2006 a retired tax partner from a Big 4 audit rm observes that such information spillover from tax
compliance and tax planning work is not uncommon. One specic example of information spillover aiding the going-
concern audit opinion involved a client that had made a loan to the CEO, who was also a founder. The loan was secured
by property purchased with the proceeds from the loan. During tax work related to a joint-venture project to develop the
land, it became clear that the land was acquired for far more than it was worth. This information directly led to the
requirement for a reserve related to this loan during the audit, which threatened the rms going-concern status.
Information spillover is less likely with other types of NAS such as information systems design and implementation
work, given the nature of these tasks.
4
Sample rms disclosures in pre-ling 10-Ks provide corroboration for Goldmans 2006 suggestion that distressed
rms may have an enhanced need for some types of tax planning. Many of these rms discuss the need to implement
tax-planning strategies in order to ensure that the benet of tax carryovers and credits are fully realized.
5
The rules identify circumstances in which the provision of tax services impairs an auditors independence, including
services related to marketing, planning, or opining in favor of the tax treatment of, among other things, transactions that
are based on aggressive interpretations of applicable tax laws and regulations. The rules also treat registered public
accounting rms as not independent of their audit clients if they enter into contingent fee arrangements with those
clients or if the rms provide tax services to certain members of management who serve in nancial reporting oversight
roles at an audit client, or to immediate family members of such persons.
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 33
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
diminished auditor independence, but consistent with suggestions that information spillover from
the provision of tax service enhances audit quality Simunic 1984; Panel on Audit Effectiveness
2000; Kinney et al. 2004.
PRIOR RESEARCH
Regulators concern about the type of nonaudit services auditors provide for their audit clients
and the magnitude of the fees they generate relative to audit services is based on the view that
auditor independence is vital to the production of high-quality audits, and that fees paid to auditors
for both audit and nonaudit services may impair such independence. The SEC-mandated fee
disclosure data has generated numerous academic studies investigating the relation between audi-
tor independence and NAS for example, Frankel et al. 2002; Antle et al. 2002; Ashbaugh et al.
2003; Chung and Kallapur 2003; Francis and Ke 2003; Reynolds et al. 2004; Larcker and
Richardson 2004; Kinney et al. 2004. Overall, these studies do not nd support for diminished
auditor independence from the provision of NAS.
DeFond et al. 2002 and Geiger and Rama 2003 utilize auditors going-concern opinions to
more directly test the effects of NAS on auditor independence. Both these studies conclude that
nonaudit fees have no adverse effect on auditors reporting judgments. In contrast to these studies,
I disaggregate nonaudit fees into a tax and a nontax component, recognizing that tax services may
have potentially different effects on audit quality since the review of the registrants tax returns
and reserves require substantial knowledge about the audit client Sage and Sage 2005, 31.
Further, my denition of audit failure is restricted to nancially distressed rms that failed to
receive a going-concern qualication and subsequently led for bankruptcy.
Prior research specically examining the issue of auditor-provided tax services has focused on
more indirect outcome measures such as rms propensity to restate previously led nancial
statements Kinney et al. 2004 and rms reporting of income tax expense Gleason and Mills
2006. Kinney et al. 2004 document a negative association between tax service fees and restate-
ments, and Gleason and Mills 2006 document that corporations that use their auditors to provide
tax services record lower IRS deciency as additional tax expense. While this evidence suggests
some benecial effect on rms nancial and tax reporting quality from auditor-
provided tax services, there is no evidence, to date, that such information spillover from
auditor-provided tax services specically improves the going-concern opinion. Given the negative
public exposure auditors face when an audit client fails to receive a going-concern and subse-
quently les for bankruptcy, as well as the regulatory initiatives afoot to restrict tax services by
auditors, I examine the relation between auditor-provided tax services and the going-concern
opinion among bankruptcy ling rms.
RESEARCH DESIGN
Sample
I obtain an initial sample of 587 rms ling for Chapter 11 bankruptcy protection from New
Generation Inc.s Bankruptcy Datasource for the period 20012004. Compustat and detailed fee
data requirements result in a nal sample of 209 bankrupt rms153 rms that correctly received
a going-concern opinion prior to ling, and 56 rms that did not.
Going-Concern Model
To investigate whether auditor independence measured by the going-concern audit opinion
is impaired by the presence of auditor-provided tax service I estimate the following logistic
regression model:
34 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
GCOPINION=
0
+
1
RETURN+
2
VOLATILITY +
3
ZFCINDX +
4
LEV +
5
CHLEV
+
6
BIG5 +
7
logASSETS +
8
SWORKCAP +
9
INVEST +
10
SOPCASH
+
11
ISSDEBT +
12
ISSTOCK+
13
DEBTRED+
14
SELLOFF
+
15
RESTRUCT +
16
OENEG+
17
ROA +
18
REPLAG1 +
19
REPLAG2
+
20
DEFAULT +
2123
FEE VARIABLES +
2426
YEAR DUMMIES +
1
where:
GCOPINIONan indicator variable equal to 1 if the rm received a going-concern
opinion prior to the bankruptcy ling, and 0 otherwise;
RETURNthe rms stock return over the last scal year prior to the ling
date;
VOLATILITY the variance in the rms stock returns over the last scal year prior
to the ling date;
ZFCINDX the Zmijewski index measuring bankruptcy risk for the scal year
prior to the ling;
LEV total liabilities divided by total assets at the end of the scal year
prior to the ling;
CHLEV change in LEV over the last two reporting periods prior to the ling;
BIG5 an indicator variable equal to 1 if the rm is audited by a Big 5
audit rm, and 0 otherwise;
log(ASSETS) natural logarithm of total assets at the end of the scal year prior to
the ling;
SWORKCAP working capital current assets current liabilities divided by total
assets at the end of the scal year prior to the ling;
INVEST the sum of cash and long- and short-term investment securities
scaled by total assets at the end of the scal year prior to the ling;
SOPCASHoperating cash ows divided by total assets at the end of the scal
year prior to the ling;
ISSDEBT an indicator variable equal to 1 if the rm issues debt in the scal
year prior to the ling, and 0 otherwise;
ISSTOCKan indicator variable equal to 1 if the rm issues equity in the scal
year prior to the ling, and 0 otherwise;
DEBTREDan indicator variable equal to 1 if the rm reduces long-term debt in
the scal year prior to the ling, and 0 otherwise;
SELLOFF an indicator variable equal to 1 if the rm sells assets in the scal
year prior to the ling, and 0 otherwise;
RESTRUCT an indicator variable equal to 1 if the rm engages in restructuring
activities in the scal year prior to the ling, and 0 otherwise;
OENEGan indicator variable equal to 1 if the rm has negative stockholders
equity at the end of the scal year prior to the ling, and 0
otherwise;
ROA net income divided by total assets at the end of the scal year prior
to the ling;
REPLAG1 number of days between the audit report ling date and the
bankruptcy ling date;
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 35
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
REPLAG2 number of days between the scal year end and the earnings
announcement date;
DEFAULT an indicator variable equal to 1 if the rm is in default on any of its
debt in the year prior to ling, and 0 otherwise;
NONTAXRATIOthe ratio of nontax NAS fees to total fees paid to the auditor;
TAXRATIOthe ratio of tax fees to total fees paid to the auditor;
log(AUDIT FEE) the natural logarithm of the sum of audit fees paid to the auditor;
log(TAX FEE) the natural logarithm of the sum of tax fees paid to the auditor;
log(NONTAX NAS FEE) the natural logarithm of the sum of nontax nonaudit fees paid to the
auditor; and
YEARDUMMIES indicator variable equal to 1 if the audit report is issued in each of
the four sample years, and 0 otherwise.
The going-concern model utilized in the study is primarily based on the DeFond et al. 2002
study, but relies on several other studies to motivate the use of additional explanatory variables
that are correlated with the going-concern opinion e.g., Mutchler et al. 1997; Reynolds and
Francis 2000; Geiger and Rama 2003. Bankrupt rms generally have several years of poor
nancial performance prior to the bankruptcy ling; therefore I include several nancial and
market derived variables in an attempt to capture the nancial condition prevailing at the time of
the audit report that would create substantial doubt about the rms ability to continue as a
going-concern. The model includes RETURN and VOLATILITY, two market-based measures that
DeFond et al. 2002 nd signicantly correlated with the going-concern audit opinion. I expect
a negative positive correlation between the going-concern opinion and RETURN and
VOLATILITY. To capture the probability of default, I use ZFCINDX, which is the nancial
condition index from Zmijewskis 1984 model. Higher values of ZFCINDX indicate a greater
likelihood of default, suggesting a positive relation between ZFCINDX and the issuance of a
going-concern opinion. I include LEV and CHLEV because rms with high and increasing lever-
age are more likely to be closer to debt covenant violations Beneish and Press 1993; Reynolds
and Francis 2000; DeFond et al. 2002. Other measures attempt to capture cash ows, working
capital constraints, and earnings performance. SWORKCAP is working capital divided by total
assets and is expected to be negatively correlated with the going-concern opinion. DeFond et al.
2002 nd a negative correlation between SOPCASH cash ow from operating activities and the
going-concern opinion. I include two measures for earnings performance: ROA current earnings
performance and OENEG a measure of long-term earnings performance. I expect a negative
correlation between ROA and the likelihood of issuing a going-concern opinion, and positive
correlation with OENEG. Like Ohlson 1980, I include OENEG, an indicator variable equal to 1
if the rm has negative owners equity, to capture the persistence of the losses, as a rm is more
likely to report negative owners equity after several years of losses.
Consistent with Statement of Accounting Standards SAS No. 59, when auditors have sub-
stantial doubt about the ability of an entity to continue as going-concern, they are required to
evaluate the plans that management proposes to counteract the adverse circumstances. Behn et al.
2001 document that auditors view various plans differently: plans to issue debt or equity are
negatively correlated with a going-concern opinion, while plans to reduce spending or dispose of
assets are positively correlated with a going-concern opinion. Geiger and Rama 2003 nd a
positive correlation with cost reduction plans and asset sales and issuing a going-concern opinion.
I include ve variables to capture the managerial actions that are potentially correlated with the
going-concern opinion. ISSDEBT and ISSTOCK are indicator variables equal to 1 if the rm issues
debt or equity respectively in the scal year prior to ling; DEBTRED is an indicator
variable equal to 1 if the rm reduces long-term debt and 0 otherwise; SELLOFF is an
36 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
indicator variable equal to 1 if the rm sells assets and 0 otherwise; RESTRUCT is an indicator
variable equal to 1 if the rm incurs any restructuring costs and 0 otherwise; and DEFAULT is an
indicator variable equal to 1 if the rm is in default on any of its debt in the year prior to ling and
0 otherwise.
Other variables in Equation 1 include log(ASSETS), BIG5, REPLAG1, and REPLAG2. Prior
research documents an inverse relation between the likelihood of issuing a going-concern opinion
and rm size Mutchler 1985; Hopwood et al. 1989; Chen and Church 1992; Carcello et al. 1995;
Geiger and Raghunandan 2001; Geiger and Rama 2003. Log(ASSET), the natural logarithm of
total assets at the end of the scal year prior to ling, is the proxy measure for rm size. INVEST
is the sum of cash and long- and short-term investment securities scaled by total assets. DeFond et
al. 2002 include this as a measure of liquidity because rms with large cash and investment
securities are more likely to weather adverse circumstances and potentially forestall bankruptcy.
BIG5 is also included as a control variable because DeFond et al. 2002 nd a positive correlation
between being audited by a Big 5 auditor and receiving a going-concern opinion, consistent with
potentially higher litigation costs inuencing auditor reporting decisions. Additionally, Geiger and
Rama 2006 investigate whether the type of auditor i.e., Big 4, national, or regional/local is
related to the likelihood that the auditor will erroneously issue an unqualied opinion prior to the
bankruptcy ling of the client. They nd that Big 4 rms are more likely to correctly issue a
going-concern opinion, consistent with providing a higher quality audit.
I include REPLAG1, the natural logarithm of the number of days from the audit report date to
the bankruptcy date, because Geiger et al. 2005 nd a signicant correlation between this
variable and the issuance of a going-concern opinion. REPLAG2, the natural logarithm of the
number of days between the scal year-end and the earnings announcement date, is included
because prior research such as Carcello et al. 1995 and DeFond et al. 2002 nd a signicant
association between audit opinion and longer reporting delays. Finally, I include year dummies for
the years covered over the sample period to allow for changes in auditor reporting behavior over
the period of the study because Geiger et al. 2005 document an increasing propensity by auditors
to issue going-concern opinions after January 2002.
RESULTS
Descriptive Statistics
Panel A, Table 1 reports industry distribution for the sample. The data shows a wide industry
distribution, although there is some evidence of industry clustering with three industries
electronic, communications, and business servicestogether accounting for one-third of the
sample. Panel B documents that the probability of incorrectly issuing a clean audit opinion prior
to a bankruptcy ling is decreasing over the sample period, from a high of 48 percent in 2001 to
a low of 15 percent in 2004, suggesting an increasing propensity by auditors to issue going-
concern opinions, similar to Geiger et al. 2005.
6
Panel C reports the distribution across audit
years as well as the frequency of tax fee disclosures among sample rms. For the majority of the
sample rms the audit year predates the SEC requirement to disclose tax fees, which is reected
in the infrequency with which tax fees are reported in the earlier period of the sample. As
6
The increasing propensity of auditors to issue GC opinions may reect increased litigation risk. However, prior research
suggests that litigation risk from auditor-provided NAS is relatively low Antle et al. 1997; Palmrose 2000; Anakwe
2003. Anakwe 2003 analyzes legal lings against CPA rms for nonaudit work provided by CPA rms, and provides
evidence specic to tax services. She reports that only 27 of the 143 cases related to lings against Big 5 auditors, and
eight 30 percent of those cases involved tax work tax compliance, planning, and representation services. Those eight
cases resulted in a total judgment of $4.5 million against Big 5 auditors.
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 37
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
TABLE 1
Industry, Filing Year, and Exchange Distribution of a Sample of
209 Bankruptcy Filing Firms
153 Firms with Going-Concern Qualications and 56 Firms
without Going-Concern Qualications Prior to Filing
Panel A: Sample Distribution by Industry
SIC Code Industry
GC
Sample
No GC
Sample Total
% of
Total
13 Oil and gas exploration 2 2 4 1.9
16 Heavy construction 0 1 1 0.5
17 ConstructionSpecial 1 1 2 1.0
22 Textile mill products 0 1 1 0.5
23 Apparel 2 0 2 1.0
24 Lumber and wood products 1 0 1 0.5
25 Furniture and xtures 2 0 2 1.0
26 Paper and allied products 1 0 1 0.5
28 Chemicals 6 3 9 4.3
29 Petroleum and coal products 1 0 1 0.5
30 Rubber and plastics products 2 0 2 1.0
32 Stone, clay, etc. products 2 0 2 1.0
33 Primary metals 10 2 12 5.7
34 Fabricated metal products 2 0 2 1.0
35 Machinery and computer 7 4 11 5.3
36 Electronic and electrical 16 7 23 11.0
37 Transportation equipment 7 0 7 3.3
38 Measuring equipment 4 0 4 1.9
42 Motor freight transport 2 1 3 1.4
44 Water transportation 1 1 2 1.0
45 Transportation by air 3 0 3 1.4
48 Communications 15 10 25 12.0
49 Utility services 7 2 9 4.3
50 WholesaleDurable 5 3 8 3.8
51 WholesaleNondurable 1 0 1 0.5
52 Building materials 1 1 2 1.0
54 Food stores 1 0 1 0.5
55 Automotive dealers 1 0 1 0.5
56 Apparel and accessory 3 1 4 1.9
57 Home furniture and equip 0 1 1 0.5
58 Eating and drinking places 4 1 5 2.4
59 Miscellaneous retail 3 1 4 1.9
(continued on next page)
38 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
expected, the frequency of tax fee disclosure increases over the sample periodfrom 19.3 percent
for audit years 20002001 to 54.7 percent over the 20022004 audit years.
7
Table 2 reports descriptive statistics for the sample. The mean rm reports negative stock
7
The percentage of the bankrupt rms disclosing tax fees from 2000 to 2002 24 percent is lower than the 42 percent
reported by Omer et al. 2006.
TABLE 1 (continued)
61 Non-depository credit institution 4 3 7 3.3
63 Insurance carriers 2 0 2 1.0
67 Holding, other investment 1 1 2 1.0
73 Business services 19 9 28 13.4
75 Automobile services 3 0 3 1.4
78 Motion pictures 1 0 1 0.5
79 Recreation services 1 0 1 0.5
80 Health services 4 0 4 1.9
83 Social services 1 0 1 0.5
87 Engineering services 3 0 3 1.4
99 Non-classiable 1 0 1 0.5
Total 153 56 209 100
Panel B: Sample Distribution by Filing Year and Audit Opinion
2001 % 2002 % 2003 % 2004 % Total %
GC Concern 13 52 57 75 55 73 28 85 153 73
No GC Concern 12 48 19 25 20 27 5 15 56 27
Total 25 100 76 100 75 100 33 100 209 100
% of total 12 36 36 16 100
Panel C: Sample Distribution by Audit Year and Tax Fee Disclosure
Auditor Report
Frequency
Tax
Fees
Disclosed % Disclosing
GC NGC Total Yes No Total
Audit Year 2000 46 20 66 4 62 6.1
2001 55 24 79 24 55 30.4
2002 34 10 44 18 26 40.9
2003 17 2 19 16 3 84.2
2004 1 0 1 1 0 100
Total 153 56 209 63 146 30.1
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 39
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
TABLE 2
Descriptive Statistics for the 209 Bankruptcy Filing Firms withAvailable Auditor Fee and
Compustat Data for the Fiscal Year Preceding the Bankruptcy Filings
Variables Mean Median Std. Dev. Maximum Minimum
NASRATIO 0.43 0.40 0.22 0.95 0.04
NONTAXRATIO 0.37 0.34 0.25 0.95 0
TAXRATIO 0.06 0 0.12 0.66 0
Total Fees
$thousands
1,333 492 2,305 16,300 26.75
Audit Fees
$thousands
529 265 991 11,900 17.5
Nonaudit Fee
$thousands
750 151 1,640 11,960 0
Nontax NAS Fees
$thousands
732 120 1657 11,960 0
Tax Fees
$thousands
72 0 244 1,927 0
ZFCINDX 3.47 1.24 7.40 47.25 2.84
ASSET $million 1,110 181 2,777 19,415 2.14
RETURN 0.36 0.37 0.38 2.06 0.98
VOLATILITY 1.78 0.53 2.88 10.00 0.04
LEV 0.69 0.69 0.35 2.17 0.10
CHLEV 0.22 0.11 0.45 2.27 0.46
ROA 0.43 0.24 0.75 0.28 4.83
OENEG 0.27 0 0.45 1.00 0
SWORKCAP 0.12 0.04 0.69 0.88 4.07
SOPCASH 0.21 0.03 0.63 0.79 5.42
INVEST 0.16 0.06 0.22 1.00 0
ISSDEBT 0.59 1.00 0.49 1.00 0
ISSTOCK 0.65 1.00 0.48 1.00 0
SELLOFF 0.38 0 0.49 1.00 0
DEBTRED 0.87 1.00 0.34 1.00 0
RESTRUCT 0.22 0 0.42 1.00 0
REPLAG1 279 267 154 672 14
REPLAG2 48 45 24 163 15
GCOPINION 0.73 1.00 0.44 1.00 0
BIG5 0.85 1.00 0.36 1.00 0
DEFAULT 0.31 0 0.46 1.00 0
Variable Denitions:
NASRATIO the ratio of total NAS fees to total fees paid to the auditor;
NONTAXRATIO the ratio of nontax NAS fees to total fees paid to the auditor;
TAXRATIO the ratio of tax service fees to total fees paid to the auditor;
Total Fees the sum of audit and nonaudit fees paid to the auditor;
(continued on next page)
40 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
returns, return on assets, negative working capital, and negative operating cash ows.
8
The mean
and median sample rm is highly leveraged and relatively small compared to the Compustat
population, with mean median assets totaling $1.1 billion $181 million.
9
The average sample
rm was audited by a Big 5 auditor. Many rms took managerial actions prior to the bankruptcy
ling such as issuing new equity and/or debt, selling assets, restructuring activities, and reducing
long-term debt. Table 2 also reports that the mean and median values of nonaudit services to total
fees NASRATIO are 0.43 and 0.40, respectively. The total fees paid to rm auditors ranged from
$27,000 to $16.3 million. Tax advisory service fees, on average, account for 6 percent of total fees
8
I examine whether the sample rms exhibited signs of nancial distress as dened in Hopwood et al. 1994; that is,
whether the rms had at least one of the following: 1 negative working capital in the current year, 2 a loss from
operations in any of the three years prior to bankruptcy, 3 a retained earnings decit in year minus three, or 4 a
bottom-line loss in any of the last three years before bankruptcy. Only four of the sample rms failed to meet the
Hopwood et al. 1994 denition of nancial stress. The results are unchanged when these rms are deleted.
9
The mean median rm in the Compustat population has total assets of $5.5 billion $183.8 million over the sample
period.
TABLE 2 (continued)
Audit Fees the fees paid to the auditor for audit services;
Nonaudit Fees the fees paid to the auditor for all NAS;
Nontax Fees the fees paid to the auditor for all NAS other than tax service;
Tax Fees the fees paid to the auditor for tax advisory services;
ZFCINDX the Zmijewski index measuring bankruptcy risk for the scal year prior to ling;
ASSET total assets at the end of the scal year prior to ling;
RETURN the rms stock returns over the last scal year prior to the ling;
VOLATILITY the variance of stock returns over the year prior to the ling;
LEV total liabilities divided by total assets at the end of the scal year prior to the ling;
CHLEV change in LEV over the last two reporting periods preceding ling;
ROA net income divided by total assets at the end of the scal year prior to the ling;
OENEG an indicator variable equal to 1 if the rm has negative stockholders equity at the end of the scal
year prior to the ling, and 0 otherwise;
SWORKCAP working capital current assetscurrent liabilities divided by total assets at the end of the scal year
prior to the ling;
SOPCASH operating cash ows divided by total assets at the end of the scal year prior to the ling;
INVEST the sum of cash and long- and short-term investment securities scaled by total assets at the end of
the scal year prior to the ling;
ISSDEBT an indicator variable equal to 1 if the rm issues debt in the scal year prior to the ling, and 0
otherwise;
ISSTOCK an indicator variable equal to 1 if the rm issues equity in the scal year prior to the ling, and 0
otherwise;
SELLOFF an indicator variable equal to 1 if the rm sells assets in the scal year prior to the ling, and 0
otherwise;
DEBTRED an indicator variable equal to 1 if the rm reduces long-term debt in the scal year prior to the
ling, and 0 otherwise;
RESTRUCT an indicator variable equal to 1 if the rm has restructuring activities in the scal year prior to the
ling, and 0 otherwise;
REPLAG1 number of days between the audit ling date and the bankruptcy ling date;
REPLAG2 number of days between the scal year end and the earnings announcement date;
GCOPINION an indicator variable equal to 1 if the rm received a going-concern opinion prior to the bankruptcy
ling, and 0 otherwise;
BIG5 an indicator variable equal to 1 if the rm is audited by a Big 5 audit rm, and 0 otherwise; and
DEFAULT an indicator variable equal to 1 if the rm is in default on any of its debt in the year prior to ling,
and 0 otherwise.
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 41
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
while nontax NAS fees account for an average of 37 percent of total fees paid to the auditor.
10
These nontax service fees include: information technology systems fees up to 2002, audit-related
fees, and a category simply dened as other fees. There is signicant variation in the amount of
nonaudit fees paid to the incumbent auditor, ranging from a minimum of $0 to a maximum of
almost $12 million.
Table 3 compares sample variables between the bankrupt ling rms that correctly received a
going-concern opinion prior to the ling GC and those that did not NGC. I focus on differences
in median values as the means are more affected by outliers. Table 3 documents that the GC rms
are more likely to have negative stockholders equity, lower working capital, and are more lever-
aged than the NGC rms. The two sub-samples have similar bankruptcy risk, stock and accounting
performance, and are of similar size. With respect to fee data, the GC sample reports signicantly
higher audit fees and tax service fees than the NGC rms.
Multivariate Test Results
Table 4 reports the results from the estimation of Equation 1. The results reported for Model
1 in the table reect the estimation of Equation 1 without the inclusion of any fee variable and
report a model R
2
of 41.7 percent, implying that the model has fairly good explanatory power.
Similar to DeFond et al. 2002, I document a positive correlation between a going-concern
opinion and volatility and leverage, but no correlation between changes in leverage and operating
cash ows. I also nd that rms with higher working capital and stock performance prior to the
ling are less likely to receive a going-concern opinion. However, going-concern opinions are
more likely if the rm engages in restructuring activities prior to ling. These results suggest that
managerial actions are important control variables.
11
Models 24 reect the estimation of Equation 1 with fee variables included. The pseudo-R
2
is essentially unchanged when the audit fee and grossed-up nonaudit fee variables are added to the
regression see Model 2. I nd no signicant correlation between either the audit fee or the
nonaudit fee variables and the probability of issuing a going-concern opinion. Model 3 reports the
results when nonaudit fee is decomposed into two components: the ratio of total nontax NAS fees
to total fees, and a ratio for tax fees to total fees. Model 4 reports the regression results with log
of tax fees and log of the nontax nonaudit fees as explanatory variables. Both models show
signicant improvement over the base model Model 1. For example, the pseudo-R
2
for Model 3
is 10.6 percent higher than for Model 1 46.1 41.7 and the pseudo-R
2
for Model 4 is 18.2
percent higher than Model 1 49.3 41.7. More importantly, both models report a signicant and
positive correlation between the tax fee variable and the going-concern opinion. The results
suggest that auditors are more likely to correctly issue going-concern opinions prior to bankruptcy
lings if they also provide tax services for their clients. Nontax NAS services fees are not signi-
cantly correlated with the going-concern opinion.
Controlling for Expected Fees and Endogeneity
DeFond et al. 2002, 1264 argue that auditor independence may be inuenced by the
amount of client fees relative to their expected amounts, rather than the nominal amounts. I
10
The mean median tax fees of $72,000 0 is smaller relative to a mean median of $178,000 0 for the Compustat
population over the sample period, suggesting that the average sample rm paid lower tax fees relative to the average
Compustat rm.
11
The ISSDEBT variable has the opposite sign from that predicted from prior research. One explanation for this result is
that instead of being a mitigating factor in the auditors decision to issue a going-concern qualication, the need to seek
new nancing may actually be a conrmatory factor. This may be particularly valid for this sample of rms, given that
the average stressed rm is already likely to be highly leveraged.
42 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
TABLE 3
Comparison of Going-Concern Opinion Sample of 153 Bankruptcy Filing Firms with
Non-Going-Concern Opinion Sample of 56 Bankruptcy Filing Firms
Mean Median
GC
Sample
n = 153
No GC
Sample
n = 56
2-Sample
Test
Two-Tailed
p-value
GC
Sample
n = 153
No GC
Sample
n = 56
Wilcoxon
2-Sample
Test
Two-Tailed
p-value
NASRATIO 0.419 0.454 0.344 0.389 0.415 0.342
0.000 0.000 0.000 0.000
NONTAXRATIO 0.348 0.425 0.058* 0.309 0.409 0.052*
0.000 0.000 0.000 0.000
TAXRATIO 0.072 0.029 0.016** 0.000 0.000 0.002***
0.000 0.043 0.000 0.000
Total Fees 1429.76 1069.54 0.235 515.000 382.921 0.298
0.000 0.000 0.000 0.000
Audit Fees 599.244 337.052 0.014** 283.500 226.500 0.077*
0.000 0.000 0.000 0.000
Nonaudit Fees 830.520 732.492 0.684 180.850 141.806 0.744
0.000 0.000 0.000 0.000
Nontax 740.755 708.566 0.893 128.000 101.372 0.837
NAS Fees 0.000 0.000 0.000 0.000
Tax Fees 89.765 23.926 0.014** 0.000 0.000 0.002***
0.000 0.095 0.000 0.031
ZFCINDEX 3.470 3.477 0.996 1.320 0.618 0.140
0.000 0.008 0.000 0.000
ASSET 1351.1 452.5 0.001*** 193.7 153.6 0.132
0.000 0.000 0.000 0.000
RETURN 0.388 0.262 0.089* 0.395 0.353 0.178
0.000 0.000 0.000 0.000
VOLATILITY 1.962 1.293 0.055* 0.527 0.542 0.411
0.000 0.000 0.000 0.000
LEV 0.740 0.535 0.000*** 0.746 0.558 0.000***
0.000 0.000 0.000 0.000
CHLEV 0.236 0.191 0.519 0.118 0.112 0.436
0.000 0.002 0.000 0.000
OENEG 0.320 0.143 0.004*** 0.000 0.000 0.012**
0.000 0.004 0.000 0.008
ROA 0.399 0.526 0.363 0.244 0.232 0.671
0.000 0.000 0.000 0.000
SWORKCAP 0.208 0.118 0.000*** 0.002 0.098 0.000***
(continued on next page)
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 43
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
therefore estimate separate fee determination regressions models for audit fees, nonaudit fees, tax
fees, nontax NAS fees, and the various fee ratios, relying on the models used in DeFond et al.
2002, but augment the tax fee determination model with three additional variables that poten-
tially inuence the demand for tax servicesthe size of the operating loss carryforwards, and the
TABLE 3 (continued)
Mean Median
GC
Sample
n = 153
No GC
Sample
n = 56
2-Sample
Test
Two-Tailed
p-value
GC
Sample
n = 153
No GC
Sample
n = 56
Wilcoxon
2-Sample
Test
Two-Tailed
p-value
0.000 0.032 0.110 0.000
SOPCASH 0.162 0.345 0.160 0.015 0.076 0.042**
0.000 0.007 0.000 0.000
INVEST 0.140 0.223 0.034** 0.048 0.138 0.028**
0.000 0.000 0.000 0.000
ISSTOCK 0.627 0.696 0.349 1.000 1.000 0.359
0.000 0.000 0.000 0.000
ISSDEBT 0.667 0.393 0.000*** 1.000 0.000 0.000***
0.000 0.000 0.000 0.000
SELLOFF 0.340 0.500 0.041** 0.000 0.500 0.037**
0.000 0.000 0.000 0.000
DEBTRED 0.902 0.786 0.058* 1.000 1.000 0.028**
0.000 0.000 0.000 0.000
RESTRUCT 0.261 0.089 0.001*** 0.000 0.000 0.008***
0.000 0.024 0.000 0.063
REPLAG1 293.51 238.04 0.008** 278.00 243.50 0.040**
0.000 0.000 0.000 0.000
REPLAG2 50.29 43.52 0.022** 45.000 45.00 0.133
0.000 0.000 0.000 0.000
BIG5 0.850 0.857 0.893 1.000 1.000 0.895
0.000 0.000 0.000 0.000
DEFAULT 0.346 0.214 0.053* 0.000 0.000 0.033**
0.000 0.000 0.000 0.000
Variables are as dened in Table 2. *, **, and *** indicate that the difference between the going-concern and the
no-going concern samples is signicant at the 10%, 5%, and 1% levels in two-tail tests. The tests for mean differences are
based on the t-test assuming unequal variances and the tests for median differences are based on the Wilcoxon two-sample
test.
44 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
TABLE 4
The Relation between Going-ConcernAudit Opinions for a Sample of 209 Bankruptcy Filing
Firms andAuditor Fee Variables
Explanatory Variables
Predicted
Sign
Model 1
Coefcient
(p-value)
Model 2
Coefcient
(p-value)
Model 3
Coefcient
(p-value)
Model 4
Coefcient
(p-value)
RETURN / 0.985 0.992 1.117 1.013
0.06 0.07 0.05 0.08
VOLATILITY 0.166 0.170 0.213 0.235
0.09 0.06 0.03 0.02
ZFCINDX 0.047 0.038 0.048 0.039
0.58 0.67 0.62 0.70
LEV 1.668 1.602 1.427 1.572
0.09 0.11 0.15 0.12
CHLEV 0.929 1.098 1.388 1.869
0.49 0.43 0.37 0.26
BIG5 0.184 0.154 0.270 0.343
0.77 0.81 0.67 0.60
Log(ASSET) 0.000 0.057 0.045 0.041
0.99 0.78 0.77 0.84
SWORKCAP 1.384 1.425 1.872 2.110
0.02 0.02 0.00 0.00
SOPCASH 0.227 0.299 0.579 0.560
0.73 0.66 0.40 0.44
INVEST 0.695 0.653 1.175 0.889
0.51 0.54 0.30 0.45
ISSDEBT 0.942 0.891 0.890 0.889
0.04 0.06 0.06 0.07
ISSTOCK 0.227 0.293 0.245 0.188
0.62 0.54 0.61 0.71
DEBTRED 0.247 0.178 0.262 0.317
0.67 0.77 0.67 0.60
SELLOFF 0.950 0.997 1.190 1.304
0.98 0.98 0.99 0.99
RESTRUCT 1.620 1.518 1.929 1.795
0.01 0.02 0.00 0.00
OENEG 0.734 0.720 0.924 0.942
0.34 0.35 0.24 0.25
ROA 0.126 0.158 0.114 0.151
0.71 0.65 0.76 0.69
REPLAG1 0.790 0.837 0.901 0.982
0.98 0.98 0.99 0.99
(continued on next page)
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 45
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
level and change in the valuation allowance account.
12
Using the residuals from the above model
estimations to proxy for the unexpected fee variables, I re-estimate Equation 1 inserting the
unexpected fee variables for the nominal fee variables. The results in Table 5 with the unexpected
fee variables as explanatory variables are almost identical to those reported in Table 4.
The empirical ndings documented so far may be potentially misspecied because the going-
concern opinion, audit fees, and nonaudit fees are endogenously determined. Prior research
Whisenant et al. 2003; DeFond et al. 2002; Geiger and Rama 2003 document that nonaudit and
12
DeFond et al. 2002 rely heavily on other research in developing these models. For example: Craswell et al. 1995;
Whisenant et al. 2003; Firth 1997; Parkash and Venable 1993; Frankel et al. 2002. The inclusion of these additional
variables increased the explanatory power of the model by 11 percent.
TABLE 4 (continued)
Explanatory Variables
Predicted
Sign
Model 1
Coefcient
(p-value)
Model 2
Coefcient
(p-value)
Model 3
Coefcient
(p-value)
Model 4
Coefcient
(p-value)
REPLAG2 0.731 0.660 0.611 0.627
0.17 0.22 0.27 0.30
DEFAULT 0.761 0.694 0.887 0.944
0.16 0.21 0.12 0.12
NONTAXRATIO / 1.129
0.33
TAXRATIO / 5.171
0.05
Log(Audit Fees) / 0.377 0.277
0.34 0.85
Log(Nonaudit Fees) / 0.123
0.59
Log(Tax Fees) / 0.178
0.00
Log(Nontax Fees) / 0.104
0.41
Max-Rescaled R
2
41.7% 42.2% 46.1% 49.3%
Percent 84.7 84.9 86.4 87.6
Concordant
Variable Denitions:
NONTAXRATIO the ratio of nontax NAS fees to total fees paid to the auditor;
TAXRATIO the ratio of tax service fees to total fees paid to the auditor;
Log(Audit Fees) the natural logarithm of the sum of audit fees paid to the auditor;
Log(Nonaudit Fees) the natural logarithm of the sum of nonaudit fees paid to the auditor;
Log(Tax Fees) the natural logarithm of the sum of tax fees paid to the auditor; and
Log(Nontax Fees) the natural logarithm of the sum of NAS service fees minus tax fees paid to the auditor.
All other variables are as dened in Table 2.
46 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
TABLE 5
Going-Concern Opinion Models for a Sample of 209 Bankruptcy Filing Firms with
Unexpected Fee Variables Included as Independent Variables
Explanatory Variables
Predicted
Sign
Model 1
Coefcient
(p-value)
Model 2
Coefcient
(p-value)
Model 3
Coefcient
(p-value)
Model 4
Coefcient
(p-value)
RETURN / 0.985 0.988 1.228 1.182
0.06 0.06 0.03 0.04
VOLATILITY 0.166 0.170 0.200 0.230
0.09 0.06 0.04 0.02
ZFCINDX 0.047 0.042 0.061 0.035
0.58 0.63 0.52 0.70
LEV 1.668 1.677 1.676 1.741
0.09 0.09 0.10 0.08
CHLEV 0.929 0.994 1.092 1.656
0.49 0.47 0.47 0.28
BIG5 0.184 0.156 0.095 0.044
0.77 0.81 0.88 0.95
Log (ASSET) 0.000 0.006 0.036 0.047
0.99 0.97 0.80 0.75
SWORKCAP 1.384 1.407 1.806 1.935
0.02 0.02 0.00 0.00
SOPCASH 0.227 0.265 0.081 0.138
0.73 0.69 0.91 0.84
INVEST 0.695 0.647 1.177 0.936
0.51 0.54 0.29 0.41
ISSDEBT 0.942 0.912 0.854 0.828
0.04 0.05 0.08 0.09
ISSTOCK 0.227 0.253 0.247 0.191
0.62 0.59 0.60 0.69
DEBTRED 0.247 0.209 0.415 0.418
0.67 0.73 0.49 0.49
SELLOFF ? 0.950 0.997 1.203 1.291
0.98 0.98 0.99 0.99
RESTRUCT 1.620 1.547 1.754 1.734
0.01 0.02 0.00 0.01
OENEG 0.734 0.713 1.014 0.959
0.34 0.35 0.21 0.26
ROA 0.126 0.120 0.024 0.009
0.71 0.73 0.95 0.98
REPLAG1 0.790 0.815 0.964 1.043
0.98 0.98 0.99 0.99
(continued on next page)
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 47
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
audit fees are jointly determined. To address this issue, I utilize a two-stage procedure recom-
mended by Nelson and Olsen 1978 and used by DeFond et al. 2002. Table 6 reports the results
from the second-stage structural regression of the going-concern opinion. In the model with
grossed-up nonaudit fees, I document a marginally signicant and positive correlation with audit
fee and the going-concern opinion and no signicant correlation with nonaudit fees. In the model
with nonaudit fees decomposed into the tax and nontax components, I again document a positive
and signicant coefcient on the tax fee variable, a positive and marginally signicant coefcient
on the audit fee variable, and an insignicant coefcient on the nontax nonaudit fee variable.
TABLE 5 (continued)
Explanatory Variables
Predicted
Sign
Model 1
Coefcient
(p-value)
Model 2
Coefcient
(p-value)
Model 3
Coefcient
(p-value)
Model 4
Coefcient
(p-value)
REPLAG2 0.731 0.724 0.627 0.710
0.17 0.18 0.27 0.23
DEFAULT 0.761 0.720 1.074 1.111
0.16 0.19 0.07 0.07
UE / 0.372
0.75
NONTAXRATIO / 7.875
UE TAXRATIO 0.01
UE Log(Audit Fees) / 0.265 0.223
0.50 0.58
UE Log(Nonaudit Fees) / 0.039
0.86
UE Log(Tax Fees) / 0.181
0.00
UE Log(Nontax Fees) / 0.158
0.28
Max-Rescaled R
2
41.7% 41.9% 47.0% 49.5%
Percent 84.7 84.8 87.0 87.7
Concordant
Variable Denitions:
UE NONTAXRATIO residual from the applicable fee determination model;
UE TAXRATIO residual from the applicable fee determination model;
UE Log(Audit Fees) residual from the applicable fee determination model;
UE Log(Nonaudit fees) residual from the applicable fee determination model;
UE Log(Tax Fees) residual from the applicable fee determination model;
UE Log(Nontax fees) residual from the applicable fee determination model.
All other variables are as dened in Table 2.
48 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
TABLE 6
Structural Model fromTwo-Stage Procedure (after controlling for endogeneity) for a Sample
of 209 Bankruptcy Filing Firms from 2001-2004
Dependent Variable = GCOPINION Dependent Variable = GCOPINION
RETURN 1.007 0.09 RETURN 0.814 0.28
VOLATILITY 0.151 0.11 VOLATILITY 0.254 0.03
ZFCINDX 0.061 0.54 ZFCINDX 0.074 0.53
LEV 1.531 0.13 LEV 0.514 0.67
CHLEV 1.018 0.49 CHLEV 3.914 0.05
BIG5 0.154 0.82 BIG5 2.360 0.98
Log (ASSET) 0.220 0.52 Log (ASSET) 1.143 0.05
SWORKCAP 1.431 0.02 SWORKCAP 3.620 0.00
SOPCASH 0.246 0.71 SOPCASH 3.111 0.00
INVEST 0.965 0.37 INVEST 1.859 0.27
ISSDEBT 1.005 0.04 ISSDEBT 0.863 0.16
ISSTOCK 0.200 0.53 ISSTOCK 0.458 0.47
DEBTRED 0.383 0.56 DEBTRED 0.299 0.71
SELLOFF 0.891 0.04 SELLOFF 1.666 0.00
RESTRUCT 1.525 0.02 RESTRUCT 2.485 0.00
OENEG 0.862 0.21 OENEG 1.188 0.22
ROA 0.285 0.33 ROA 0.656 0.28
REPLAG1 0.709 0.95 REPLAG1 0.982 0.97
REPLAG2 0.495 0.43 REPLAG2 0.457 0.57
DEFAULT 0.770 0.21 DEFAULT 0.081 0.92
AUDHAT 1.403 0.08 AUDHAT 2.342 0.07
NONHAT 0.552 0.25 TAXHAT 1.169 0.00
NONTAXHAT 0.196 0.42
Chi-squared for model 25 degrees of
freedom 73.810 p = 0.0001
Max-rescaled R
2
0.433
Chi-squared for model 26 degrees of
freedom 127.061 p = 0.0001
Max-rescaled R
2
0.663
Percent Concordant 85.6 Percent Concordant 93.3
Variable Denitions:
AUDHAT predicted value of audit fee;
NONHAT predicted value of nonaudit fee;
TAXHAT predicted value of tax service fee component of nonaudit fee;
NONTAXHAT predicted value of nontax service fee component of nonaudit fee.
All other variables are as described in Table 2. The models include year dummy variables estimates not reported.
p-values are reported in parentheses.
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 49
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
These results suggest that the earlier nding from Tables 4 and 5, which document that auditor
independence is enhanced by tax work, is not affected by endogeneity. In fact, controlling for the
joint determination in audit and nonaudit fees improves the explanatory power of the model.
13
Robustness Checks
Conservatism versus Accuracy
The positive relation between the tax fee variable and the going-concern opinion is consistent
with information spillover. However, an alternative explanation for the positive relation is that the
joint provision of audit and tax services induces auditors to make more conservative decisions.
This implies that auditors would be more likely to give qualied reports even when they are
unwarranted. To rule out this alternative explanation, I re-estimate Equation 1 using a matched
sample of nonbankrupt rms from the Compustat population.
14
I nd no evidence that conserva-
tism is driving the result reported in the study. Specically, the coefcient on the tax fee variable
while positive is insignicant 0.062, p = 0.216 in a two-tailed test. While the positive coefcient
is consistent with some conservatism effects, a comparison of the coefcients on the tax variable
between the two samples suggests that this is not likely what is driving the results in the bank-
ruptcy setting. The coefcient is much smaller than that in the bankrupt sample 0.062 compared
to 0.178.
First-Time Going-Concern Opinions
Prior studies suggest that issuing a rst-time going-concern opinion may be more difcult for
an auditor relative to issuing a modied opinion in subsequent periods, and as such, may require
a different decision model Kida 1980; Mutchler 1984; Geiger et al. 1998; Carcello and Neal
2003; Geiger and Rama 2003. Accordingly, I re-estimate Equation 1 but restrict the going-
concern rms to those rms with rst-time going-concern opinions. The restricted sample of 181
rms includes 125 rst-time going-concern opinions and the 56 rms that had unqualied opin-
ions. The results are robust to this alternative specication. The regression pseudo-R
2
increased to
0.567, which suggests that the model performance improved with the rst-time going-concern
restriction. The coefcient on the tax fee variable is 0.236 and is highly signicant. The coef-
cients on the other two fee variables are insignicant at conventional levels.
Mandatory Tax Fee Disclosure
The SEC requirement to separately disclose the tax fee components of nonaudit fees only
became mandatory in 2003.
15
While the SEC encouraged issuers to adopt the disclosure provisions
earlier, the disclosure of tax service fees was voluntary prior to the 2002 audit year. Omer et al.
2006 provide evidence that many rms did not disclose tax fees prior to the SEC requirement
13
I also test whether the going-concern opinion, audit fee, and nonaudit fee are endogenous. I nd no evidence based on
the Hausman 1978 test that the going-concern opinion is jointly determined with audit and nonaudit fees for this
sample of bankrupt rms.
14
To ensure that the matched rms include some going-concern opinions since the dependent variable is the going-
concern opinion, I match based on size asset, probability of bankruptcy Zmijewski index and rm performance
ROA in the year prior to the bankruptcy ling. The matched rms are not signicantly different from the sample rms
in size, probability of bankruptcy, protability, level of debt, tax fees, and audit fees. However, the matched rms have
higher ratios of tax fees to total fees and audit to total fees. Sixty-nine matched rms received going-concern opinions
in the audit report prior to the ling year.
15
The new disclosure requirements were effective for periodic annual lings and proxy or nancial statement lings for
the rst scal year ending after December 15, 2003. Thus, the new disclosure requirements were not mandatory until the
calendar-year 2003 periodic annual lings were made in 2004. Companies were required to disclose fees paid to the
principal auditor in four categories for the two most recent years.
50 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
only 42 percent of their sample of 5,727 rm years from 2000 to 2002 disclosed tax fees.
16
The
lack of disclosure potentially introduces noise in the tax fee variable, since a zero tax fee reported
prior to the SEC requirement may be actually nonzero but not disclosed. This would seem to bias
against nding any signicance on the tax fee variable. To mitigate this potential problem,
I re-estimate Equation 1 but restrict the sample to audit years after the SEC requirement
20022004 audit years.
17
The results are qualitatively unchanged coefcient on the tax fee
variable is 0.227, p = 0.035 based on one-tailed test. I also re-estimate the model using a sample
that includes rms that voluntarily reported tax fees prior to the SEC requirement 20002001
audit years in addition to the post-SEC sample. The coefcient on the tax fee variable is 0.346
and is signicant at the 1 percent level. The coefcients on the other two fee variables are
insignicant at conventional levels.
18
Inuential Observations
To determine whether the results are driven by a few sample rms with large tax service fees,
I delete those rms with tax fees above the mean of $72,000.
19
The results are qualitatively
unchanged when these rms are excluded from the sample. I also perform a more rigorous test for
outliers suggested by Belsley et al. 1980 and deleted eight rms with residuals which suggest
that these are potentially inuential observations. The results are qualitatively unchanged when
these rms are excluded from the sample. In fact, both the audit and tax fee variables are positive
and signicant in the re-estimated models.
Other Fees
Finally, I decompose nonaudit fees into all of its componentsnamely audit-related tax
service fees, information systems design fees, and a category listed as other. I nd no signicant
effect for any of the other NAS fee components, and more importantly report the positive and
signicant result for the tax service fee component only.
SUMMARY
This study examines whether auditor independence is impaired through the provision of tax
services by focusing on auditors issuance of going-concern opinions among a sample of
bankruptcy-ling rms. The evidence from the bankruptcy setting is important to the continuing
debate on auditor independence because the bankruptcies of several large corporations such as
Enron brought the issue to center stage and motivated several provisions of the Sarbanes-Oxley
Act of 2002. SOX severely restricted many types of nonaudit services audit rms could provide to
their clients while allowing some tax services, suggesting some benecial effect from auditor-
provided tax services.
I document a signicant and positive relation between the likelihood of correctly issuing a
going-concern opinion in the last audit report prior to the bankruptcy ling and the level of tax
services fees. I nd no signicant correlation with the going-concern opinion and either audit fees
16
Mishra et al. 2005 nd evidence of higher tax to audit fee ratios after the SEC requirement, consistent with higher tax
fees being reported.
17
Restricting the sample to audit years after mandatory disclosure 64 observations and estimating the regression with the
full set of explanatory variables led to failure of the logistic regression to converge. The results I report above are based
on the 64 rms, but the model only includes those explanatory variables that are signicant at least at the 10 percent
level, in addition to the fee variables.
18
Including the 28 pre-SEC voluntary disclosers increases the sample size to 92 observations. I also examine whether the
results are sensitive to any potential confusion about how to disaggregate total fees into the audit and other components
in the year after the SEC requirement. I re-estimate the regression after deleting observations for the 2000 audit year the
rst year that rms were required to publicly report disaggregated fees. The results are qualitatively unchanged.
19
Thirty-two rms fall in this category, 28 of which received a going-concern opinion.
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 51
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
or the nontax component of NAS fees. This result holds after controlling for unexpected fees,
endogeneity among the fee variables, and various robustness checks. This result is inconsistent
with diminished auditor independence, but rather suggests that audit quality is enhanced through
information spillover from auditor-provided tax services. The results of the study contribute to the
current debate on auditor independence and NAS by providing evidence that the level of tax
service fees does not appear to systematically impair auditor independence in failing rms.
REFERENCES
Anakwe, B. 2003. An examination of the factors associated with lawsuits against accountants for nonaudit
services. Unpublished dissertation, Rutgers, The State University of New Jersey.
Antle, R., E. Gordon, G. Narayanamoorthy, and L. Zhou. 2002. The joint determination of audit fees,
nonaudit fees, and abnormal accruals. Working paper, Yale School of Management.
, P. Grifn, D. Teece, and O. Williamson. 1997. An Economic Analysis of Auditor Independence for a
Multi-Client, Multi-Service Public Accounting Firm. Report prepared on behalf of the AICPA in
connection with the presentation to the Independence Standards Board of Serving the Public Interest:
A New Conceptual Framework for Auditor Independence. Berkeley, CA: The Law & Economics
Consulting Group Inc.
Ashbaugh, H., R. Lafond, and B. Mayhew. 2003. Do nonaudit services compromise auditor independence?
Further evidence. The Accounting Review 78: 611639.
Behn, B., S. Kaplan, and K. Krumweide. 2001. Further evidence on the auditors going-concern report: The
inuence of management plans. Auditing: A Journal of Practice & Theory 20: 1328.
Belsley, D., E. Kuh, and R. Welsch. 1980. Regression Diagnostics. New York, NY: John Wiley & Sons, Inc.
Beneish, D., and E. Press. 1993. Costs of technical default. The Accounting Review 68: 233257.
Carcello, J., D. Hermanson, and F. Huss. 1995. Temporal changes in bankruptcy-related reporting. Auditing:
A Journal of Practice & Theory 74: 133143.
, and T. Neal. 2003. Audit committee composition and auditor dismissals following new going-
concern reports. The Accounting Review 78: 95117.
Chen, K., and B. Church. 1992. Default on debt obligations and the issuance of going-concern opinions.
Auditing: A Journal of Practice & Theory Fall: 3049.
Chung, H., and S. Kallapur. 2003. Client importance, nonaudit fees, and abnormal accruals. The Accounting
Review 78: 931955.
Craswell, A., J. Francis, and S. Taylor. 1995. Auditor brand name reputations and industry specializations.
Journal of Accounting and Economics 20: 297312.
DeFond, M., K. Raghunandan, and K. Subramanyam. 2002. Do nonaudit fees impair auditor independence?
Evidence from going-concern auditor opinions. Journal of Accounting Research 40: 12471274.
Dietz, D. 2002. Auditors are timidThey failed to warn in most big rm bankruptcies since 1996. Pittsburg
Post-Gazette April 26: C8.
Firth, M. 1997. The provision of nonaudit services by accounting rms to their audit clients. Contemporary
Accounting Research 14: 111.
Francis, J., and B. Ke. 2003. Do fees paid to auditor increase a companys likelihood of meeting analysts
earnings forecasts? Working paper, University of Missouri.
. 2004. What do we know about audit quality? The British Accounting Review 36: 345368.
Frankel, R., M. Johnson, and K. Nelson. 2002. The relation between auditors fees for nonaudit services and
earnings management. The Accounting Review 77: 71105.
Geiger, M., K. Raghunandan, and D. Rama. 1998. Costs associated with going-concern modied audit
opinions: An analysis of auditor changes, subsequent opinions and client failures. Advances in Ac-
counting 16: 117139.
, and . 2001. Bankruptcies, audit reports, and the Reform Act. Auditing: A Journal of Practice &
Theory 20: 187195.
, and D. Rama. 2003. Audit fees, nonaudit fees, and auditor reporting on stressed companies. Auditing:
A Journal of Practice & Theory 22: 5369.
52 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
, K. Raghunandan, and D. Rama. 2005. Recent changes in the association between bankruptcies and
prior audit opinions. Auditing: A Journal of Practice & Theory 24: 2135.
, and D. V. Rama. 2006. Audit rm size and going-concern reporting accuracy. Accounting Horizons
20: 117.
Gleason, C., and L. Mills. 2006. Do auditor-provided tax services compromise auditor independence with
respect to tax expense? Working paper, University of Arizona and University of Iowa.
Goldman, D. 2006. Interview comments on information spillover through auditor-provided tax services.
October 11.
Hausman, J. 1978. Specication tests in econometrics. Econometrica 46: 12511271.
Hilzenrath, D. 2001. The numbers crunch: After Enron, new doubts about auditors. Washington Post De-
cember 5: A1.
Hopwood, W., J. McKeown, and J. Mutchler. 1989. A test of the incremental explanatory power of opinions
qualied for consistency and uncertainty. The Accounting Review January: 2848.
, , and . 1994. A reexamination of auditor versus model accuracy within the context of the
going-concern opinion modication. Contemporary Accounting Research Spring: 409431.
Kida, T. 1980. An investigation of auditors continuity and related qualication judgments. Journal of Ac-
counting Research Autumn: 506523.
Kinney, W., Z-V. Palmrose, and S. Scholz. 2004. Auditor independence, nonaudit services, and restatements:
Was the U.S. Government right? Journal of Accounting Research 42: 561588.
Larcker, D., and S. Richardson. 2004. Fees paid to audit rms, accrual choices and corporate governance.
Journal of Accounting Research 42: 625658.
Lindberg, D., and F. Beck. 2004. Before and after Enron: CPAs views of auditor independence. The CPA
Journal 74: 3639.
Mishra, S., K. Raghunandan, and D. Rama. 2005. Do investors perceptions vary with types of nonaudit fees?
Evidence from auditor ratication voting. Auditing: A Journal of Practice & Theory 24: 925.
Mutchler, J. 1984. Auditors perceptions of the going-concern decision. Auditing: A Journal of Practice &
Theory Spring: 1729.
. 1985. A multivariate analysis of the auditors going-concern opinion decision. Journal of Accounting
Research Autumn: 668682.
, W. Hopwood, and J. McKeown. 1997. The inuence of contrary information and mitigating factors on
audit opinion decisions of bankrupt companies. Journal of Accounting Research 35: 295310.
Nelson, F., and L. Olsen. 1978. Specication and estimation of simultaneous equations model with limited
dependent variables. International Economic Review 19: 695710.
Oates, M. 2002. Auditor independence, Sarbanes-Oxley, and tax services. The Tax Executive Sept-Oct.
Available at: http://www.ndarticles.com/p/articles/mi_m6552/is_5_54/ai_94779549.
Ohlson, J. 1980. Financial ratios and the prediction of bankruptcy. Journal of Accounting Research 18:
109131.
Omer, T., J. Bedard, and D. Falsetta. 2006. Auditor-provided tax services: The effects of a changing regula-
tory environment. The Accounting Review 81: 10951117.
Palmrose, Z-V. 2000. Empirical Research on Auditor Litigation: Considerations and Data. Studies in Ac-
counting Research No. 33. Sarasota, FL: American Accounting Association.
Panel on Audit Effectiveness. 2000. The Panel on Audit Effectiveness Report and Recommendations. Stam-
ford, CT: Public Oversight Board.
Parkash, M., and C. Venable. 1993. Auditee incentives for auditor independence: The case of nonaudit
services. The Accounting Review 68: 113133.
Public Company Accounting Oversight Board PCAOB. 2004. Proposed Ethics and Independence Rules
Concerning Independence, Tax Services and Contingent Fees, PCAOB Rulemaking Docket No, 017.
PCAOB Release No. 2004-015. Available at: http://www.pcaobus.org/Rules_of_the_Board/
Documents/Docket_017/Release2004-015.pdf., Dec. 14.
Reynolds, K., and J. Francis. 2000. Does size matter? The inuence of large clients on ofce-level auditor
reporting decisions. Journal of Accounting and Economics 30: 375400.
, D. Deis, and J. Francis. 2004. Professional service fees and auditor objectivity. Auditing: A Journal of
Practice & Theory 23: 2952.
Auditor Independence and Auditor-Provided Tax Service: Evidence from Going-Concern Audit Opinions 53
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association
Sage, L., and J. Sage. 2005. Tax service implications of the Sarbanes-Oxley Act of 2002. Managerial Finance
31: 2934.
Securities and Exchange Commission SEC. 2000. Revision of the Commissions Auditor Independence
Requirements. Release Nos. 33-7919; 34-43602. Washington, D.C.: SEC.
. 2006. Rule Filings with Respect to the PCAOBs Proposed Ethics and Independence Rules Concerning
Independence, Tax Services, and Contingent Fees. March 28. Washington, D.C: SEC.
Sharma, D., and J. Sidhu. 2001. Professionalism vs. commercialism: The association between nonaudit
services NAS and audit independence. Journal of Business Finance & Accounting 28: 595630.
Simunic, D. 1984. Auditing, consulting and auditor independence. Journal of Accounting Research 22:
679702.
U.S. House of Representatives. 2002. The Sarbanes-Oxley Act of 2002. Public Law 107-204 H. R. 3763.
Washington, DC: Government Printing Ofce.
Venuti, E. 2004. The going-concern assumption revisited: Assessing a companys future viability. The CPA
Journal May: 4044.
Weil, J. 2001. Going concerns: Did accountants fail to ag problems at dot-com casualties? Wall Street
Journal February 9: C1.
. 2003. Travel-Billing probe has a bigger scope. Wall Street Journal September 26.
Whisenant, S., S. Sankaraguruswamy, and K. Raghunandan. 2003. Evidence on the joint determination of
audit and nonaudit fees. Journal of Accounting Research 41: 721744.
Zmijewski, M. 1984. Methodological issues related to the estimation of nancial distress prediction models.
Journal of Accounting Research Supplement: 5982.
54 Robinson
Auditing: A Journal of Practice &Theory November 2008
American Accounting Association

Вам также может понравиться