-
Forced auditor change, industryspecialization and audit fees
Winifred D. ScottCollege of Business, Zayed University, Dubai,
United Arab Emirates, and
Willie E. GistSchool of Accountancy, College of Business, Ohio
University,
Athens, Ohio, USA
Abstract
Purpose The purpose of this study is to explore the effect of
industry specialization on theabsorption and competitive pricing
(or lack thereof) of audits of large Andersen clients (S&P
1500companies) who switched to the remaining Big 4 international
accounting firms in 2002 due to thedemise of Arthur Andersen LLP
(Andersen). Did the audit clients pay a premium or discount in
auditfees to their new auditor who specialized in their
industry?
Design/methodology/approach Ordinary least squares regression is
used to test hypothesis of apositive association between industry
specialization and audit fees charged to former Andersens
auditclients in 2002 following Andersens demise. This study
provides more control over size effects by design.Test variables
are constructed based on national market share of audit fees within
an industry. Logisticregression is used to examine the likelihood
of choosing new auditor that is an industry specialist.
Findings Results support hypothesis, consistent with auditor
differentiation explanation.Proportion of clients that had engaged
an industry specialist in 2001 increased from 38 percent(84
clients) to 48 percent (105 clients) in 2002. No evidence of
price-gouging in 2002 although clientswho aligned with industry
specialist paid a 23.2 percent premium in audit fees. Large clients
lostbargaining power to negotiate lower fees. Findings are robust
to the inclusion of additional alternativemeasures of company
size.
Research limitations/implications Results of logistic regression
analysis imply that large auditclients with former auditor of
tarnished reputation, long auditor tenure and high leverage are
morelikely to switch to an industry specialist to possibly signal
audit/financial reporting quality. Largesample companies may limit
the ability to generalize findings to smaller companies.
Practical implications Mandatory audit firm rotation (currently
being debated in the profession)will have costly effect on the
pricing of Big 4 audits for companies wanting to signal audit
andfinancial reporting quality to affect market perception, and
large companies would likely lose theirability to bargain for lower
audit fees.
Originality/value The paper focus on the alignment of Andersen
clients and impact on audit feeswith Big 4 industry specialists
resulting from the sudden increase in audit market concentration.
Priorto Andersens collapse, evidence on the association of audit
fees premium and industry specialists wasmixed, and little
attention has been given to the influence of auditor industry
specialization on bothaudit fees and alignment of former Andersen
clients with a Big 4 specialist. This paper fills that void.
Keywords Andersen, Industry specialization, Auditor switching,
Involuntary auditor change,Audit fees, Audit market concentration,
Price-gouging, Bargaining power, Mandatory auditor
rotation,Auditors, Auditing
Paper type Research paper
The current issue and full text archive of this journal is
available at
www.emeraldinsight.com/0268-6902.htm
The authors are very thankful for the valuable and constructive
comments of anonymousreferees in significantly improving this
manuscript.
Data availability. Data are publicly available from the sources
identified in the paper.
Managerial Auditing JournalVol. 28 No. 8, 2013pp. 708-734q
Emerald Group Publishing Limited0268-6902DOI
10.1108/MAJ-11-2012-0779
MAJ28,8
708
-
I. IntroductionAudits play an important role in corporate
governance. They provide independentassurance to investors and
other stakeholders that management prepared financialstatements are
not materially misstated in accordance with generally accepted
accountingstandards. Understanding the clients industry enhances
the auditors professionalskepticism about the proper recognition
and valuation of transactions and events relatedto that industry.
Consequently, audit firms differentiate themselves from other
competingaudit firms by specializing in certain industries in order
to provide better quality serviceto their audit clients than a
non-industry specialist audit firm (Habib, 2011). The issue
ofauditor industry specialization is relevant to the auditing
profession as firms organizetheir practices along industry lines to
increase the effectiveness and quality of their audits(American
Institute of Certified Public Accountants (AICPA, 1998); Owhoso et
al., 2002;Bell et al., 1997; GAO, 2003a; Knechel et al., 2007;
Cenker and Nagy, 2008; Cahan et al.,2008). Industry specialization
is often perceived as a proxy for audit quality. In 2001,Andersen
served as an industry specialist auditor to many of its large
clients.
When Andersen was barred from conducting and reporting on audits
for Securitiesand Exchange Commission (SEC) registered companies in
2002, its audit clients had tofind a new auditor in a hurry because
they needed to file their annual audited 10-kstatements with the
SEC. Although the SEC issued temporary rules on filingrequirements
(SEC Release 33-8070, 2002), companies did not want to delay filing
ofaudited financial statements since investors favor the issuance
of timely financialreports. Filing unaudited financial statements
had the potential to hurt investors andreduce the value of the
firm, and was not the preferred course of action. This
forcedauditor change (in contrast to a voluntary auditor change)
for hundreds of companies atone time was very unique in the audit
market. Since the shredded reputation ofAndersen negatively
affected the stock price of its clients (Chaney and Philipich,
2002),it seems reasonable that many of Andersens former audit
clients would have wanted tosignal a positive perception about the
expressed opinion on their financial statementsby selecting an
industry specialist as their new auditor. A number of studies
indicatethat several former Andersen clients selected their new
auditor by simply followingtheir Andersen audit partner to their
new auditor (Blouin et al., 2007; Vermeer et al.,2008; Kohlbeck et
al., 2008). Vermeer et al. (2008) report lower fees paid by
followers,whereas Kohlbeck et al. (2008) find no evidence of a
premium or discount to followerscompared to audit fees paid by
non-follower audit clients. Kealy et al. (2007) find thatclients
who were with Andersen for a long time faced greater professional
skepticismabout the quality of prior audits and paid larger audit
fees than clients with short tenurewith Andersen. Little attention,
however, has been given to the influence of auditorindustry
specialization on both the audit fees and new auditor selection of
the formerAndersen clients. The collapse of Andersen brought about
a sudden increase in theaudit market concentration in the large
client segment and induced a forced auditorchange environment, in
which we examine the auditor specialization and audit feesrelation.
In addition to being able to explore specialization effect in this
environment,such a study may have implications for:
. mandatory auditor rotation;
. client-auditor alignment; and
. price-gouging behavior.
Forced auditorchange
709
-
The purpose of this study is to explore the effects of industry
specialization oncompetitive pricing (or lack thereof) of audits
for large Andersen clients that switched tothe remaining Big 4
international accounting firms (hereafter, the Big 4) in 2002.
Giventhe necessity of the involuntary switching of former Andersen
clients, did the auditclients pay a premium or discount in audit
fees to their new auditor who specialized intheir industry? Whether
a fee premium or discount is associated with auditor
industryspecialization has not been convincingly documented in the
literature; the results aremixed (Mayhew and Wilkins, 2003;
Casterella et al., 2004; Hay et al., 2006; Ghosh andLustgarten,
2006; Kohlbeck et al., 2008; Habib, 2011). This study contributes
to ourunderstanding of the effect that industry specialization and
involuntary auditorswitching had on the audit fees of hundreds of
large audit clients who changed auditorat the same time for the
same reason. We believe that given the tarnished reputation
ofAndersen, the large former Andersen clients wanted to signal the
quality of theirfinancial reporting, and would likely have done so
by engaging a Big 4 industryspecialist. Furthermore, a more likely
competitive response by a Big 4 network firm(discussed below)
absorbing these clients would be to increase audit fees (rather
thandiscount them) to better reflect the value of the audit
(whether due to actual or perceivedhigher quality) and earn an
appropriate return on the additional investment made bythe firm to
differentiate its product as a specialist.
This study is restricted to the former Andersen clients who had
been part of theS&P 1500 in 2001 and who chose one of the
remaining Big 4 audit firms as their newauditor. We limit our study
to the largest of Arthur Andersens former clients toprovide more
control over size effects by design, and to reduce or eliminate
possibleconfounding effects between any premiums resulting from
industry specialization orfrom low bargaining power of the smaller
audit clients. For example, Casterella et al.(2004) find that
premiums for industry specialization arise when clients have
lowbargaining power (in the case of smaller clients). Also, by
focusing our study on largerclients we avoid the size effect issue
reported by Francis et al. (2005) and Craswell et al.(1995) whereby
the premiums for industry leadership in their samples are driven by
theupper half of company size. We find that 48 percent of the 221
former Andersen clientsselected an industry specialist auditor in
2002. While 40 former Andersen audit clientslost the privilege of
having an industry specialist auditor in 2002, 61 former
Andersenclients (who did not have Andersen as their industry
specialist in 2001) gained theadvantage of having an industry
specialist auditor to express an opinion on thereliability and
faithful representation of their 2002 financial statements to
investorsand other stakeholders.
Given the size, resources, and national/international presence
of the S&P 1500, thesecompanies are more likely concerned with
the firm-wide and international reputation oftheir auditors as
oppose to the auditors local-office reputation. A Big 4 network
firmrefers to the organizational structures and operations of
national and internationalaccounting firm networks that may produce
positive synergies which benefit verylarge companies to a great
degree (Carson, 2009). Therefore, we examine nationalindustry
leadership as opposed to city-specific industry leadership. Francis
et al. (2005)examine both national and city-specific industry
expertise and find that they jointlyaffect audit fees.
City-specific industry expertise may be an appropriate
considerationfor the companies in their sample with average total
assets of $1.9 billion, however, it isdifficult for us to argue
that the average company of $11 billion total assets as in our
MAJ28,8
710
-
sample would fixate on local-office rather than
national/international expertise of itsauditors from a Big 4
network firm.
Ordinary least squares (OLS) regression analysis is used to test
the associationbetween the audit fees of Andersens former clients
and auditor industry specialization.As hypothesized we find that
the association between audit fees and auditor
industryspecialization is positive and significant at 0.05 level or
better, consistent with theFrancis et al. (2005) model that
considers only national industry leaders. Our findingsupports the
product differentiation explanation. OLS results indicate that the
formerAndersen clients paid, on average, a fee premium of 23
percent to their new industryspecialist auditor. We also find that
these very large companies did not have thebargaining power to
negotiate lower fees, in contrast to Casterella et al. (2004).
Sinceprior studies (Huang et al., 2007; Kohlbeck et al., 2008)
indicate a potential large-clientsize effect on pricing audit
services, additional tests are conducted to examine whetherthe
results are driven by client size. We find the inferences of our
results unchanged.Tests also did not indicate that price-gouging or
low-balling were pervasive in thepricing of audits of Andersens
former clients. Further, results of logistic regressionanalysis
indicate that the likelihood of choosing a new industry specialist
auditorincreased as the length of the client-auditor tenure with
Andersen increased, consistentwith Kealy et al. (2007).
The remainder of the paper is organized into four sections. The
next sectionprovides some background and the hypothesis
development. Section III discusses themethodology and Section IV
describes the sample selection. Results of the analyses
arepresented in Section V and the conclusion, contribution, and
implications are discussedin Section VI.
II. Background and hypothesis developmentAuditor switching and
the pricing of audit servicesTheoretical models of audit pricing
suggest that when a client voluntarily switchesauditors, the client
should initially enjoy lower audit fees because
non-incumbentauditors low-ball or discount the initial audit
engagement to earn the right to futurequasi-rents of audit fees
(DeAngelo, 1981; Beck et al., 1988). Prior to Andersens
demise,empirical studies reported evidence of persistent initial
price cutting (Simon andFrancis, 1988; Turpin, 1990; Yardley et
al., 1992; Whisenant et al., 2003).
Voluntary auditor switching, in general, focuses on matters such
as pressuringincumbent auditors to issue clean audit opinions,
brand name reputation, industryspecialization, market power, and
low-balling/price-gouging (DeAngelo, 1981; Chowand Rice, 1982;
Palmrose, 1986a, b; Ettredge and Greenberg, 1990; Yardley et al.,
1992;Craswell et al., 1995; Deis and Giroux, 1996; AICPA, 1998;
Owhoso et al., 2002;Balsam et al., 2003; Krishnan, 2003; Knechel et
al., 2007; Kohlbeck et al., 2008). AfterAndersens demise, the
Herfindahl-Hirschman Index for audit firms increased to 2,566,well
above the score of 1,800 that indicates audit firms have the
potential to exercisemarket power (Eisenberg and Macey, 2003)[1].
In this unique setting regulators wereconcerned about excessive
pricing for the hundreds of involuntary auditor
switchingcompanies.
Some studies indicate that several former Andersen audit clients
chose to followtheir Andersen audit partner to the new auditor. For
example, Blouin et al. (2007) findthat slightly more than half of
their sample, 226 out of 407, followed their Andersen
Forced auditorchange
711
-
audit partner to the new auditor. In their follow/non-follow
logistic regression model,they find that audit clients with greater
switching costs as well as industries with themost number of
clients in a single industry (CLIENT variable) were more likely
tofollow Andersens audit team. However, their CLIENT variable may
also be capturinga lack of competition and less of an indicator of
switching costs. While Vermeer et al.(2008) find that half their
sample of 575 former Andersen clients who followed theAndersen
audit partner/team to the new auditor paid lower audit fees,
Kohlbeck et al.(2008) in contrast find that former Andersen clients
who were early switchers andthose who followed the audit team did
not experience fee discounts or premiums.
Other studies indicate that the larger audit clients of Andersen
were more likely tobe early switchers (Kohlbeck et al., 2008;
Barton, 2005; Chen and Zhou, 2007). Forexample, Chen and Zhou
(2007) find that companies with larger audit committees withgreater
financial expertise and companies with larger boards were more
likely todismiss Andersen sooner and choose a Big 4 successor
auditor.
In addition, some studies indicate that the perceived riskiness
of former Andersenaudit clients influence audit pricing. One
measure of client risk of former Andersenclients is client-auditor
tenure. The longer that a company was a client of Andersen,
thegreater the skepticism about the quality of prior audits and
greater the risk that priorfinancial statements of former Andersen
clients were not audited independently.Regulatory limits on
client-auditor tenure have not been set by the SEC or
PCAOB,however, a GAO (2003b) report finds that the average
client-auditor tenure ofFortune-1000 companies is 22 years. In our
usable sample of large Andersen clients(S&P 1500 companies),
the average client-auditor tenure at the time of Andersensdemise is
15.75 years with 37 percent of sample companies having 17 years or
moretenure with Andersen. Kealy et al. (2007) find a positive and
significant associationbetween audit firm tenure and audit fees
paid to the successor auditors by formerAndersen clients. They
interpret the results as supporting the perception that
longclient-auditor tenure is a factor that increases the risk of a
new client. On the otherhand, some view short auditor-tenure as
risky. For example, Landsman et al. (2009) usean multinomial
logistic auditor switch model (involving lateral, upward,
anddownward moves to/from the Big N auditors) to examine whether
company-specificrisk factors and client misalignment are
differentially associated with Big N auditorswitch decisions in the
pre-Enron period (1993-2001) and post-Enron period(2002-2005). In
their study, short tenure is viewed as a risk proxy that increases
thelikelihood of audit failure. Although their evidence is
consistent with the Big 4becoming more sensitive to client risk in
the post-Enron period, their post-Enronsample excludes former
Andersen clients from the analysis. Our study controls for
theeffect that client-auditor tenure has on the pricing of audit
services.
Attention to the influence of auditor industry specialists on
both the involuntaryauditor switches by former Andersen clients and
audit fees is generally lacking. Thisstudy fills that void.
Although Blouin et al. (2007) include an industry specialist
variablein their follow/non-follow model, the association between
auditor industry specialistand audit fees was not examined. Huang
et al. (2007) examine the association betweenindustry specialist,
client bargaining power, and audit fees, however, they
excludedformer Andersen audit clients from their sample. Huang et
al. (2007) fail to find audit feepremiums charged to small and
large audit clients in 2003, but in 2004 evidenceindicates that the
smaller audit clients paid an industry specialist fee premium.
MAJ28,8
712
-
Several (Kohlbeck et al., 2008; Huang et al., 2007) of the
studies discussed aboveindicate a potential size effect on pricing
audit services. In this study, the analysisfocuses on very large
Andersen clients providing more control over a potential sizeeffect
in our research design. The audit market for the S&P 1500
public corporations isheavily concentrated because the Big 4/Big 5
firms audit approximately 98 percent ofthese companies (GAO, 2008).
This study includes only those Andersen clients whoswitched to the
remaining Big 4 firms, hence, brand name reputation is not
adifferentiating factor.
Industry specialization and the pricing of audit servicesAuditor
industry specialists are perceived to offer a higher level of audit
effectivenessand quality relative to non-industry specialists (Bell
et al., 1997; AICPA, 1998;Owhoso et al., 2002; Balsam et al., 2003;
Krishnan, 2003; Carcello and Nagy, 2004;Knechel et al., 2007).
According to a GAO (2003a) survey, 81 percent of the
respondentscited industry specialization or expertise as an
important factor in choosing a newauditor. Habib (2011) states that
it is costly to develop specialization in an industrybecause a
significant amount of resources are required by the audit firm. But
oncedeveloped, a specialists knowledge of an industry and its
accounting will increase theauditors ability to detect and curb
earnings management and minimize intentionalerrors (Balsam et al.,
2003). Auditor expertise in an industry is an important factor
inreducing litigation risk, and improving auditor retention and
audit quality (Cenker andNagy, 2008). Knechel et al. (2007) find
that firms who switch from (to) a nonspecialistBig 4 auditor to
(from) a specialist Big 4 auditor experience positive
(negative)abnormal stock returns during 2000-2003, however, their
sample excludes auditorchanges that involve former Andersen
clients. Their results indicate that industryspecialization matters
to investors. Hence, it seems reasonable to assume that
formerAndersen audit clients may have wanted to contract with a Big
4 industry specialist tosignal a positive perception about
financial reporting and audit quality.
The effect of industry specialization on audit fees is still an
open question. Theassociation of audit fees and industry
specialization can have three outcomes positive,negative, or no
association. A positive association indicates a fee premium. A
feepremium from industry specialization would be consistent with
auditor differentiationthrough the acquisition of industry
specialized knowledge and with seeking to recouphigher audit
production costs (Palmrose, 1986a). Craswell et al. (1995) failed
to findconsistent support for the presence of an industry
specialist audit fee premium in thepost merger years of 1990, 1992,
and 1994. A significant and negative relationshipbetween industry
specialization and audit fees indicates a fee discount (Casterella
et al.,2004). A fee discount from auditor industry specialization
would be consistent withauditor production efficiency or production
economies where auditors pass on their costsavings to clients. An
insignificant relationship between audit fees and auditor
industryspecialization could mean industry specialization has no
effect on audit fees. Thisneutral position could also mean the
presence of both differentiation and productioneconomies offsetting
each other. Furthermore, studies show that the effect of
industryspecialization on audit fees is not only mixed, but varies
with firm size (Habib, 2011).
The present study explores the effect of industry specialization
at the national levelon audit fees from 2001 to 2002 and
involuntary auditor change by former Andersenlarge audit clients.
Prior studies suggest that the effects of the forced auditor
change
Forced auditorchange
713
-
may vary between the smaller and larger companies (Kohlbeck et
al., 2008; Huang et al.,2007) or according to whether the firm is a
national or city-specific leader (Francis et al.,2005). The effect
of industry specialization on audit fees related to city-specific
industryleaders is not examined. Given the size, resources, and
national/international presenceof our sample clients, they would
more likely be looking for national level expertise asopposed to
local-office expertise. National and international accounting firm
networksmay produce positive synergies that benefit very large
companies. As stated byReichelt and Wang (2010):
[. . .] at the firm-wide (national) level, positive synergies
arise when accounting firms captureindustry expertise through
knowledge-sharing practices, such as internal benchmarking ofbest
practices, the use of standardized industry-tailored audit
programs, and extending thereach of professionals from their
primary local-office clientele to other clients through traveland
internal consultative practices.
Thus, while industry knowledge of individual auditors in local
offices may play a rolein helping to establish the national
reputation of accounting firms, it would be difficultto argue that
very large companies (average total assets of $11 billion) with
nationaland international operations fixate on local expertise
rather than national/internationalexpertise of their auditors[2].
City-specific industry expertise was more likely anappropriate
consideration for smaller audit clients, such as those in the
Francis et al.sample with an average total assets of $1.9 billion,
than for the clients in our sample.
Further, the Big 4 accounting firms are identified as global
audit firm networks(Carson, 2009) which create industry specialist
groupings to share knowledge, staff,and resources with the
intention of improving audit quality. Carson (2009) argues thatthe
industry specialist teams of these large audit firms are supported
by knowledgemanagement databases and common industry-specific work
programs and training;and that given the significant investments in
audit technology, global audit firmnetworks are efficient
mechanisms for developing, retaining, and transferring
codifiedknowledge. These networks have been developed in part
because large companies,especially multinational operations, demand
consistent auditing throughout the world.
Many studies on the determinants of audit fees have shown that
audit fees arehigher for larger companies, but very large audit
clients may be of economicimportance to the audit firm and may have
bargaining power to attain lower audit fees(Casterella et al.,
2004). By limiting our study to the largest of Arthur Andersen
formerclients we reduce or eliminate possible confounding effects
between any premiums thatresult from industry specialization or
that may arise from low bargaining power ofsmaller audit clients.
Casterella et al. (2004) find that premiums for
industryspecialization arise when clients have low bargaining power
(in the case of smallerclients). Companies in our sample may not
have bargaining power to attain lower auditfees due to the forced
change resulting from the demise of their former auditor.Therefore,
a secondary issue in testing the auditor specialization effect is
to control andtest for the bargaining power of former Andersen
large clients.
The unique event of involuntary auditor switching by hundreds of
firms,simultaneously, presents an opportunity to draw implications
on competitive pricing(or lack thereof), price-gouging, auditor
alignment and mandatory auditor rotation.While prior research has
been mixed with respect to the effect of auditor specializationon
audit fees, we believe (like Carson, 2009) that a more competitive
response by anetwork firm to a client seeking a specialist to
signal its financial reporting quality,
MAJ28,8
714
-
especially given the tarnished reputation of its former auditor,
is to increase the auditfees to better reflect the value of the
audit (whether due to actual or perceived higherquality) and
achieve an appropriate return on the additional capital invested by
thefirm to differentiate its product. Thus, our hypothesis (in the
alternative form) is:
H1. There is a positive association between national industry
specialization andaudit fees charged to former Andersens largest
clients, ceteris paribus.
This is a one-tailed test. A positive and significant
coefficient will indicate that theformer Andersen clients incurred
a fee premium for having a national industryspecialist to audit
their financial statements, possibly to signal the quality of
theirfinancial reporting.
III. MethodologyTo test our hypothesis we construct the national
market share specialist variable,SPECMS, based on prior studies[3].
By accounting firm, industries of specializationbefore and after
Andersens demise are identified based on the national market share
ofaudit fees of the S&P 1500 companies. Industry membership of
our sample isdetermined by SIC codes similar to those of Frankel et
al. (2002) and Whisenant et al.(2003)[4]. The national market share
of audit fees per year for each accounting firm,SPECMS_c, is used
to identify the auditor industry specialists for that year.SPECMS_c
is calculated as the sum of audit fees of all companies audited by
anaccounting firm in a given industry divided by the sum of all
audit fees across all firmswithin the same industry, similar to
Casterella et al. (2004). Prior studies determiningauditor industry
specialization used sales revenues and total assets as proxies for
auditfees (Palmrose, 1986a; Mayhew and Wilkins, 2003; Balsam et
al., 2003; Neal and Riley,2004) since actual audit fee data was not
readily available publicly. Francis et al. (2005)is the first study
of industry specialist pricing in the USA to use newly mandated
auditfees disclosures beginning with 2000 fiscal-year data. While
we also use mandatoryaudit fees disclosures to determine
specialization, it is worth noting that Francis et al.investigate
Big 5 industry expertise prior to the demise of Andersen, whereas
ourstudy focuses on the alignment of former Andersen clients with a
Big 4 industryspecialist after Andersens demise.
In our analysis we use two measures of specialization:
(1) the calculated percentage of audit fees in an industry is
our continuous measure(SPECMS_c); and
(2) SPECMS is our dichotomous measure based on an audit fee
market sharespecialist minimum threshold.
Based upon the methodology used by Palmrose (1986a), an audit
fee market sharespecialist minimum threshold is 24 percent in 2001
among the Big 5, and 30 percent in2002 among the remaining Big
4[5]. If the audit fee market share for a firm is equal to
orgreater than the minimum threshold then SPECMS equals 1,
otherwise 0.
To test the hypothesis, OLS regression analysis is used to
examine the associationbetween industry specialization and audit
fees. The natural log of audit fees (LnAF) isregressed on a set of
variables that control for auditee size, profitability,
complexity,risk, client bargaining power and industry membership
similar to those used in priorstudies. The audit fees OLS model is
specified as follows:
Forced auditorchange
715
-
LnAF b0 b1LnTA b2LOSS b3ROA b4AR b5NewFIN b6TENURE b7LEV b8INV
b9MERGER b10EXDisc b11FOREIGN b12SpecItems b13AAfee b14POWER b15REG
b16SPECMS or SPECMS_c 1
Variable definitions:
LnAF Natural log of audit fees, the dependent variable.
LnTA Natural log of total assets.
LOSS Indicator variable equals 1 if the client reported a net
loss for the year,and 0 otherwise.
ROA Return on assets, measured as net income divided by total
assets.
AR Accounts receivable divided by total assets.
NewFIN Indicator variable equals 1 for clients that issued new
equity greaterthan $10 million and long-term debt greater than $1
million, and 0otherwise.
TENURE The number of years Andersen audited the company.
LEV Leverage, measured as long-term debt divided by total
assets.
INV Inventory divided by total assets.
MERGER Indicator variable equals 1 if the client engaged in
merger activity, and0 otherwise.
EXDisc Indicator variable equals 1 if the client reported
extraordinary items ordiscontinued operations during the period,
and 0 otherwise.
Foreign Indicator variable equals 1 for foreign operations if
the client reportedforeign currency adjustments, and 0
otherwise.
SpecItems Indicator variable equals 1 if the client recognized
special items, and 0otherwise. Special items are unusual in nature
or infrequent inoccurrence, but not both.
AAfee Indicator variable equals 1 if audit client paid fees to
Anderson foraudit work performed on 2002 financial statements prior
to beingbarred, and 0 otherwise.
POWER Natural log of company audit fees divided by the sum of
industryaudit fees for all companies in the industry audited by the
companysauditor.
REG Indicator variable equals 1 if client is a member of the
financialservices industry or utilities industry, and 0
otherwise.
SPECMS Industry specialist indicator variable measured based on
nationalmarket share of audit fees; equals 1 if minimum threshold
forspecialist is met, and 0 otherwise.
MAJ28,8
716
-
SPECMS_c Industry specialist continuous variable measured based
on nationalmarket share of audit fees.
1 Error term.
Control variables expected to influence the level of audit fees
(LnAF) are included in themodel for proper specification and to
avoid an omitted variable problem. In theliterature, total assets
have been found to explain much of the variability in audit
fees(Gist, 1994; Palmrose, 1986a; Simunic, 1980). The proxy for
auditee size (LnTA) isexpected to be positively associated with
LnAF. A profitable company is considered tohave less business risk
and is not expected to be charged an audit risk premium.Therefore,
the reporting of a net loss (LOSS) or a negative return on assets
(ROA)suggests increased business risk and is expected to have a
positive coefficientreflecting an increasing effect on LnAF. Other
factors that increase risk such as AR,INV, and LEV are expected to
be positively associated with LnAF. Prior studies tendnot to
associate long audit tenure with low audit quality (Nagy, 2005),
yet the uniqueenvironment of forced auditor change from an auditor
who was barred fromconducting and reporting on audits may lead to
the perception of longer tenure posinggreater risk to the new
auditor (Kealy et al., 2007). A positive coefficient for TENURE
isconsistent with increased skepticism by the new auditor.
Obtaining new financing maylower audit risk because of the
additional scrutiny of management by creditors thatoccurs during
the loan process. A negative coefficient for NewFIN is expected.
Factorsthat increase audit complexity such as SpecItems, EXDisc,
Foreign, and MERGER areexpected to be positively associated with
LnAF. A variable (AAfee) to control for feespaid to Andersen for
audit work completed in 2002 prior to its dismissal is included
inthe model. AAfee may have a decreasing effect on year 2002 audit
fees if it decreasesthe amount of audit effort by the new
auditor.
POWER is a continuous variable intended to capture the
importance of a singleclient company in an industry to its auditor.
Casterella et al. (2004) argues that thelarger the audit client the
greater is its economic importance to the auditor, which leadsto
greater bargaining power of the audit client to attain lower audit
fees. A negativecoefficient for POWER represents the ability of
large audit clients to negotiate and usetheir bargaining power to
obtain lower fees. On the other hand, one can question howmuch
bargaining power Andersen clients really had to negotiate lower
audit fees. Sinceformer Andersen clients were operating in a
different environment of mandatoryauditor change where hundreds of
clients needed to find a new auditor quickly toreplace Andersen,
this may have reduced clients bargaining power.
Nevertheless,controlling and testing bargaining power of former
Andersen clients will help to isolateits effect from that of the
auditor industry specialization variable. A positive coefficientfor
POWER may indicate large audit clients inability to negotiate and
use theirbargaining power to obtain lower fees. A positive
coefficient for POWER could alsorepresent a size effect for very
large audit clients.
Simunic (1980) and Palmrose (1986a) reports significantly lower
audit fees in theregulated industries of financials and utilities.
We therefore use indicator variables tomeasure and control the
effects of regulated financial services and utilities
industries(REG) in our model. Many studies (Simunic, 1980;
Palmrose, 1986a; Davidson and Gist,1996; Carson, 2009) in the
literature have captured and controlled the effects ofregulated
industries on audit fees or audit effort using dummy variables.
Forced auditorchange
717
-
A separate regression model is run using each measure of auditor
industryspecialist market share. A positive coefficient would
indicate an increase in audit feesconsistent with audit quality
differentiation. In contrast, a negative coefficient wouldindicate
a decrease in audit fees consistent with auditor production
efficiency where theaudit firm passes cost savings on to the audit
client.
IV. Data collectionAuditor identification and company
characteristics of the S&P 1500 were collectedfrom the
University of Pennsylvanias Wharton Research Database (WRDS). Of
theS&P 1500 companies, 1,406 public companies audited by the
Big 5 internationalaccounting firms in 2001 were identified. To
control for audit quality (or brand name)only audits by the Big 5
(Big 4) firms were considered in this study (Palmrose, 1988).We
excluded 25 Andersen client companies from the sample that switched
to a non-Big5 auditor in 2002 for a total of 1,381 companies. Of
the 1,381 companies, 269 (20 percent)were audited by Andersen. To
be included in the analysis, a former Andersen auditclient had to
have auditor fees publicly available for both 2001 and 2002; 48
clients didnot have both years of audit fees data (most often due
to mergers or bankruptcyfilings). Thus, the final sample consists
of 221 former Andersen audit clients who wereabsorbed by the
remaining Big 4 firms in 2002. Prior to the felony conviction,
Andersenwas still performing services for its audit clients. Hand
collected data was obtainedfrom the proxy statements on audit fees
received by Andersen for audit workperformed on the 2002 financial
statements (AAfee) prior to being barred fromconducting and
reporting on audits for SEC-registered companies. Under the
SECs(2001) proxy disclosure rule S7-13-00, registrants are required
to disclose audit fees forthe most recent fiscal year.
V. Analyses and resultsDescriptive statisticsTable I shows
descriptive data on the distribution of the S&P 1500 in 2001
and 2002among the international accounting firms before and after
the demise of Andersen.Panel A indicates that
PricewaterhouseCoopers (PWC) enjoyed the largest share of
theS&P 1500 audit market with 362 audits or 26.2 percent of the
market. Ernst & Young(E&Y) was second with 335 audits or
24.3 percent. Andersen followed E&Y in thirdplace with 269 or
19.5 percent of the market. KPMG had the lowest share with
195audits or 14.1 percent. There are no statistically significant
differences in the mean andmedian total assets, sales, and net
income of clients among auditors.
Panel B of Table I presents descriptive data of the Big 4 after
the collapse ofAndersen. While PWC had the most S&P 1500 audit
clients in 2001, PWC lost that leadby a very slight margin to
E&Y in 2002. Both E&Y and PWC have 30 percent of themarket.
KPMG maintained the lowest share of audit clients (18 percent)
among theS&P 1500. Analysis of variance and median tests
indicate that clients average andmedian total assets, sales, and
net income do not differ significantly among the Big 4after
Andersens demise.
Panel C of Table I shows the distribution of the 221 former
Andersen audit clientsacross the remaining Big 4 accounting firms
in 2002. Deloitte & Touche (D&T) gainedthe most former
Andersen audit clients (65 clients or 29.4 percent), and PWC gained
thefewest (40 clients or 18.1 percent). PWC gained significantly
fewer Andersen clients than
MAJ28,8
718
-
expected based on PWCs (large) 2001 share of the market shown in
Panel A ( p 0.00,x 2-test). In fact, PWC gained significantly less
than one-quarter of Andersens clients( p 0.09, x 2-test). It may be
that because PWC had more clients in 2001 than the otherBig 4, it
had less opportunity to add new clients. Panel C also shows the
total assets, sales
Panel A: distribution of S&P 1500 audit clients in 2001a
AA D&T E&Y KPMG PWC TotalNumber of clients 269 220 335
195 362 1,381Percentage of clients 20 16 24 14 26 100Total assets
($ in millions)
Mean 7,435 14,603 7,848 16,073 10,862 10,795Median 1,314 1,283
1,472 1,219 1,645 1,391
Sales ($ in millions)Mean 3,963 6,750 4,386 4,844 5,888
5,139Median 1,160 1,282 1,263 1,271 1,294 1,261
Net income ($ in millions)Mean 153 229 231 289 208 154Median 41
43 37 36 44 41
Panel B: distribution of S&P 1500 audit clients in 2002b
D&T E&Y KPMG PWC TotalNumber of clients 294 402 249 399
1,344Percentage of clients 22 30 18 30 100Total assets ($ in
millions)
Mean 13,831 7,329 14,253 12,520 11,575Median 1,475 1,552 1,390
1,717 1,523
Sales ($ in millions)Mean 5,479 4,589 4,426 5,591 5,051Median
1,280 1,160 1,179 1,276 1,227
Income ($ in millions)Mean 156 174 79 182 51Median 54 41 43 49
45
Panel C: where did they go?c
2002 size and profitability distribution of former Andersen
clientsD&T E&Y KPMG PWC Total ANOVA F-statistic
Significance
Number of clients 65 61 55 40 221Percentage of clients 29 28 25
18 100Total assets ($ in millions)
Mean 8,236 3,615 3,888 7,691 5,780 1.632 0.183Median 1,887 1,424
1,390 1,031 1,390
Sales ($ in millions)Mean 4,504 3,267 3,488 4,392 3,889 0.477
0.698Median 1,505 1,109 1,207 1,159 1,193
Net income ($ in millions)Mean 116 47 270 392 101 2.909
0.035Median 54 44 30 45 44
Notes: aThere are no statistically significant differences in
the mean and median total assets, sales,and net income of clients
among auditors; banalysis of variance and median tests indicate
that clientsaverage and median total assets, sales, and net income
do not differ significantly among the Big 4 afterAndersens demise;
cthere are no statistically significant differences in the mean and
median totalassets and sales of clients among auditors; the mean
net income between KPMG and PWC isstatistically different at the
0.05 level; AA Arthur Andersen, E&Y Ernst & Young, D&T
Deloitte & Touche, PWC PricewaterhouseCoopers
Table I.Distribution of S&P audit
clients
Forced auditorchange
719
-
and net income of former Andersen audit clients partitioned by
their new auditor.Statistically, client firm size is similarly
distributed across auditors. It seems though thatthe most
profitable audit clients selected PWC as their new auditor.
Table II provides descriptive statistics on audit fees that
clients paid to Andersen in2001 and to their new Big 4 auditors in
2002. In 2001, Andersen charged its clients anaverage of $1.07
million in audit services fees (median of $.46 million). In 2002,
however,the former Andersen clients paid their new auditors an
average of $582,000 more in auditfees than was paid to Andersen in
2001. Analysis of variance test indicates that the feedifferences
between years 2001 and 2002 are statistically significant at the
0.01 level.However, the average new audit fees that former Andersen
clients paid in 2002 do notsignificantly differ among the Big 4
accounting firms ( p . 0.10).
The average change in audit fees for the former Andersen clients
increased 54.6percent and is significantly different from 0 ( p ,
0.01). As a benchmark, the averageincrease in audit fees for
non-Andersen audit clients is 36.1 percent (not shown) and isalso
significantly different from 0 ( p , 0.01). The former Andersen
clients experiencedlarger audit fees increase, on average, relative
to audit clients that were not forced to
Audit fees ($ in 000s.)
2001 fees paid to AndersenMean 1,065Median 463n 221
2002 fees paid to new auditorsE&Y Mean 1,729
Median 622n 61
D&T Mean 1,563Median 757n 65
KPMG Mean 1,580Median 478n 55
PWC Mean 1,750Median 594n 40
Total Mean 1,647Median 613n 221
Statistical testsDo fees differ between 2001 and
2002?F-statistic 8.208Significance 0.004Do fees differ among new
auditors in 2002?F-statistic 0.076Signficance 0.973
Notes: AA Arthur Andersen, E&Y Ernst & Young, D&T
Deloitte & Touche, PWC PricewaterhouseCoopers
Table II.Descriptive statistics ofaudit fees of formerAndersen
clients in 2001and 2002
MAJ28,8
720
-
change auditors. Further tests are warranted to determine
whether the larger audit feesrepresents evidence of
price-gouging.
Industry specialist by auditorAuditor market specialist by
industry among the S&P 1500 companies shifted slightlyin 2002.
In 2002 (2001), an auditor is identified as SPECMS if a companys
auditor has30 percent (24 percent) or more market share based on
audit fees. Of the 221 Andersenclients in 15 industries, four
industries made up over half of Andersens audit clients:46 clients
in durable manufacturing, 28 clients in utilities, 20 clients in
services, and 20clients in extractive. E&Y and D&T each
gained 16 of the 46 audit clients in thedurable manufacturers
industry, while PWC absorbed only five and maintained itsindustry
specialist position. D&T absorbed 14 of the 28 utilities audit
clients andmaintained its leadership position in the industry.
Across the 15 industries and the Big4 auditors, 21 industry
specialists (SPECMS) were identified. Further analysisindicates
that of the 84 clients for which Andersen served as SPECMS in 2001,
52percent were able to obtain a Big 4 industry specialist in 2002.
Of the remaining 137Andersen clients, 45 percent (61 clients)
selected a new auditor who specialized in theirindustry, possibly
to communicate financial reporting quality. Overall, the number
ofAndersen clients who had engaged an industry specialist increased
from 38 percent (84clients) in 2001 to 48 percent (105 clients) in
2002, a net increase of 10 percentattributable to the alignment
with a Big 4 industry specialist.
Correlation analysisTable III shows positive and significant
correlations for year 2001 between LnAF andLnTA (0.80), EXDisc
(0.31), TENURE (0.27), and POWER (0.45). As expected, SPECMSand
SPECMS_c are highly correlated at (0.86). In addition, positive and
significantcorrelations exist between LnTA and NewFIN (0.36),
TENURE (0.25), LEV(0.41),EXDisc (0.33), POWER (0.37), and REG
(0.34). Positive and significant correlations foryear 2002 are
similar to year 2001. The correlations do not appear to present a
problemwith collinearity. This observation is confirmed by the
calculated variance inflationfactors (VIFs) reported for the
models. Variable definitions are given in Section III.
Regression analysisOLS regression models for testing the
hypothesis are presented in Table IV. Models inPanel A are based on
audit fees paid to Andersen in 2001, while the models in Panel B
arebased on audit fees paid by former Andersen clients to new
auditors in 2002. For eachyear, the audit fees model is run three
times. The first model is the base model withoutthe test variable.
The second model includes the SPECMS indicator variable. In the
thirdmodel, the SPECMS_c continuous variable is substituted for the
SPECMS variable. Inthe 2001 OLS regressions (Panel A), the adjusted
R 2 of 72.8 percent indicates the powerof the independent variables
(excluding the specialist test variable) to explain thedependent
variable, LnAF, the natural log of audit fees. In 2002 (Panel B),
the adjustedR 2 indicates that the independent variables (excluding
the specialist test variable)explain 76.2 percent of audit fees.
For both years, the models are significant at the 0.01level. Also
for both years, the F-tests for incremental explanatory power of
the industryspecialist variables are significant at the 0.05 level.
The OLS assumptions are notviolated. Since all VIFs are below 3.0,
multicollinearity does not appear to be a problem.
Forced auditorchange
721
-
PanelA:year2001 Ln
AF
Ln
TA
LO
SS
RO
AA
RN
ewF
INT
EN
UR
EL
EV
INV
ME
RG
ER
EX
Dis
cF
OR
EIG
NS
pec
Item
sP
OW
ER
RE
GS
PE
CM
SL
nT
A0.
804
**
0.00
0L
OS
S2
0.12
02
0.17
4*
*
0.07
60.
010
RO
A2
0.01
80.
002
20.
646
**
0.78
80.
973
0.00
0A
R0.
095
20.
079
20.
082
0.05
60.
159
0.24
00.
223
0.40
7N
ewF
IN0.
211
**
0.36
4*
*2
0.08
02
0.04
72
0.07
50.
002
0.00
00.
233
0.48
70.
270
TE
NU
RE
0.27
2*
*0.
253
**
0.10
22
0.08
02
0.11
50.
084
0.00
00.
000
0.12
90.
236
0.08
70.
215
LE
V0.
187
**
0.41
1*
*0.
139
*2
0.21
3*
*2
0.28
7*
*0.
272
**
0.21
9*
*
0.00
50.
000
0.03
90.
001
0.00
00.
000
0.00
1IN
V2
0.01
72
0.15
3*
0.05
30.
047
0.03
52
0.07
10.
156
*2
0.18
0*
*
0.80
60.
023
0.42
90.
491
0.60
30.
293
0.02
00.
007
ME
RG
ER
0.12
10.
145
*0.
037
20.
058
20.
037
0.22
4*
*0.
010
0.11
70.
004
0.07
10.
031
0.58
70.
393
0.58
90.
001
0.87
70.
083
0.94
8E
XD
isc
0.31
3*
*0.
331
**
0.00
82
0.10
62
0.16
6*
0.13
20.
141
*0.
298
**
20.
129
0.11
60.
000
0.00
00.
908
0.11
80.
013
0.05
00.
036
0.00
00.
055
0.08
5F
OR
EIG
N0.
102
20.
010
0.04
82
0.06
00.
050
0.01
40.
060
20.
119
20.
003
0.08
72
0.11
90.
129
0.88
00.
477
0.37
30.
463
0.84
10.
378
0.07
80.
962
0.19
80.
077
Sp
ecIt
ems
0.13
10.
038
0.29
2*
*2
0.27
1*
*0.
013
20.
022
0.09
30.
023
0.03
50.
010
0.05
80.
006
0.05
10.
576
0.00
00.
000
0.84
30.
747
0.16
60.
729
0.60
90.
888
0.38
90.
933
PO
WE
R0.
446
**
0.36
7*
*2
0.10
70.
110
0.06
40.
157
*0.
174
**
0.03
90.
027
0.11
30.
164
*0.
022
0.11
20.
000
0.00
00.
113
0.10
40.
343
0.01
90.
010
0.56
60.
691
0.09
40.
015
0.74
30.
098
RE
G0.
105
0.34
3*
*2
0.21
6*
*2
0.00
32
0.12
80.
157
*2
0.10
90.
278
**
20.
341
**
20.
060
0.15
2*
20.
205
**
20.
234
**
20.
087
0.11
80.
000
0.00
10.
969
0.05
80.
019
0.10
50.
000
0.00
00.
376
0.02
40.
002
0.00
00.
198
SP
EC
MS
0.20
3*
*0.
211
**
0.00
12
0.07
00.
033
0.14
1*
0.13
8*
0.22
1*
*2
0.09
40.
019
0.13
22
0.09
32
0.05
72
0.04
90.
252
**
0.00
20.
002
0.99
00.
302
0.62
10.
036
0.04
00.
001
0.16
60.
783
0.05
00.
166
0.39
60.
466
0.00
0S
PE
CM
S_
c0.
200
**
0.24
1*
*2
0.07
32
0.03
12
0.03
70.
142
*0.
136
*0.
241
**
20.
084
0.03
80.
137
*2
0.09
82
0.11
42
0.14
0*
0.26
5*
*0.
864
**
0.00
30.
000
0.28
00.
647
0.58
90.
035
0.04
40.
000
0.21
30.
577
0.04
20.
145
0.09
10.
037
0.00
00.
000
(continued
)
Table III.Correlation matrix
MAJ28,8
722
-
PanelB:year2002 Ln
AF
Ln
TA
LO
SS
RO
AA
RN
ewF
INT
EN
UR
EL
EV
INV
ME
RG
ER
EX
Dis
cF
OR
EIG
NS
pec
Item
sA
Afe
eP
OW
ER
RE
GS
PE
CM
SL
nT
A0.
823
**
0.00
0L
OS
S0.
144
*2
0.02
50.
032
0.71
0R
OA
20.
131
20.
043
20.
662
**
0.05
10.
529
0.00
0A
R0.
026
20.
106
0.05
72
0.06
00.
704
0.11
60.
398
0.37
7N
ewF
IN0.
179
**
0.26
7*
*0.
105
20.
133
*2
0.03
50.
008
0.00
00.
118
0.04
90.
607
TE
NU
RE
0.27
4*
*0.
233
**
0.09
82
0.06
22
0.12
80.
033
0.00
00.
000
0.14
50.
356
0.05
80.
629
LE
V0.
299
**
0.44
1*
*0.
165
*2
0.18
3*
*2
0.27
9*
*0.
329
**
0.19
8*
*
0.00
00.
000
0.01
40.
006
0.00
00.
000
0.00
3IN
V2
0.12
82
0.16
1*
0.05
70.
029
0.05
02
0.13
6*
0.16
3*
20.
220
**
0.05
80.
016
0.40
10.
669
0.45
70.
044
0.01
60.
001
ME
RG
ER
0.06
72
0.00
22
0.04
30.
008
0.03
42
0.04
52
0.11
82
0.05
00.
107
0.31
90.
980
0.52
10.
911
0.61
50.
503
0.08
00.
462
0.11
4E
XD
isc
0.30
4*
*0.
199
**
0.37
9*
*2
0.22
3*
*2
0.12
10.
105
0.10
30.
280
**
20.
074
20.
018
0.00
00.
003
0.00
00.
001
0.07
20.
119
0.12
60.
000
0.27
60.
786
FO
RE
IGN
0.14
5*
0.01
60.
129
20.
120
0.04
82
0.01
20.
040
20.
138
*0.
000
0.10
82
0.02
50.
031
0.80
80.
056
0.07
50.
482
0.86
50.
553
0.04
11.
000
0.10
80.
716
Sp
ecIt
ems
0.13
8*
0.03
50.
252
**
20.
194
**
20.
045
20.
105
0.10
50.
028
0.19
2*
*0.
155
*0.
183
**
0.16
3*
0.04
00.
607
0.00
00.
004
0.50
60.
119
0.11
80.
683
0.00
40.
021
0.00
60.
015
AA
fee
20.
020
0.00
20.
006
20.
054
20.
051
20.
079
20.
064
0.02
02
0.14
6*
20.
039
20.
015
20.
053
0.05
80.
764
0.98
00.
925
0.42
80.
448
0.24
00.
345
0.76
50.
030
0.56
60.
829
0.43
30.
387
PO
WE
R0.
550
**
0.50
7*
*0.
137
*2
0.04
52
0.07
70.
141
*0.
111
0.24
6*
*2
0.10
22
0.05
70.
153
*0.
039
0.07
80.
000
0.00
00.
000
0.04
10.
509
0.25
30.
036
0.10
10.
000
0.12
90.
401
0.02
30.
568
0.24
80.
996
RE
G0.
200
**
0.33
8*
*2
0.08
12
0.00
22
0.14
3*
0.12
92
0.10
90.
329
**
20.
336
**
20.
076
0.14
7*
20.
232
**
20.
309
**
0.03
20.
125
0.00
30.
000
0.23
30.
973
0.03
40.
055
0.10
50.
000
0.00
00.
263
0.02
90.
001
0.00
00.
638
0.06
4S
PE
CM
S0.
145
*0.
155
*2
0.20
8*
*0.
147
*0.
098
20.
039
0.00
20.
011
20.
041
0.10
22
0.05
42
0.09
52
0.16
6*
0.06
22
0.12
70.
200
**
0.03
20.
021
0.00
20.
029
0.14
70.
568
0.98
00.
869
0.54
40.
130
0.42
20.
160
0.01
40.
361
0.05
90.
003
SP
EC
MS
_c
0.10
60.
134
*2
0.17
8*
*0.
089
0.00
40.
016
0.06
20.
062
20.
011
0.08
12
0.04
82
0.02
02
0.15
0*
0.04
12
0.20
3*
*0.
207
**
0.82
6*
*
0.11
50.
047
0.00
80.
189
0.94
80.
813
0.36
30.
359
0.86
80.
229
0.47
40.
770
0.02
60.
541
0.00
20.
002
0.00
0
Note:
Cor
rela
tion
issi
gn
ifica
nt
at:
* 0.0
5an
d*
* 0.0
1le
vel
s(t
wo-
tail
ed),
firs
tv
alu
eis
the
corr
elat
ion
coef
fici
ent
and
the
seco
nd
val
ue
issi
gn
ifica
nce
lev
el
Table III.
Forced auditorchange
723
-
Var
iab
leE
xp
ecte
dsi
gn
Coe
ff.e
stim
ate
tS
ig.
VIF
Coe
ff.e
stim
ate
tS
ig.
VIF
Coe
ff.
esti
mat
et
Sig
.V
IF
PanelA:year2001basedon
auditfees
paid
Andersen(n
221)
Ln
AFb
0b
1L
nT
Ab
2L
OS
Sb
3R
OAb
4A
Rb
5N
ewF
INb
6T
EN
UR
Eb
7L
EVb
8IN
Vb
9M
ER
GE
Rb
10
EX
Dis
cb
11
FO
RE
IGNb
12
Sp
ecIt
emsb
13
AA
feeb
14
PO
WE
Rb
15
RE
Gb
16
SP
EC
MS
(or
SP
EC
MS
_c)1
DependentvariableLnAFisbasedon
auditfees
paid
toAndersenin
year2001
(Con
stan
t)1.
809
7.54
80.
000
1.84
77.
777
0.00
01.
767
7.42
80.
000
Ln
TA
0.
589
17.2
660.
000
1.9
0.58
317
.257
0.00
01.
90.
578
16.9
760.
000
1.9
LO
SS
(0
.055
)(0
.394
)0.
347
2.0
(0.0
65)
(0.4
69)
0.32
02.
0(0
.042
)(0
.300
)0.
382
2.0
RO
A2
(0.6
52)
(1.0
46)
0.14
81.
9(0
.638
)(1
.034
)0.
151
1.9
(0.6
21)
(1.0
07)
0.15
81.
9A
R
1.17
63.
266
0.00
11.
11.
058
2.94
30.
002
1.2
1.11
83.
132
0.00
11.
1N
ewF
IN2
(0.2
00)
(1.9
11)
0.02
91.
2(0
.215
)(2
.076
)0.
020
1.2
(0.2
10)
(2.0
32)
0.02
21.
2T
EN
UR
E
0.00
91.
088
0.13
91.
20.
006
0.76
30.
223
1.2
0.00
60.
774
0.22
01.
2L
EV
(0
.680
)(2
.504
)0.
007
1.6
(0.7
35)
(2.7
28)
0.00
31.
6(0
.737
)(2
.733
)0.
003
1.6
INV
0.
546
1.55
30.
061
1.2
0.55
61.
600
0.05
61.
20.
544
1.56
40.
060
1.2
ME
RG
ER
(0
.012
)(0
.130
)0.
448
1.1
(0.0
13)
(0.1
42)
0.44
41.
1(0
.018
)(0
.199
)0.
421
1.1
EX
Dis
c
0.22
82.
488
0.00
71.
20.
217
2.39
70.
009
1.2
0.21
82.
403
0.00
91.
2F
orei
gn
0.
235
2.19
70.
015
1.1
0.24
92.
351
0.01
01.
10.
251
2.36
50.
009
1.1
Sp
ecIt
em
0.10
71.
234
0.10
91.
20.
114
1.33
70.
091
1.2
0.12
21.
417
0.07
91.
2P
OW
ER
^0.
962
2.82
10.
005
1.3
1.04
73.
088
0.00
21.
31.
142
3.30
20.
001
1.4
RE
G2
(0.2
20)
(1.7
93)
0.03
71.
6(0
.266
)(2
.164
)0.
016
1.7
(0.2
52)
(2.0
61)
0.02
01.
6S
PE
CM
S
0.20
52.
407
0.00
81.
2S
PE
CM
S_
c
0.87
52.
346
0.01
01.
2A
dju
sted
R2
(per
cen
t)72
.873
.573
.4F
-sta
tist
ic43
.15
41.5
941
.52
Sig
nifi
can
ce0.
000
0.00
00.
000
(continued
)
Table IV.OLS regression
MAJ28,8
724
-
Var
iab
leE
xp
ecte
dsi
gn
Coe
ff.e
stim
ate
tS
ig.
VIF
Coe
ff.e
stim
ate
tS
ig.
VIF
Coe
ff.
esti
mat
et
Sig
.V
IF
PanelB:year2002basedon
auditfees
paid
totheremainingBig
4auditors(n
221)
Ln
AFb
0b
1L
nT
Ab
2L
OS
Sb
3R
OAb
4A
Rb
5N
ewF
INb
6T
EN
UR
Eb
7L
EVb
8IN
Vb
9M
ER
GE
Rb
10
EX
Dis
cb
11F
OR
EIG
Nb
12
Sp
ecIt
emsb
13
AA
feeb
14
PO
WE
Rb
15R
EGb
16S
PE
CM
S(o
rS
PE
CM
S_
c)1
DependentvariableLnAFisbasedon
auditfees
paid
byAndersenform
erclientsto
new
auditor
inyear2002
(Con
stan
t)1.
722
6.87
0.00
1.77
87.
129
0.00
1.66
06.
624
0.00
0.0
Ln
TA
0.
581
16.3
900.
001.
90.
565
15.7
730.
002.
00.
568
15.9
150.
002.
0L
OS
S
0.23
61.
657
0.05
2.2
0.25
41.
803
0.04
2.2
0.25
51.
804
0.04
2.2
RO
A2
(0.0
28)
(0.0
59)
0.48
1.9
(0.0
90)
(0.1
92)
0.42
1.9
(0.0
44)
(0.0
95)
0.46
1.9
AR
1.
197
3.13
40.
001.
11.
080
2.82
70.
001.
21.
162
3.06
20.
001.
1N
ewF
IN2
(0.0
90)
(0.9
43)
0.17
1.2
(0.0
76)
(0.8
01)
0.21
1.2
(0.0
88)
(0.9
28)
0.18
1.2
TE
NU
RE
0.
025
2.86
30.
001.
20.
024
2.79
00.
001.
20.
023
2.69
30.
001.
2L
EV
(0
.618
)(2
.201
)0.
011.
7(0
.613
)(2
.203
)0.
011.
7(0
.633
)(2
.272
)0.
011.
7IN
V
(0.5
31)
(1.3
82)
0.08
1.3
(0.5
50)
(1.4
43)
0.08
1.3
(0.5
80)
(1.5
17)
0.07
1.3
ME
RG
ER
0.
224
2.24
20.
011.
10.
197
1.97
50.
021.
10.
205
2.06
10.
021.
1E
XD
isc
0.
314
3.32
60.
001.
30.
314
3.36
20.
001.
30.
317
3.38
40.
001.
3F
orei
gn
0.
241
2.30
10.
011.
10.
250
2.41
40.
011.
20.
229
2.20
40.
011.
2S
pec
Item
0.
081
0.85
70.
201.
30.
102
1.08
30.
141.
30.
098
1.04
10.
151.
3A
Afe
e
(0.0
31)
(0.3
21)
0.37
1.1
(0.0
48)
(0.5
10)
0.31
1.1
(0.0
44)
(0.4
63)
0.32
1.1
PO
WE
R^
1.90
84.
050
0.00
1.4
2.12
64.
464
0.00
1.5
2.18
74.
489
0.00
1.5
RE
G2
(0.0
14)
(0.1
15)
0.45
1.6
(0.0
49)
(0.3
98)
0.35
1.6
(0.0
55)
(0.4
39)
0.33
1.6
SP
EC
MS
0.
209
2.26
20.
011.
2S
PE
CM
S_
c
0.63
52.
034
0.04
1.2
Ad
just
edR
2(p
erce
nt)
76.2
76.6
76.5
F-s
tati
stic
47.9
046
.13
45.8
5S
ign
ifica
nce
0.00
00.
000
0.00
0
Table IV.
Forced auditorchange
725
-
For both years in each model run, the coefficients for LnTA, AR,
EXDisc, and Foreignare positive and significant at the 0.01 levels
indicating that size, risk and complexityare positively associated
with audit fees. POWER is also positive and significant at the0.01
level in each model. Rather than observing negative coefficients,
indicating thebargaining power of large audit clients to negotiate
lower audit fees consistent with theresults of Casterella et al.
(2004), these findings indicate that large audit clients ofAndersen
are charged a premium fee, possibly to compensate the auditor for
greateraudit effort due to increased risk and complexity of the
audit. Another possibleexplanation for the result of the POWER
variable is that former Andersen clients didnot have much
bargaining power due to the forced change and the need to
quicklyengage a new auditor. In 2001, but not in 2002, the
coefficients for INV (proxy for risk)and SpecItem (proxy for
complexity) are consistently positive and significant asexpected at
the 0.10 levels. In year 2001, but not in 2002, the coefficients
for NewFIN(proxy for increased scrutiny) and REG (regulated
industries) are significantlynegative as expected in each run at
the 0.05 level or better. In 2002, but not in 2001, thecoefficients
for LOSS (proxy for low profitability), Merger (proxy for audit
complexity)and TENURE (proxy for professional skepticism) are
consistently positive andsignificant as expected at the 0.05 levels
or better. The TENURE variable indicates thataudit fee premiums are
associated with clients who were with the Andersen accountingfirm
for many years, and is consistent with increased risk perception
and a healthydose of professional skepticism by the new auditor
(Kealy et al., 2007).
Support is provided for H1. As Table IV, Panel B shows for year
2002, formerAndersen clients who were absorbed by a Big 4 industry
specialist paid an audit feespremium. The coefficients for SPECMS
(0.209) and SPECMS_c (0.635) are positive andsignificant at the
0.01 and 0.05 levels, respectively. For year 2001, Panel A shows
thatclients for which Andersen represented an industry specialist
paid an audit feespremium to Andersen. The coefficients for SPECMS
(0.205) and SPECMS_c (0.875) arepositive and significant at the
0.01 level. These findings suggest that Andersen and thenew
auditors who specialized in an industry based upon market share (of
audit fees)charged the Andersen clients an incremental fee
consistent with the productdifferentiation explanation. We use the
procedure described by Craswell et al. (1995)and Simon and Francis
(1988) to calculate the audit fee premium[6]. On average, theformer
Andersen audit clients paid a premium of 23.2 percent to their Big
4 industryspecialist auditor in 2002; this compares to a 22.8
percent premium paid to Andersenby clients for which it represented
an industry specialist in 2001.These results aresubstantially
unchanged when replacing the regulated industry indicator variable
withindicator variables for all industries except for one, the
effect of which is embedded inthe intercept.
Additional analysesMembers of the Big 4 could exercise their
dominant market power to charge formerAndersen clients
uncompetitive fees (Yardley et al., 1992; AccountingWEB.com,
2002;Wall Street Journal, 2003). As we reported previously, former
Andersen clientsexperienced larger audit fee increases from 2001 to
2002, on average, relative to auditclients that were not forced to
change auditors. Since regulators have expressedconcern for
potential price-gouging (GAO, 2003a, 2008), we test for it
following themethodology employed by Ettredge and Greenberg (1990),
which requires analyzing
MAJ28,8
726
-
the difference in deflated residuals (residual/audit fees)
between 2001 and 2002 forevidence of price-gouging and/or
low-balling. Regressions are run for each year and theresiduals
saved. Next, ratios are calculated by dividing the residuals
derived from theregression by its dependent variable. Then the 2001
ratios are subtracted from the 2002ratios to calculate the change
in deflated residuals. Results of parametric(nonparametric) tests
of the mean (median) values (not shown) indicate that
neitherprice-gouging nor low-balling were pervasive in the pricing
of audits of Andersensformer clients after controlling for factors
that are related to audit fees.
Next, we use a logistic model to explore the probability of
factors that mayinfluence former Andersen clients switching to an
industry specialist since auditorexpertise in an industry is an
important factor affecting the positive perception ofaudit
effectiveness and audit quality relative to non-industry
specialists (AICPA, 1998;Owhoso et al., 2002; Balsam et al., 2003;
Cenker and Nagy, 2008). We expect thatformer Andersen clients are
motivated to signal a perception of lower risk, high auditquality
and increased reliability of financial statements. In the logistic
model, auditorindustry specialist, a dichotomous variable, is the
dependent variable. This testprovides an indication of the
likelihood of former Andersen audit clients switching to anew
industry specialist Big 4 auditor in 2002. The model appears to
appropriatelycapture variation in the dependent variable as
evidenced by the inability to reject thenull of an appropriate
model fit indicated by the Omnibus test of model
coefficients(results not shown). In SPSS binary logistic regression
report, significance levels bythe traditional x 2 method are an
alternative to the Hosmer-Lemeshow x 2-testgoodness-of-fit. The
pseudo R 2 is 27.6 percent for the logistic regression.
Resultsindicate that large audit clients who had long auditor
tenure with Andersen and highleverage were more likely to switch to
an industry specialist. This finding is notinconsistent with a
management risk perspective in which former Andersen clientslikely
want to signal to the market their financial reporting quality
because of theirlong tenure with an auditor who had been barred
from performing audits, and becauseof higher than normal business
risk.
Since our sample consists of very large companies as a design
feature controllingextraneous size effects, additional tests are
performed to determine whether companysize still could be driving
the results as suggested by prior studies (Huang et al., 2007;Hay
and Jeter, 2011): for example, Hay and Jeter (2011) state that the
specialistpremium applies most consistently to larger client
companies; and Carcello and Nagy(2004) find a positive and
significant interaction between industry specialization andclient
size. Although client size (LnTA) and bargaining power (POWER) are
controlledin our model, an additional size variable (MedTA) is
included to ascertain whetherthere is any change in coefficient of
our specialization test variable. MedTA equals 1 iftotal assets is
above its median, and 0 otherwise. When the regressions in Table IV
arererun with MedTA included, MedTA is significant ( p 0.01) and
the industryspecialization variables continue to be positive and
significant at the 0.05 level orbetter. Similarly, another
additional size measure, MedPOWER, is included in theoriginal
regressions. MedPOWER equals 1 if POWER is above its median, and
0otherwise. When MedPOWER is tested, the findings on the
specialization test variablesremain unchanged. Hence, our findings
are robust to the inclusion of additionalalternative measures of
company size in the regression models, suggesting that size isnot
driving our results as in prior research studies.
Forced auditorchange
727
-
We further explore the industry specialization variable by
partitioning it into twoindicator variables:
. first variable indicates industry specialist in 2002 that was
also considered anindustry specialist in 2001 (n 44); and
. second variable indicates industry specialist in 2002 that was
not considered anindustry specialist in 2001 (n 61).
The adjusted R 2 increased slightly to 77.4 percent from 76.6
percent, and only the firstindicator variable is positive and
significant (at the 0.01 level). This finding suggeststhat auditors
who became industry specialist because of the new Andersen clients
intheir portfolio did not charge a specialist premium to former
Andersen clients. Perhapsas knowledge and experience is acquired
from serving these new clients an audit feespremium may be
charged.
Finally, we ran a change regression to ascertain whether a
forced change isassociated with a significant fee premium for
clients with specialist auditors. Theadjusted R 2 is 54 percent,
and the auditor industry specialization variable is notsignificant.
When the specialization variable is partitioned into two indicator
variablesas discussed above, neither indicator variable is
significant. This finding suggests thatthe forced change is not
driving the result on the auditor specialization variable in
thelevel regressions. Further, while we test for and find a
significant increase in audit fees(after controlling for other
factors) from 2001 to 2002, there is no evidence ofprice-gouging as
previously reported in this section.
VI. Conclusion, contribution, and implicationsThis study
explores the effect of industry specialization and competitive
pricing (orlack thereof) of audits of the S&P 1500 former
Andersen clients who switched to one ofthe remaining Big 4 auditors
in 2002. The collapse of Andersen increased audit
marketconcentration significantly in the large client segment and
induced an abrupt forcedauditor change for hundreds of audit
clients at one time, allowing us the opportunityto examine the
effect of auditor specialization on audit fees in this
uniqueenvironment and its implications for mandatory auditor
rotation. Because prior auditorspecialization studies report size
effects, we limit our study to the largest ofArthur Andersens
former clients to provide more control over size, by design, and
toreduce or eliminate possible confounding effects between any
premiums resulting fromindustry specialization and those that may
arise from a companys potentialbargaining power to obtain lower
audit fees. We argue that given the size, resources,and
national/international presence of the S&P 1500, that these
companies are morelikely concerned with the firm-wide and
international reputation of their auditors asoppose to the auditors
local-office reputation. A basis for this study is that we
believe,given the tarnished reputation of Andersen, its large
former clients:
. wanted to send a positive signal about the quality of their
financial reporting;and
. would likely have done so by engaging a Big 4 industry
specialist.
We also believe that a more likely competitive response by a Big
4 network firm wouldbe to increase audit fees (rather than discount
them) to better reflect the value of the
MAJ28,8
728
-
audit and earn an appropriate return on the investment in
specialization. Consequently,we examine if there is a positive
association between national industry specializationand audit fees
charged to former Andersens largest clients.
As hypothesized, we find that the association between audit fees
and auditorindustry specialization is positive and significant at
the 0.05 level or better, whichsupports the product differentiation
explanation. Our finding is consistent with theFrancis et al.
(2005) model, which considers national-only industry leaders. On
average,the former Andersen audit clients paid a premium of 23.2
percent to their Big 4industry specialist auditor in 2002; this
compares to a 22.8 percent premium paid toAndersen by clients for
whom it represented an industry specialist in 2001.
This study makes a contribution to the literature beyond Francis
et al. (2005) byfocusing on the alignment of former Andersen
clients with a Big 4 industry specialist(in a forced auditor
change) after the demise of Andersen. In contrast, Francis et
al.investigate Big 5 industry expertise prior to Andersens demise.
While both this studyand Francis et al. (2005) use the newly
required mandatory audit fees disclosures, wefocus our study on
large S&P 1500 clients, and are able to avoid size effect
issuesreported by Francis et al. and other prior studies because
the premium for industryleadership is driven by the upper half of
company size.
This study contributes to the literature beyond the primary
finding as it relates toclient bargaining power, price-gouging, and
the likely characteristics of companiesinfluencing the choice of an
auditor industry specialist. When testing the auditorindustry
specialization hypothesis, we include a POWER variable in the model
as aproxy for client bargaining power to attain lower audit fees,
similar to Casterella et al.(2004). We find that the POWER variable
is positive and significant ( p , 0.01),suggesting that POWER
represents a size effect of the large and complex formerAndersen
audit clients. Further, results of the POWER variable remained
unchangedwhen conducting additional analyses. Contrary to
Casterella et al. (2004), ourexplanation is that large former
Andersen clients did not have much bargaining powerdue to the
forced change and the need to quickly engage a new auditor. This
may haveimplications for mandatory auditor rotation (discussed
below), whereby largecompanies would likely lose their ability to
bargain for lower audit fees.
Additionally, given the significant increase in audit market
concentration for largepublic companies along with involuntary
auditor switching, regulators have expressedconcern about potential
price-gouging (GAO, 2003a, 2008). We test for excessivepricing (and
low-balling) using a method employed by Ettredge and Greenberg
(1990),but could not find any evidence of price-gouging by the Big
4 as had been feared byregulators after the demise of Anderson. It
is difficult to extend pricing implications ifone or more of the
international accounting firms is abruptly barred from
providingaudit services to their clients because the accounting
profession will be faced with ahost of complex problems (including
lost investor confidence) along with an increase inaudit market
concentration and the possibility of excessive pricing.
Finally, our study provides further evidence of how companies
perceive theimportance of having the financial statements audited
by an industry specialist in orderto signal quality financial
reporting to the market. First, we analyzed the companiesthat had a
specialist (non-specialist) in 2001 and retained a specialist in
2002. The dataindicates that the tarnished audit quality reputation
of Andersen resulted in a net 10percent increase in the number of
Andersens former clients who obtained an industry
Forced auditorchange
729
-
specialist in 2002 (over 2001). Second, we employed a logistic
model using an auditorindustry specialist dichotomous variable as
the dependent variable to determinecompany characteristics that are
more likely to influence the engagement of anindustry specialist
among the large former Andersen clients. We find that
formerAndersen clients with the characteristics of long tenure and
high business risk weremore likely to switch to an industry audit
specialist. Supporting the basic premise ofour study, this finding
implies that companies align with a specialist for the majorbenefit
of signaling financial reporting and audit quality to the market.
Our study isalso consistent with a GAO (2003a) survey which
indicated that 81 percent ofrespondents cited industry
specialization or expertise as an important factor inchoosing a new
auditor.
Implications of this research are that mandatory audit firm
rotation (which iscurrently being debated in the profession) or
involuntary change of auditors will have acostly effect on the
pricing of audit services for those companies that use a Big
4network firm. The GAO (2003b) report states that most public
companies will only usethe Big 4 firms for audit services. Given
this preference, these public companies mayonly have one or two
real choices for an auditor of record under any mandatoryrotation
system given the importance of industry expertise and the
Sarbanes-OxleyActs auditor independence requirements. Under the
Sarbanes-Oxley Acts auditorindependence requirements, audit firms
are prohibited from providing both audit andnon-audit services to
the same client. With the limited choices for an auditor of
recordamong the Big 4, involuntary auditor changes for those firms
wanting an auditor thatspecializes in their industry will more than
likely incur additional cost in the form ofhigher audit fees.
The sample only includes relatively large companies, which may
limit the ability togeneralize the findings to smaller
companies.
Notes
1. The Herfindahl-Hirschman Index measures the relative
concentration of market power heldby the largest firms in an
industry and represents the degree to which the industry
isoligopolistic.
2. For a general discussion of network synergies, please refer
to studies by Katz and Shapiro(1985) and Bental and Spiegel (1995)
about scope and size of relevant (appropriate) networks,network
externalities, and the quality of a network product.
3. Methodologies used to identify firms as industry audit
specialists lack consistency and arethus difficult to compare and
evaluate findings (Neal and Riley, 2004).
4. Industry membership is determined by SIC code as follows:
mining and construction(1000-1999, excluding 1300-1399), food
(2000-2111), textiles and printing/publishing(2200-2799), chemicals
(2800-2824 and 2840-2899), biotechnology/pharmaceuticals(2830-2836
and 8731-8734), extractive (1300-1399 and 2900-2999), durable
manufacturers(3000-3999, excluding 3570-3579 and 3670-3674),
computers (3570-3579 and 7370-7379),transportation (4000-4899),
retail-wholesale (5000-5999, excluding 5200-5961),
services(7000-8999, excluding 7370-7379), financial services
(6021-6798), utilities electric and gas(4900-4940), retail-other
(5200-5961), and other (000-0999, 9000-9999).
5. The specialist cutoff is based on studies by Palmrose (1986a)
and Neal and Riley (2004).When the audit market consisted of the
Big 8, each firm without specialization holds anequal market share
of 12.5 percent. Palmrose (1986a) specified an auditor as a
specialist if the
MAJ28,8
730
-
market share was 20 percent or greater above this, thereby
specifying the specialist cutoff at15 percent (0.125 1.2 0.15).
Using this same methodology, the calculated specialist cutoffis 24
percent (30 percent) when the industry consisted of the Big 5 (Big
4) auditors.
6. The procedure described by Craswell et al. (1995) and Simon
and Francis (1988) is used tocalculate the audit fee premium. The
percentage shift in audit fees in the fitted regressionmodel is
estimated to infer the magnitude of change in audit price
attributable to industryspecialization. Therefore, in addition to
the statistical tests of parameter b15 to determinewhether there is
a significant intercept shift, the magnitude of the intercept shift
iscalculated. The shift in the intercept term affects audit fees in
the fitted model in thefollowing manner.
exz 2 ex
ex
where, ex audit fees of clients without an industry specialist
auditor; e(x z) audit feesof clients with an industry specialist
auditor, where z is the upward shift in the intercept termdue to
the SPECMS variable. The above equation simplifies to ez 2 1, which
is solved usingthe mean parameter value (z) of the SPECMS variable
in the fitted regression model. Thisexpresses the mean shift in
industry specialist audit fees as a percentage of
non-industryspecialist audit fees.
References
AccountingWEB.com (2002), Accounting firms expect double-digit
hikes in audit fees,August 13.
AICPA (1998), CPA vision project identifies top five issues for
profession, The CPA Letter,Vol. 78, April, p. 12.
Balsam, S., Krishnan, J. and Yang, J.G.S. (2003), Auditor
industry specialization and theearnings response coefficient,
Auditing: A Journal of Practice and Theory, Vol. 22September, pp.
71-97.
Barton, J. (2005), Who cares about auditor reputation?,
Contemporary Accounting Research,Vol. 22, pp. 549-586.
Beck, P.J., Frecka, T.J. and Solomon, I. (1988), A model of the
market for MAS and audit ser