Asthana Et Al (2010)

International Journal of Auditing doi:10.1111/j.1099-1123.2010.00417.
x
Int. J. Audit. 14: 274–293 (2010)
Corporate Governance, Audit Firm

Reputation, Auditor Switches, and
Client Stock Price Reactions: The
Andersen Experience ija_417 274..293
Sharad C. Asthana,1 Steven Balsam2 and Jagan Krishnan2

1
University of Texas at San Antonio
2
Temple University
The financial scandal surrounding the collapse of Enron

caused erosion in the reputation of its auditor, Arthur
Andersen, leading to concerns about Andersen’s ability to
continue in existence and ultimately to the firm’s demise. In
this paper we investigate the role of corporate governance on
the timing of the auditor switch by former Andersen clients.
After controlling for factors associated with switching costs, we
find clients with strong corporate governance were more likely
to switch early. We also find that clients switching from
Andersen experienced positive abnormal returns during the
three-day window surrounding the announcement of the
switch. We attribute this positive response to the reduction in
uncertainty associated with the cost of finding a new auditor.
Key words: Auditor reputation, auditor change, Arthur Andersen,

Enron, corporate governance
SUMMARY auditor change by former Andersen clients. After

controlling for costs associated with the auditor
The financial fraud surrounding the collapse of change, we find clients with strong corporate
Enron tarnished the reputation of its auditor, governance were more likely to change to a new
Arthur Andersen, one of the Big 5 CPA firms at the auditor early. We also find that clients changing to
time. This raised serious concerns about Arthur another auditor from Andersen experienced
Andersen’s ability to survive and led to a mass positive stock price reaction during the three-day
exodus of Andersen’s clients within a short period window surrounding the announcement of auditor
of time. Arthur Andersen was found guilty of change. We attribute this positive response to the
obstruction of justice which ultimately led to the reduction in uncertainty associated with the cost of
firm’s voluntary decision to cease its practice by finding a new auditor.
August 31, 2002. In this paper we investigate the
role of corporate governance on the timing of the
INTRODUCTION
Correspondence to: Jagan Krishnan, The Fox School of Business
and Management, Temple University, Alter Hall, 1801 Liacouras The financial collapse of Enron Corporation during
Walk, Philadelphia, PA 19122, USA. Email: krish@temple.edu the last quarter of 2001 drew extensive attention
ISSN 1090-6738
© 2010 Blackwell Publishing Ltd
Corporate Governance, Audit Firm Reputation, Auditor Switches, and Client Stock Price Reactions: The Andersen Experience 275
to its auditor, Arthur Andersen, from the media, suggest that clients with good governance are more
regulators, and the accounting profession. likely to switch earlier. In addition, we find that
Disclosures about Andersen’s audit of Enron switching costs are positively associated with the
caused a decline in its reputation and as a result, time taken to find a new auditor.
Andersen clients suffered significant loss of market We also examine the valuation effects
value (Chaney & Philipich, 2002). The first material surrounding the auditor changes. Prior research has
disclosure, that Andersen had shredded argued that auditor reputation adds credibility to
Enron-related documents, was made on January the client’s financial statements (Simunic & Stein,
10, 2002. The second disclosure, on February 2, 1987; Francis & Wilson, 1988), and is therefore
2002, was the release of the Powers Report, which impounded in the client’s stock price (e.g., Beatty,
detailed extensive problems in the audit of Enron 1989; Baber et al., 1995). Studies have shown
and revealed that the national office of Andersen (Chaney & Philipich, 2002; Callen & Morel, 2003;
was aware of these problems. From February 2002 Doogar et al., 2003; Krishnamurthy et al., 2006), that
to June 2002, Andersen faced a civil investigation Andersen clients lost significant market value
by the Securities and Exchange Commission (SEC), during the periods of key negative disclosures
several lawsuits, and was indicted and tried by the about Andersen. However, a more recent study
Justice Department for, and ultimately found guilty by Nelson et al. (2008) suggests that the
of, obstruction of justice. Following Andersen’s contemporaneous macroeconomic news and the
indictment in March 2002, its clients began peculiar industry composition of Arthur Andersen
changing auditors en masse. Andersen was clients were responsible for these negative stock
convicted on June 15, 2002 and discontinued its returns and not the erosion of Andersen’s
operations soon after, forcing its remaining clients reputation. Thus, the impact on the market value of
to seek a new audit firm.1 Unlike a voluntary Andersen’s clients was unclear. However, it was a
auditor change where the change is motivated by foregone conclusion that the auditor had suffered
client cost–benefit considerations, for Andersen irreparable reputational damage. By the spring of
clients the change was involuntary and driven by an 2002 this led to uncertainty regarding Andersen’s
exogenous shock (Barton, 2005). ability to continue in business, a situation that
In this paper we examine the influence of created uncertainty amongst its clients about
corporate governance on the timing and the firm finding a new auditor. To investigate whether the
valuation effects of the switch away from Andersen. market reacted positively to the reduction in such
Previous studies have examined the timing from uncertainties, we examine the stock price reaction to
different perspectives. Barton (2005) examines the 635 auditor change announcements made by former
auditor switching decision from the perspective of Andersen clients.3 We find the market reaction to
the capital market, looking at factors such as analyst the change announcement was positive on average.
following, press coverage, and institutional Our study provides additional insights on the
ownership. Blouin et al. (2007) focus specifically on events surrounding the collapse of Enron and its
the distinction in the switching behavior of clients auditor Arthur Andersen. Further, the results of our
that followed former Andersen partners and those study are relevant to the debate on mandatory
that did not. Bewley et al. (2008) examine the auditor rotation, which the Sarbanes-Oxley Act
switching decision from a signaling perspective, required the Government Accountability Office
whereas Chen and Zhou (2007) focus on the role of (GAO) to examine. That is, the forced auditor
the audit committee in the decision to switch.2 In change that some clients of Andersen faced captures
this study, we focus on the role of corporate some elements of mandatory rotation (Nagy, 2005).
governance in the timing of switch. We conjecture
that, ceteris paribus, clients with strong governance
MOTIVATION
would be the first to change auditors because of
their desire to disassociate from an auditor with a DeAngelo (1981) describes the auditor–client
tainted reputation and possibly their ability to relationship as one of bilateral monopoly because
attract a new auditor. In multivariate analyses, we the net benefit to the auditor and the client exceeds
control for switching costs and previous share price what could be obtained from an alternative
losses suffered by former clients of Andersen at the relationship. The client would incur switching
time of disclosures pertaining to Andersen’s audit of costs and start-up costs in the event of a switch to a
Enron. After controlling for these factors, our results new auditor. There could also be agency costs,
© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

276 S. C. Asthana, S. Balsam and J. Krishnan
insurance costs, and costs associated with reasons. First, firms are switching away from an
reputation loss and adverse stock price reaction auditor with a damaged reputation and, second, the
due to switching.4 For the auditor, the loss of future switch announcement resolves any uncertainty
revenues and prestige associated with clients are regarding the firm’s ability to find a new auditor.
factors that provide incentives to continue the Therefore, we test the following hypothesis:
relationship. The events surrounding disclosures
H2: Andersen clients that publicly disclosed they
about Enron and its auditor Andersen created an
were changing auditors experienced positive
imbalance to the bilateral monopoly because of
stock price reactions around the announcement
major disincentives for the clients of Andersen to
date.
continue the relationship. During the period from
February to June 2002, when Andersen faced civil
investigation by the SEC, several lawsuits, and was DATA AND RESULTS
indicted and tried by the Justice Department for, Sample selection
and ultimately found guilty of, obstruction of
justice, disclosures began to appear in the media We started with 9,861 firms on the 2001 Annual
about Andersen clients changing auditors.5 Compustat Tape (PST (primary, supplementary
The circumstances surrounding Andersen’s and tertiary) and Full Coverage). We obtained
demise are unique in that within a short period of financial data and auditor identity from
time all the firm’s clients had to select a new Compustat and stock returns from the Center for
auditor (Blouin et al., 2007; Chen & Zhou, 2007). Research in Security Prices (CRSP). We collected
Several studies have documented a positive by hand corporate governance and employee
association between corporate governance quality affiliation information (described later) from
and financial reporting quality. Corporate proxy and 10-K filings. We obtained audit and
governance has been shown to be negatively non-audit fee data from Standard & Poor’s and
associated with earnings management (Klein, 2002; proxy filings. As described in Table 1 (Panel A),
Xie et al., 2003; Bédard et al., 2004). Other studies we lost 2,657 observations for which the name of
document the role of good corporate governance the auditor was not available. Of the remaining
in auditor choice and in management support of 7,204 observations, only the 1,181 observations
auditors’ decisions. For example, companies with pertaining to Andersen clients were retained.
strong corporate governance tend to hire better Missing CRSP data and incomplete information
quality auditors (Abbott & Parker, 2000), are more further reduced the sample by 208 and 36
likely to support the auditor in a going concern observations, respectively. The final sample
opinion decision (Carcello & Neal, 2000), and are consists of 937 Andersen clients.6
less likely to dismiss an auditor following the Panel B provides information on the timing of
going concern opinion (Carcello & Neal, 2003). the auditor switch during the period February
We also expect that clients with strong corporate 6–August 31, 2002, which we manually obtained
governance would be quicker to disassociate from 8-K filings at the SEC website. For clarity, we
themselves from an auditor with a tainted break up the window by easily identifiable events
reputation. In addition, we expect that firms with beginning with the release of the Powers Report on
good governance will be more desirable/less risky February 2, 2002. The Powers Report was critical
to the new auditor and hence will be able to switch of Andersen with respect to three items: (1)
more quickly. Thus, our hypothesis is: Andersen’s Chicago Headquarters were aware of
accounting problems at Enron, (2) Andersen did
H1: Andersen clients with strong corporate
not effectively monitor the disclosures made by
governance were more likely to switch early.
Enron, and (3) Andersen did not report any
We next investigate the stock price reaction to accounting-related concerns to Enron’s board of
auditor change announcements. In general, prior directors (Chaney & Philipich, 2002). From the date
research has not found a significant abnormal of the Powers Report to the date of Andersen’s
return for auditor change announcements indictment (March 14, 2002) only 15 of Andersen’s
(Schwartz & Soo, 1996; Nichols & Smith, 1983), 937 clients switched audit firms. The relatively few
except for auditor resignations (e.g., Shu, 2000). changes may reflect the fact that clients had not
However, we expect that investors would react yet realized that Andersen would be unable to
positively to the move away from Andersen for two continue in existence, and/or the fact that they

Table 1: Sample selection and distribution

Panel A: Sample selection
Procedure Number of Number of
firms lost firms left
Firms available on 2001 Annual COMPUSTAT Tape (PST and Full Coverage) 9,861
Auditor’s name not available on COMPUSTAT (2,657) 7,204
Firm not an Andersen client (6,023) 1,181
Data not available on 2002 CRSP Tape (208) 973
Variables needed for analysis not estimable (36) 937
Panel B: Distribution of switching firms by delay
Date Event Days after Firms switching Firms still
2/5/2002† from Andersen after with Andersen
previous event till
current event
2/2/2002† Powers Report is released 0 days n/a 937
3/14/2002 Federal Govt. indicts Andersen 37 days 15 922
5/6/2002 Andersen’s criminal trial begins 90 days 261 661
6/15/2002 Andersen found guilty 130 days 359 302
8/31/2002 Andersen ceases practicing 207 days 302 0
†
Since the Powers report was released on 2/2/2002 (Saturday), the event date for our abnormal return analysis
is 2/4/2002 and our (-1, 1) event window ends on 2/5/2002. Consequently we begin our analysis of switches
on 2/6/2002. Note that the first switch we record is on 2/12/2002.
Panel C: Distribution of Switching Firms across Industries
Industry Clients Percentage Compustat
percentage
1. Agriculture, Forestry, and Fishing 5 0.53 0.36
2. Mining 48 5.12 4.12
3. Construction 10 1.07 0.98
4. Manufacturing 326 34.79 36.31
5. Transportation and Utilities 93 9.93 10.02
6. Wholesale 30 3.20 3.11
7. Retail 50 5.34 5.26
8. Financial Services 133 14.19 18.26
9. Services 172 18.36 20.50
10. Others 70 7.47 1.08
Total 937 100.00 100.00
The industry classification is based on Dopuch et al. (1987), and includes the following SIC codes:
Agriculture, Forestry, and Fishing 100–999
Mining 1000–1499
Construction 1500–1999
Manufacturing 2000–3999
Transportation and Utilities 4000–4999
Wholesale 5000–5199
Retail 5200–5999
Financial Services 6000–6999
Services 7000–8999
Others <100 and >8999
To be comparable with the sample, the Compustat distribution is based on fiscal year end data for 2001.

Table 2: Cumulative abnormal returns

Date Event Andersen Mean CMAR Mean CCAR
clients (median) (median)
1/10/2002‡ Andersen admits to shredding thousands 937 **-0.0036 ***-0.0046
of Enron-related documents. The (***-0.0031) (**-0.0021)
Department of Justice announces a
criminal investigation of Andersen.
2/2/2002†‡ Powers Report, suggesting Andersen’s 937 *-0.0020 ***-0.0067
headquarters office in Chicago was (***-0.0012) (*-0.0038)
aware of problems with Houston
office and Enron audit, is released
3/14/2002‡ Federal Government indicts Andersen 922 ***-0.0073 *-0.0021
(***-0.0072) (*-0.0022)
5/6/2002‡ Andersen’s criminal trial begins 661 ***-0.0148 ***-0.0054
(**-0.0164) (*-0.0043)
6/15/2002‡ Andersen found guilty 302 ***-0.0099 *-0.0023
(**-0.0064) (-0.0017)
Oct 2001 to Oct 2002 Thirteen-month window 937 ***-0.0168 *-0.0519
(*-0.0013) (*-0.0513)
CMAR = Cumulative market-adjusted returns of clients for the window (-1, 1) around the respective event date.
Equally weighted market index is used. Value-weighted gives similar results.
CCAR = Cumulative four-factor-adjusted Carhart returns of clients for the window (-1, 1) around the respective
event date.
†
Since the Powers Report was released on 2/2/2002 (Saturday), the event date is 2/4/2002; ‡return for (-1, 1)
window; *significant at 10% level; **significant at 5% level; ***significant at 1% level.
were unwilling to drop their auditor in the midst also examine cumulative abnormal returns for the
of their year-end audit.7 Switches accelerated 13-month window from October 1, 2001 to October
dramatically after the indictment when it became 31, 2002.8 Cumulative abnormal returns are
clear that Andersen would be unable to continue computed two ways, cumulative market-adjusted
and finding a new auditor was inevitable. A total of returns (CMAR) and cumulative four-factor returns
261 clients switched between the indictment on (Carhart, 1997) (CCAR) during the three-day
March 14 and May 6, the date the trial started, while window (-1, 1) surrounding the event-dates.9
more than half of Andersen’s remaining clients The magnitude of the abnormal return in the
(359) announced their switch during the trial. Of short windows is negative and is significantly
the 937 changes, slightly more than two-thirds different from zero in all five short windows. We
(635) occurred prior to Andersen’s conviction on see that the return to Andersen clients in the
June 15, 2002. Almost immediately following the period surrounding the shredding disclosure was
verdict, Andersen announced that it would no significantly negative, -0.46% for our sample of 937
longer audit public companies (Eichenwald, 2002). firms.10 The return associated with the disclosure of
While the vast majority of these pre-conviction the Powers Report is significantly negative, -0.67%,
period changes were auditor dismissals, there were while that associated with Andersen’s indictment,
13 auditor resignations. Panel C of Table 1 presents -0.21%.11 Using a reduced sample of clients that had
the industry distribution of the sample and the not yet announced an auditor change, we find
Compustat population. With minor variations, the additional significant negative abnormal returns of
sample is representative of the population of firms. -0.54% around the start of Andersen’s criminal trial,
In Table 2, we present the market reaction to and -0.23% around Andersen’s conviction date,
several negative disclosures about Andersen. We both statistically different from zero.12 Examining
estimate losses in value for five dates by computing the full 13-month window, which encompasses the
the mean cumulative abnormal returns during the first doubts about the assurance value of Andersen’s
three day window (-1, 1) surrounding the date. We audit, Andersen’s downfall, and the switch from

Andersen, we find a significant negative abnormal model in which the dependent variable is coded 0
return of -5.19%. Researchers disagree on the causes if the client switches from Andersen between
of such negative stock returns to former Andersen February 5, 2002 and June 15, 2002 (i.e., the
clients. For example, Chaney and Philipich (2002), pre-conviction period), and 1 if it switches after
Callen and Morel (2003), Doogar et al. (2003) and June 15, 2002 (the post-conviction period). We also
Krishnamurthy et al. (2006) interpret the negative estimate a trichotomous model, in which the
stock returns as clients’ adverse reactions to the dependent variable takes the value of 1 if the client
decline in Andersen’s reputation. On the other switches from Andersen from February 5, 2002 to
hand, Nelson et al. (2008) suggest that May 6, 2002 (the pre-trial period); the value of 2 if it
contemporaneous macroeconomic news and the switches from May 7, 2002 to June 15, 2002 (the trial
specific industry composition of Arthur Andersen period); and the value of 3 if it switches after June
clients were responsible for these negative stock 15, 2002 (the post-conviction period).
returns.
Governance
Timing of switch We examine a series of variables suggested by prior
research to proxy for good governance which may
The failure of Andersen is an important and unique
influence the timing of the switch from Andersen.
event in the history of the auditing profession that
Prior studies (e.g., Abbott & Parker, 2000) have
left over 1,500 clients of Andersen looking for
documented that companies with good governance
another auditor within a short period of time
are likely to hire a high-quality auditor. An
(Krishnan et al., 2007). Unlike a typical auditor
effective board and audit committee is likely to play
change situation where the choice of the auditor
a pivotal role in ensuring adequate oversight and
(auditor dismissals) is determined by the client
high-quality financial reporting (Chen & Zhou,
and its management or by its auditor (auditor
2007). We use several proxies for corporate
resignations), clients of Andersen were forced to
governance based on prior work (Raghunandan
seek a new auditor by the circumstances
et al., 2001; Raghunandan & Rama, 2003; Xie et al.,
surrounding Enron’s collapse and the imminent
2003; Bédard et al., 2004): SEPARATECHAIR, BDIR,
demise of Andersen. This provides us with a
ACDIR, BMEET, ACMEET, and EXPERT.
unique opportunity to study the timing of
SEPARATECHAIR is a dummy variable that
switching since the incentives to switch are similar
takes the value of 1 if the position of CEO and
across firms and the decision is influenced
Chairman of the Board are separated, and 0
primarily by the cost of switching as opposed to
otherwise; Raghunandan and Rama (2003) provide
most other switching situations in which both the
evidence that shareholders are likely to vote
costs and benefits of switching are important. Our
against an auditor ratification proposal when the
primary focus is on whether differences in
CEO also acts as the chairperson of the board (CEO
governance characteristics among clients of
duality). Conversely, shareholders support auditor
Andersen influenced the timing of the switch. To
choice made by companies that have different
test our hypothesis, we use ordinary least squares
individuals performing the roles of the CEO and
(OLS) regressions, and ordinal and multinomial
Chairman of the Board. Dechow et al. (1996)
logit regressions. We present the OLS model below:
provide evidence consistent with greater earnings
DELAY = β0 + β1SEPARATECHAIR + β2 BDIR management by firms with CEO duality.
+ β3 ACDIR + β4 BMEET + β5 ACMEET Prospective auditors could thus view the
+ β6 EXPERT + β7 LMV + β8 INDSPL separation of the positions as a good sign. We
+ β9 NONAUDIT + β10 DTENURE (1) expect the auditor switch to be quicker when the
+ β11 ADACCR + β12 DISTRESS positions are separated.
+ β13 PREVCAR1 + β14 PREVCAR 2 We include board (BDIR) and audit committee
+ β15 DFYR + β16 ROA + β17 DNEWB4 + ε size (ACDIR) to capture other facets of governance.
It has been suggested, and empirical evidence has
where DELAY is the number of days after February shown, that large boards are less effective (e.g.,
6, 2002 the client announced its switch from Jensen, 1993; Yermack, 1996). To some extent, this
Andersen. The logistic analyses differ in the was refuted by Coles et al. (2008) who argued that
dependent variable. We present a dichotomous the relationship between board size and firm value

is more nuanced and depends upon the with as much industry expertise. Consequently,
characteristics of the firm. Thus, we do not predict firms will delay switching the greater Andersen’s
a sign for this variable. By contrast, the literature on expertise in their industry. NONAUDIT (following
audit committee size provides support for a direct Frankel et al., 2002; Ashbaugh et al., 2003; Reynolds
relationship between size and better governance et al., 2004) is the ratio of non-audit fees to total fees
(e.g., DeFond et al., 2005). We predict a negative paid to Andersen.15 Our expectation is that the more
sign (i.e., less delay) for this variable. Board and non-audit services provided by Andersen, which
audit committees that do not meet frequently may by definition are less of a commodity than audit
not be able to carry out their tasks effectively (Yang services, the greater the switching costs because
& Krishnan, 2005). To capture their effectiveness, the client’s choice decision will be influenced by
we include the number of board (BMEET) and the ability of the new auditor to provide similar
audit committee (ACMEET) meetings. We expect services.16 DTENURE (following Myers et al., 2003;
negative signs for these two variables. We include Davis et al., 2009) is an indicator variable taking the
EXPERT, the number of financial experts on the value of 1 if Andersen had been the auditor for five
audit committee to capture the financial expertise or more years and 0 otherwise.17 DTENURE could
of the audit committee. Recent studies provide proxy for Andersen’s perceived independence and
evidence in support of higher quality earnings in quality of the client’s financial reporting. If the
firms where audit committees possess accounting prospective new auditor views a longer tenure
expertise (e.g., Krishnan & Visvanathan, 2008). We with Andersen as reflecting a lower quality of
posit that financial experts on the audit committee reporting, we would expect DTENURE to be
are more likely to influence an expedited switch to positively associated with the time taken to switch
a new auditor.13 Additionally, firms with greater auditors.18 ADACCR is the absolute value of 2001
financial expertise on the audit committee may be discretionary accruals deflated by total assets,
more attractive to auditors as a client either because calculated using the cross-sectional version of the
they are likely to have higher earnings quality Jones (1991) model as in DeFond and Jiambalvo
and/or are easier for external auditors to work (1994) and using the difference between net income
with. As a result, firms with greater financial and cash from operations as our measure of total
expertise on the audit committee may find a new accruals (Hribar & Collins, 2002).19 Lastly,
auditor more quickly. Thus, we expect a negative DISTRESS is an indicator variable taking the value
sign for this variable. of 1 if the Altman Z score is less than 1.81 and 0
otherwise. Our expectation is that firms with
Proxies for switching costs higher discretionary accruals (indicating lower
earnings quality) and a greater probability of
We use several variables to proxy for switching and bankruptcy will be less desirable as clients and
start-up costs incurred by Andersen clients. LMV is thus will have higher costs of finding a new
the log of the market value of equity. Larger firms auditor.20
may have to incur greater costs of finding a new
auditor because of limited choice (i.e., they Control variables
probably need to stay with the Big 4), and greater
start-up costs due to greater complexity of PREVCAR1 and PREVCAR2, are the cumulative
operations. However, larger clients may be viewed four-factor-adjusted abnormal returns from the
as prestigious and may be aggressively sought after Carhart (1997) model for the (-1, 1) windows
and thus the cost of finding a Big 4 auditor may be around Andersen’s admission of shredding of
lower. Consequently we do not make a prediction documents on January 10, 2002 and the release of
for this variable. We include INDSPL, which is a the Powers Report on February 2, 2002,
measure of Andersen’s industry specialization. respectively. We use these events as prior literature
INDSPL is Andersen’s industry market share, (e.g., Chaney & Philipich, 2002) shows a significant
based on the square root of clients’ total assets market reaction by Andersen clients during these
(Ettredge & Greenberg, 1990; Balsam et al., 2003), in windows. Given the conflicting interpretations of
the client’s two digit SIC code.14 We expect that the the negative returns around these dates by
cost of finding a new auditor will be higher if researchers (discussed earlier), we do not have any
Andersen had a large market share, because of predictions for these variables. We include them in
the difficulty involved in identifying another firm the model as control variables to pick up the effects

of any agency costs that may not be accounted for clients could decide when they wanted to switch.
by the independent variables included in the By contrast, clients of Andersen had no choice but
models.21 to switch after June 15. Thus we subdivide our
We also include the following control variables analysis into two periods, and employ a
that might influence the time taken to switch multinomial logit model to study the switching
auditors: ROA, DNEWB4, and DFYR. We include pattern. Another difference between our study and
ROA (return on assets) as firms with better that of Chen and Zhou (2007) is their sample period
performance are likely to find it easier to find a begins earlier than ours, i.e., they look at switches
successor auditor. Hence we predict a negative starting on October 15, 2001. For example, their
association between delay and ROA. We include dependent variable measures the number of days
DNEWB4 (coded as 1 if the new auditor is a Big 4 between October 15, 2001 and the date of
auditor, 0 otherwise) because the timing of the switching, where October 15, 2001 is the day before
change may depend on the kind of successor Enron announced its unexpected $1.01 billion
auditor the client is looking for (Schwartz & Soo, charge to earnings. While this date was obviously
1996). In particular, if Big 4 auditors are more crucial in the Enron saga, we do not believe it was
conservative in their acceptance of clients in the a critical date for Andersen and expect that any
wake of Andersen’s problems, it may take longer switches between October 15 and early February to
for the client to obtain a new auditor. Finally we be caused by normal (non-Enron) factors. In terms
include DFYR, a dummy variable that takes the of empirics, the main impact is the inclusion of
value of 1 if the client’s fiscal year end is December, these switches in their sample, which we feel adds
and 0 otherwise to control for the possibility that noise to their analysis. Chen and Zhou (2007) do
clients are less likely to make an auditor change in not disclose how many switches occur within this
the midst of their year-end audit. Consequently window, but we note the first quartile (in their
these clients would be unable to change during the study) is 199 days, indicating three quarters of their
months of February and March, and hence would switches occurred after May 2, 2002. Consequently
be more likely to be late switchers.22 it is unlikely many switches occurred prior to the
start of our sample period February 6, 2002. A
Prior related study (Chen & Zhou, 2007) secondary impact would be relative to our results,
an increased intercept, as for a given switch their
Chen and Zhou (2007) also study the association dependent variable will be 114 higher than ours
between the timing of the switch and audit (difference in days between October 15, 2001 and
committee characteristics. Although they find that February 6, 2002).
independence and expertise of audit committees
and board size are positively associated with Descriptive statistics
dismissal of Andersen, their results are sensitive to
the inclusion of non-December fiscal year-end Table 3 provides descriptive statistics for the
companies, i.e., they report that the audit variables in our model. In general, early switchers
committee independence and audit committee appear to have better governance, as the number of
financial expertise are not significant for December directors (BDIR), and number of audit committee
year-end companies. They interpret their evidence members (ACDIR), as well as number of meetings
as mixed (p. 1087). Given the mixed results for both groups (BMEET and ACMEET) are greater
observed in Chen and Zhou and the increased for early adopters. We also find early switchers
emphasis placed by the Sarbanes-Oxley Act on the are more likely to have a separate board chair.
role of audit committee in the appointment of However, we do not find a difference for EXPERT.
auditors, we revisit this issue in our study. Turning to our switching cost variables we find that
The major difference between their study and firms that switch early are larger, belong to
ours is that we make a distinction between industries where Andersen had a lower market
pre-conviction period and post-conviction period share, obtained fewer non-audit services from
switching. Thus we characterize the time period Andersen, and were less likely to have been with
from February 6 to June 15, 2002 (which is the date Andersen for five years. Overall, it appears that
that Andersen announced that it would stop early switchers have lower switching costs. We do
offering audit and other services), as the not observe, at least in the univariate sense, any
‘pre-conviction’ period, since prior to June 15 difference between early and late switchers for

Table 3: Firm characteristics by timing of switch

Variables Mean value t-statistics H0:
(early = late)
Early switchers Late switchers
(PERIOD21 = 0) (PERIOD21 = 1)
DELAY 86.0920 150.8700 ***-32.94
ABNDELAY -16.9920 35.7282 ***-30.17
Governance variables
SEPARATECHAIR 0.6412 0.5805 **1.67
BDIR 8.1949 6.7458 ***6.91
ACDIR 3.4873 3.1059 ***4.67
BMEET 6.9040 5.8771 ***3.65
ACMEET 4.2175 3.5636 ***4.20
EXPERT 1.2885 1.2034 1.05
Variables proxying for switching costs
LMV 5.5246 4.2257 ***7.36
INDSPL 0.1174 0.1401 ***-2.74
NONAUDIT 0.3687 0.4492 ***-4.16
DTENURE 0.6102 0.6568 *-1.30
ADACCR 0.0956 0.1074 -0.78
DISTRESS 0.2698 0.2797 -0.30
Control variables
PREVCAR1 -0.0054 -0.0030 -1.09
PREVCAR2 -0.0084 -0.0032 *-1.46
DFYR 0.6610 0.6992 *-1.28
ROA -0.0650 -0.1130 *1.70
DNEWB4 0.7331 0.7853 *-1.66
*significant at 10% level; **significant at 5% level; ***significant at 1% level. (one-tailed significance levels where signs are
predicted, two-tailed otherwise).
1
Dichotomous variable with value of 0 if client switches from Andersen between 2/5/2002 and 6/15/2002; 1 if it switches after
6/15/2002.
Variables Definition
DELAY Number of days after February 6, 2002 the client switched from Andersen.
ABNDELAY The abnormal delay from equation 1 (Table 4).
SEPARATECHAIR Dummy variable with the value of 1 if positions of CEO and Chairman of the Board are separated,
and 0 otherwise.
BDIR Number of Directors on the Board.
ACDIR Number of Directors on the Audit Committee.
BMEET Number of Board meetings.
ACMEET Number of Audit Committee meetings.
EXPERT Number of Audit Committee members that are financial experts.
LMV Natural logarithm of market value of equity (in millions) at the end of fiscal year 2001.
INDSPL Measure of industry specialization of the auditor; measured as the auditor’s market share (based on
the square root of clients’ total assets) during fiscal year 2001 in the 2-digit SIC code.
NONAUDIT Proportion of non-audit fees (includes tax, system, and other services) to total fees paid to the
auditor.
DTENURE Dichotomous variable with value of 1 if total years with Andersen as auditor are five or more, and 0
otherwise.
ADACCR Absolute value of discretionary accruals deflated by total assets during fiscal year 2001, calculated
using the cross-sectional version of the Jones (1991) model as in DeFond and Jiambalvo (1994) and
we use the difference between net income and cash from operations as our measure of total
accruals (Hribar & Collins, 2002).
DISTRESS Dummy variable that has a value of 1 if Altman’s Z score (for fiscal year 2001) is less than 1.81, and
0 otherwise.
PREVCAR1 Cumulative four-factor-adjusted Carhart returns in (-1, 1) windows around Andersen’s admission of
shredding of documents on 1/10/2002.
PREVCAR2 Cumulative four-factor-adjusted Carhart returns in (-1, 1) windows around the release of Powers
Report on 2/2/2002.
DFYR Dummy variable with the value of 1 if fiscal year end is December, and 0 otherwise.
ROA Return on assets during fiscal year 2001.
DNEWB4 Dummy variable with the value of 1 if the new auditor is one of the Big 4, and 0 otherwise.

either the quality of earnings or the probability of (SEPARATECHAIR) are negatively associated with
financial distress. While we find early switchers delay (as predicted) in the OLS and ordered logits.
had suffered greater losses due to the disclosures The number of financial experts on the audit
pertaining to Andersen, those differences are only committee (EXPERT) is significantly negative in
marginally significant (p < 0.10) and only for the OLS and two-level logit, while the number of
second of the windows. We also find some directors on the audit committee is significantly
differences in our control variables between early negative in the three-level ordered logit. In the Cox
and late switchers. We find that early switchers model, all the governance variables (except ACDIR)
have a better, albeit less negative, ROA, are less have a positive effect on the probability of switch.26
likely to have a Big 4 auditor, and are less likely to Overall the evidence suggests firms with better
have December fiscal year ends. governance were likely to switch earlier.
Turning to our switching cost variables we find
Multivariate analysis of timing of switch that firm size, LMV, is negative and strongly
significant in the OLS and ordered logits. In the
Table 4 provides the regression results for Cox model, the probability of a switch from
equation (1).23 The first regression in the table uses Andersen at any time is higher for larger firms.
the continuous variable DELAY as the dependent INDSPL is positive and significant in all models
variable. The second regression uses a Cox (1972) (except Cox) indicating that firms in industries
proportional hazard model (a similar model is also where Andersen had a high market share were
reported in Barton, 2005).24 In addition to the more likely to delay switching. In contrast,
estimation of the Cox model, we also report the NONAUDIT is only significant in the OLS and Cox
percentage change in the hazard rate.25 Since models. DTENURE is positive and significant in
different independent variables may have different the OLS and ordinal logit models, providing some
scales, the percentage change in the hazard rate is evidence that the delay is related to Andersen’s
more comparable. This is the change in the tenure as auditor.27 ADACCR and DISTRESS are
probability of a switch due to one standard both positive and significant in the OLS and
deviation change in the variable, holding all other ordered logits and negative in the Cox model,
variables constant. We also use two discrete consistent with firms that have lower earnings
specifications. In the third regression, the quality and firms in financial distress taking longer
dependent variable is PERIOD2, a dummy variable to switch. Overall the results suggest that switching
that takes the value of 0 if the switch occurs prior to costs are associated with the timing of the switch.
Andersen’s conviction on June 15, 2002, and 1 Among the control variables, PREVCAR1 is
otherwise. In the fourth regression the dependent insignificant in all the regressions. This appears to
variable PERIOD3 takes the value of 1 if the switch be consistent with Nelson et al.’s (2008) conclusion
occurs prior to the start of the trial on May 6, 2002, that the market reaction around January 10, 2002
2 if the switch occurs after the start of the trial but shredding announcements was driven by
prior to the conviction on June 15, and 3 if the microeconomic as well as industry specific news
switch occurs after the conviction. Thus the third and not by decline in Andersen’s reputation. On the
and fourth regressions are estimated as binomial other hand, PREVCAR2 is significantly positive
logit and ordered logit respectively. We selected in the OLS and ordered logits (and negative in the
these two cut-offs because the beginning of the Cox model). Given the recent finding by Nelson
trial significantly diminished the possibility of a et al. (2008) that the adverse market reaction of
settlement with the Justice Department (Beltran, Andersen’s clients around the February 2, 2002
2002) and the conviction of Andersen on June 15 Powers Report date could be due to questions of the
marked the end of Andersen’s auditing operations. integrity of financial reporting (driven by anxiety
In general, the tenor of the results is consistent about earnings news along with concerns about
across the models. There is consistent evidence that corporate scandals), PREVCAR2 might be a proxy
the quality of corporate governance is associated for the firm’s sensitivity to financial reporting
with the timing of auditor change. Among the uncertainty at this point of time. In other words,
variables measuring the quality of corporate the more sensitive a firm is to financial reporting
governance, the number of board members (BDIR), uncertainty, the greater the reaction around
the number of board and audit committee meetings February 2, 2002 and the more negative the value
(BMEET and ACMEET), and CEO separate chair of PREVCAR2. Thus, a positive coefficient on

284
Table 4: Timing of auditor change by Andersen clients
Variables Exp. Type of procedure
sign
OLS COX PROPORTIONAL ORDINAL ORDINAL MULTINOMIAL
HAZARD MODEL†† LOGIT LOGIT LOGIT†
Dependent variable
1
DELAY Probability % Change PERIOD22 PERIOD33 PERIOD33
of defection in hazard
Log(p1/p3) Log(p2/p3)
Intercept2 ***1.3179
© 2010 Blackwell Publishing Ltd

Intercept1 ***146.524 ***2.0123 ***3.931 ***-2.5217 ***-1.4957
SEPARATECHAIR - *-2.2507 *0.0817 8.5130 *-0.2725 *-0.2011 0.1178 0.0700
BDIR +/- ***-2.3483 ***0.1394 49.1858 ***-0.1787 ***-0.142 ***0.1761 **0.1135
ACDIR - -0.9932 0.0544 6.2840 -0.0728 *-0.1263 -0.0419 -0.1086
BMEET - *-0.4160 *0.0226 9.0452 **-0.0656 *-0.0277 0.0497 **0.0695
ACMEET - ***-2.1980 ***0.1449 36.0700 **-0.0997 ***-0.1298 *0.0896 0.0054
EXPERT - *-1.0785 **0.1171 12.1523 *-0.1644 -0.0767 0.0193 0.0932
LMV +/- ***-5.9285 ***0.4018 163.0497 ***-0.3660 ***-0.3568 ***0.4086 ***0.1903
INDSPL + **16.5312 0.0714 0.7835 **2.2453 *1.1209 *1.7577 **1.9035
NONAUDIT + **6.5654 **-0.4437 -10.8403 0.1502 0.1440 -0.1605 0.2285
DTENURE + ***6.3257 0.0025 0.2503 *0.0757 ***0.3952 -0.2842 0.1855
ADACCR + **11.1533 **-0.5938 -10.7486 **1.3353 *0.6108 0.3113 0.6943
DISTRESS + **5.0029 ***-0.3790 -31.5454 ***0.5588 **0.3063 -0.2607 -0.3021
Control variables
PREVCAR1 +/- 13.0165 -1.2141 -7.9801 1.7499 0.6984 0.4029 1.1322
PREVCAR2 +/- ***33.4788 ***-3.4346 -20.8284 ***3.4866 ***2.7258 -1.8637 -1.7401
DFYR + ***12.1664 ***-0.7490 -52.7161 ***0.9252 ***0.8654 ***0.9521 ***0.5555
ROA +/- **-4.7629 0.1700 6.4432 *-0.4252 -0.2134 -0.0216 0.3624
DNEWB4 +/- ***13.9315 ***-0.9994 -63.1900 ***0.9772 ***0.8643 -0.6926 -0.0656
Observations 937 937 937 937 937
Adj R2 0.3338
Pseudo R2 0.3239 0.2787 0.2964
F-value 24.87
Probability > F <0.0001
2
Wald c Value 301.8387 116.3900 199.3358 1837.65
*significant at 10% level; **significant at 5% level; ***significant at 1% level (one-tailed significance levels where signs are predicted, two-tailed otherwise).
†
We do not have any predictions for signs in this analysis.
††
The expected signs are opposite for this model.
1
DELAY is the number of days after February 6, 2002 the client switched from Andersen.
2
PERIOD2 is a dichotomous variable with value of 0 if client switches from Andersen between 2/5/2002 and 6/15/2002; 1 if it switches after 6/15/2002.
3
PERIOD3 is a trichotomous variable with value of 1 if client switches from Andersen between 2/5/2002 and 5/6/2002; value of 2 if it switches between 5/6/2002
and 6/15/2002; and value of 3 if it switches after 6/15/2002.
The Cox (1972) proportional hazard model is similar to that in Barton (2005). The dependent variable is the client’s probability of switching from Andersen at time
t, provided it has not switched earlier. For continuous variables the percentage change in hazard rate is estimated as 100[exp(Sj bj) – 1], where Sj is the sample standard
deviation of variable j and bj is the estimated coefficient. For dichotomous variables, this change is estimated as 100[exp(bj) – 1].
See Table 3 for definition of other variables.
Int. J. Audit. 14: 274–293 (2010)

S. C. Asthana, S. Balsam and J. Krishnan
PREVCAR2 in Table 4 suggests that the more BMEET, and ACMEET declines with time. Another
sensitive the firm is to reporting uncertainty, the interpretation can be that with passage of time
earlier it switches.28 Results also show that firms there were fewer choices to be made, and the role
with greater ROA switched earlier (except in the of board and audit committee declined. An
Cox model) and that those that required the services interesting result in the multinomial logit is the
of a Big 4 auditor (DNEWB4) tended to switch later. significance of ACMEET in the pre-conviction
Finally the coefficient on DFYR is positive and period and BMEET in the post-conviction period
significant, which is consistent with clients with suggesting that the entire board played an active
December fiscal year ends having to delay their role with respect to auditor change in the
search for a successor auditor/switch until the post-conviction period in contrast to the audit
completion of their year-end audit and hence being committee’s role in the pre-conviction period.
unable to switch during February and March.29
During the sample period, clients could be
leaving Andersen initially due to reputation loss Market reaction to switching
and later due to auditor failure risk. This could lead
We also examine whether our proxies for good
to changes in the coefficients over time. To test this
governance and switching costs are associated
notion, we run the CATMOD procedure in SAS
with the market reaction around the announcement
that directly fits the generalized logit model. The
of the switch from Andersen to another auditor.30
dependent variable is a trichotomous variable,
More specifically, we look at the cumulative market-
PERIOD3, that takes the value of 1 if the switch
adjusted abnormal return (CMAR) and the
occurs prior to the start of the trial on May 6, 2002,
cumulative four-factor Carhart (1997) abnormal
2 if the switch occurs after the start of the trial but
return (CCAR) for the three day window (-1, 1)
prior to the conviction on June 15, and 3 if the
surrounding the auditor change announcement.31
switch occurs after the conviction. Denoting the
To examine the factors that influence the return we
probability of switch in the three periods by p1, p2,
use the following cross-sectional regression:
and p3, the first column reports the logarithm of the
ratio of the probabilities p1 to p3 and the second CMAR CCAR = β0 + β1SEPARATECHAIR + β2 BDIR
column reports the logarithm of the ratio of the + β3 ACDIR + β4 BMEET
probabilities p2 to p3. Thus, in the first column of + β5 ACMEET + β6 EXPERT
the multinomial logit, an insignificant coefficient + β7 LMV + β8 INDSPL
implies that the relative effect of the variable on p1 + β9 NONAUDIT + β10 DTENURE
is not different from its effect on p3; a positive + β11 ADACCR + β12 DISTRESS
(negative) coefficient implies that the relative + β13 PREVCAR1 + β14 PREVCAR 2
effect of the variable on p1 is greater (less) than + β15 DFYR + β16 ROA + β17 DNEWB4
its effect on p3. Thus, insignificant coefficients + β18 ABNDELAY + ε (2)
on SEPARATECHAIR, ACDIR, EXPERT,
NONAUDIT, DTENURE, ADACCR, DISTRESS, where ABNDELAY is the abnormal delay,
and PREVCAR2 in the multinomial logit, even calculated as the actual less the expected delay
though the same coefficients are significant in other which is the fitted value determined from equation
regressions (OLS, Cox, or ordinal logit), implies 1, and all the other variables are as defined
that these variables continue to have the same above.32 We include variables that we previously
impact on the switching decisions in all three included in our analysis of delay because they
periods. On the other hand, positive coefficients capture the uncertainties and related costs
on LMV, INDSPL, BDIR, BMEET, ACMEET, and associated with switching. We also include
DFYR imply that these variables are more ABNDELAY as a control, since the market might
influential in the earlier periods. Thus, for example, react to any abnormal delay in the switching
the strong positive coefficient on size might process. We expect market response to be
indicate that the larger the firm, the more important negatively associated with our governance
it was to find a new auditor early. Similarly, firms variables, since market participants are likely to
with good governance decided to switch auditors expect firms with good governance to switch
earlier and only the firms with less effective auditors and thus be less surprised by the
governance were left in the market in later sample announcement. On the other hand, the market
period. As a result, the explanatory power of BDIR, response should be positively associated with

Table 5: Market reaction around disclosure of auditor change

Panel A: Univariate analysis
N CMAR CCAR
635 Mean = *0.0078 Mean = *0.0042
Median = *0.0018 Median = *0.0011
Panel B: Regression analysis
Variables Exp. sign Dependent variable
CMAR CCAR
Intercept 0.0087 0.0089
SEPARATECHAIR - 0.0050 0.0029
BDIR - *-0.0009 *-0.0012
ACDIR - 0.0023 0.0032
BMEET - *-0.0006 *-0.0005
ACMEET - 0.0008 0.0004
EXPERT - *-0.0002 *-0.0010
LMV +/- *-0.0015 *-0.0008
INDSPL + 0.0033 0.0027
NONAUDIT + -0.0089 -0.0077
DTENURE + 0.0001 *0.0005
ADACCR + *0.0209 *0.0124
DISTRESS + -0.0017 -0.0017
Control variables
PREVCAR1 +/ 0.0034 -0.0056
PREVCAR2 +/ ***-0.0639 ***-0.0738
DFYR +/- -0.0013 *-0.0060
ROA +/- ***-0.0277 ***-0.0234
DNEWB4 +/- -0.0035 -0.0043
ABNDELAY - ***-0.0002 *-0.0001
Observations 635 635
Adj R2 0.0909 0.0625
F-value 4.30 3.50
Prob > F <0.0001 <0.0001
ABNDELAY is the abnormal delay from equation 1 (Table 4). See Tables 2 and 3 for other variable definitions.
*significant at 10% level; **significant at 5% level; ***significant at 1% level (one-tailed significance levels where
signs are predicted, two-tailed otherwise).
switching costs since there was greater uncertainty Table 5, Panel A shows that the average abnormal
that such firms would find an auditor to switch to. return around the date of announcement by an
As a result, the market is more relieved when such Andersen client that it was dismissing its auditor
firms switch. However, even though LMV is a was positive and significant – the mean CMAR
measure of higher switching cost, it also captures (CCAR) was 0.78 (0.42)% for the three-day
several other effects and we do not predict its window.34,35 Thus, on average, there was a positive
sign.33 For reasons discussed earlier, we do not market reaction when investors found out that the
have any expectations on the signs of PREVCAR1 firm had replaced Andersen. This finding supports
or PREVCAR2. Finally, as time passed, the switch hypothesis 2.
from Andersen became more inevitable and less Panel B contains the regression results for
unexpected, with more and more clients deserting equation (2) using alternatively, cumulative market-
their former auditor. Thus, we expect a negative adjusted (CMAR), and cumulate four-factor
(less positive) sign on ABNDELAY. abnormal returns (CCAR) as the dependent

variable. Given the models are comparable (i.e., CONCLUSIONS

p-values < 0.0001, adjusted R2 9% and 6%,
respectively), our discussion focuses on the model After months of negative disclosures about Arthur
using the market adjusted return as the dependent Andersen’s practices during the audit of Enron in
variable. We find negative and significant 2001, Andersen’s clients began to switch auditors,
associations for three of our governance variables, voluntarily, with the majority doing so before June
board size (BDIR), the number of board meetings 15, 2002 when Andersen was convicted and its
(BMEET) and the number of financial experts on remaining clients were forced to change auditors.
the board (EXPERT). The results are consistent with Complementing prior research (Blouin et al., 2007;
the market expecting firms with better governance Chen & Zhou, 2007), we shed light on the role of
to fire Andersen and hire a new auditor, and corporate governance on the timing of that switch.
consequently the market reaction to the news of the We expect that clients with strong corporate
actual switch is smaller. Overall, we see a weak governance switch auditors early, while those with
association between abnormal returns and our weak governance take longer to find a new auditor.
proxies for switching costs. We see negative and Our findings reflect these expectations. We find
significant association for firm size (LMV) and a that the timing of the switch is related to corporate
positive and significant association with earnings governance measures; boards seem to have taken
quality (ADACCR), indicating that the switch was an active role in the period following Andersen’s
viewed more favorably for smaller firms and for conviction while audit committees seem to have
firms with lower earnings quality. The coefficient on played a role in the pre-conviction period. Other
PREVCAR1 is insignificantly different from zero, factors (e.g., client size, auditor industry
consistent with Nelson et al.’s (2008) argument specialization, provision of non-audit services,
that the market reaction around the shredding auditor tenure, quality of earnings and financial
disclosures by Andersen was macro-events related distress) hypothesized to be associated with the
and not due to Andersen’s damaged reputation. The switching costs are associated with the timing of
negative coefficient on PREVCAR2 is also consistent the switch.
with Nelson et al.’s (2008) conjecture that the market We also examine the market reaction to the
reaction around the Powers Report date may proxy announcement that the firm was replacing
for the firm’s sensitivity to financial reporting Andersen as its auditor. We expect that the change
uncertainty. Thus, the negative coefficient on would cause a positive reaction due to the
PREVCAR2 implies that the more sensitive a firm is resolution of uncertainties associated with
to reporting uncertainty, the greater the relief switching and its related cost. Consistent with our
around disclosure of audit switch. We also find expectations, we find that the mean cumulative
a negative and significant coefficient on ROA. abnormal return during a three-day window
Finally, the coefficient on ABNDELAY is negative surrounding the date of announcement of auditor
suggesting that the market reacted less positively to change was positive and statistically significant.
the news of the switch with each passing day, as the The results of our study are relevant to debates
collapse of Andersen became more inevitable and about mandatory auditor rotation. The
the switch less unexpected.36 Sarbanes-Oxley Act of 2002 required the GAO
Overall, our market-based tests suggest that to evaluate the potential effects of requiring
there was a positive market reaction around the mandatory rotation for audit firms. The GAO
disclosure of an auditor switch from Andersen and concluded (GAO, 2003) that mandatory rotation
that the magnitude of this reaction was negatively may not be the ‘most efficient way to enhance
(positively) associated with good governance (high auditor independence’, but suggested that the
switching costs). Overall we observe a less positive SEC and PCAOB monitor the effectiveness of
impact for firms with good governance which is the Sarbanes-Oxley Act and determine ‘whether
consistent with market participants being less further revisions, including mandatory audit firm
surprised around the disclosure of the switch. rotation, may be needed to enhance auditor
Conversely we observe a more positive impact independence and audit quality to protect the
for firms with higher switching costs, which is public interest’. One common element between a
consistent with market participants being more mandatory auditor rotation regime and the collapse
surprised/relieved on finding out about the switch of Andersen is that the clients of auditors are forced
from Andersen. to find a new auditor in both situations.37 The

results of our study suggest that deliberations includes only 937 clients because of data
about mandatory rotation should take into availability constraints. Therefore the results
consideration the fact that switching costs differ may not be generalizable to the population of
across companies. all Andersen clients. However, our sample size
of 937 is comparable to those in other studies
that examine post-Enron auditor switches (for
ACKNOWLEDGEMENTS
example, Barton, 2005, with n = 891; Blouin
The authors appreciate the helpful comments of et al., 2007, with n = 407; and Bewley et al., 2008,
two anonymous reviewers, Chris Agoglia, Jim with n = 711).
Bierstaker, Nandini Chandar, Paul Chaney, Lucy 7. About two-thirds of Andersen’s clients had
Chen, Barbara Grein, Naser Khaledi, Jayanthi December 31 fiscal year ends and consequently
Krishnan, Jong Eun Lee, Emeka Nwaeze, Mike their audits would likely be occurring during
Peters, Eric Press, David Ryan, Heibatollah Sami, this timeframe.
Shelly Ye, Yinqi Zhang, Yinghong Zhang, and 8. We use the extended window to capture any
seminar participants at the 2004 American after-effects to the switching announcement.
Accounting Association Annual Meeting, Temple 9. CCAR is calculated as the prediction error
University, Drexel University, and the Philadelphia from Fama and French’s (1993) model with
Area Research Colloquium. Steven Balsam and Carhart’s (1997) momentum modification
Jagan Krishnan acknowledge research support estimated across the period (-250, -5).
from Temple University Fox School of Business The Fama–French–Carhart four-factor model
and Management’s Merves Research Fellowships. presents a time-series model describing the
evolution of excess security returns (relative
to a risk-free rate) as a function of excess
NOTES
market returns, a high-minus-low market-to-
1. On May 31, 2005, the Supreme Court book ratio factor, a small-minus-big market
overturned Andersen’s conviction on the capitalization factor, and a momentum
grounds that ‘the trial judge erred by granting adjustment. CMAR, on the other hand, is a
the government’s request to loosen the simple market adjusted measure (firm’s
standard jury instructions’ (Bravin, 2005). return minus the market return for the same
2. Chen and Zhou (2007) find mixed evidence of day) that does not depend on an estimation
the association between audit committee period.
factors and the timing of switch by Andersen 10. Our mean CCAR, while statistically
clients. We provide a detailed discussion of the significant, is smaller in magnitude, i.e., less
differences between our study and theirs in a negative, than that found in Chaney and
later section. Philipich (2002), who found abnormal returns
3. As explained below, we only examine the of -1.32% for their smaller sample of 284
market reaction to changes announced before firms. To a large extent we feel the difference
Andersen’s conviction. is driven by sample composition. That is,
4. See, for example, Nichols and Smith (1983), while the average firm in our sample has a
Schwartz and Menon (1985), Francis and market value of slightly more than $1.7 billion,
Wilson (1988), Williams (1988), Johnson and the average firm in Chaney and Philipich’s is
Lys (1990), Krishnan and Krishnan (1997), Shu three times that size, with a market value of
(2000), Hackenbrack and Hogan (2002), and almost $5 billion.
Sankaraguruswamy and Whisenant (2004). 11. Chaney and Philipich (2002) found
5. For example, on January 29, 2002, the Wall insignificant abnormal returns on the
Street Journal reported that Delta Airlines was indictment date.
considering changing its auditor. The same 12. Krishnan (2005) provides evidence consistent
article also reported that ‘several other large with ‘aggressive accounting practices’ by
companies are now reviewing their Andersen’s Houston clients. In untabulated
relationship with Andersen’ (Brannigan & analysis we examine whether these abnormal
Opdyke, 2002). returns are affected by whether the client was
6. Although Arthur Andersen audited over 1,500 audited by Andersen’s Houston office and
publicly traded clients, our final sample whether the client operated in the same

industry as Enron. For example, we find that in 19. Blouin et al. (2007) examine whether accruals
the January 10 window the abnormal returns quality improved following the switch by
were -1.77% for firms operating in Enron’s Andersen clients, but do not find any evidence
industries and -0.40% for all other Andersen of improvement following the switch to the
clients, both significantly less than zero. new auditor.
Consequently, the results indicate a strong 20. We acknowledge that firms with high accruals
and significant industry effect. Comparable quality (low ADACCR) may decide to switch
differences were also observed for the sooner (Bewley et al., 2008) and that good
February 4 window. Similarly for the January corporate governance is likely to be associated
10 window the abnormal returns for clients of with high quality accruals (Klein, 2002). In
Andersen’s Houston office were more negative contrast, Bradshaw et al. (2001) find that auditor
(-2.48%) than for Andersen’s non-Houston change is negatively associated with high
clients (-0.36%). While both returns were accrual companies.
significantly less than zero, the results show 21. Excluding these two variables from the models
a strong and significant Houston effect. A does not affect the conclusions.
comparable Houston effect is observed for the 22. During our sample period the 10-K had to be
May 6 window. No significant industry or filed with the SEC within 90 days after the end
Houston effects are detected for the other of the fiscal year. Hence if the fiscal year end
windows. was December 31, the 10-K must be filed by
13. To identify the financial experts on audit March 31. Prior research, e.g., Easton and
committees, we use the definition of Zmijewski (1993), finds that most firms file
‘accounting financial expertise’ in DeFond et al. close to that deadline.
(2005) which is similar to that suggested in SEC 23. The correlations between the variables (not
(2002). tabulated) are low. The highest correlation is
14. Recent studies (e.g., Ferguson et al., 2003) 0.4519 between DISTRESS and INDSPL and
provide evidence that auditors earn a premium the highest VIF is 4.8609 indicating that
only when they specialize jointly at the local multicollinearity is not a problem.
and national level implying that the quality of 24. In this regression, the dependent variable is the
industry specialists could differ depending on client’s probability of switching from Andersen
the degree of specialization at the local and at time t, provided it has not switched earlier.
national levels. We acknowledge this as an 25. For continuous variables the percentage change
avenue for future research. Also, our results are in hazard rate is estimated as 100[exp(Sj bj) – 1],
similar when market share is calculated on the where Sj is the sample standard deviation
basis of sales and total assets. of variable j and bj is the estimated coefficient.
15. We also performed a sensitivity test by using For dichotomous variables, this change is
the ratio of total client fees to total auditor fees estimated as 100[exp(bj) – 1].
in place of non-audit fees. Our results are 26. The expected signs in the Cox model are
qualitatively similar. opposite to those in other regressions.
16. We acknowledge that many non-audit services 27. To rule out the possibility that the tenure
were banned by the Sarbanes-Oxley Act of variable may be confounded by age of the
2002. However, the switching decision by company, we re-estimated the models after
Andersen clients took place several months adding an additional variable, AGE, measured
prior to the enactment of SOX. as the number of years the company has been
17. Our results reported later are qualitatively publicly traded. The TENURE variable
unchanged when three, seven, and nine continues to be significant.
years are used as cutoffs instead of five 28. We thank an anonymous reviewer for this
years. interpretation.
18. An alternative interpretation is that the long 29. The results are unchanged when we redefine
auditor tenure may capture client satisfaction DFYR as including all fiscal years ending in
with the auditor (Barton, 2005) and therefore November, December, and January. Also, our
the delay in switching to a new auditor may results are qualitatively similar when we
reflect client satisfaction and not low audit restrict our sample to those with fiscal year
quality. ending in December.

30. We confine our market test to the 635 firms REFERENCES

that switched from Andersen prior to the
conviction date of June 15, 2002, since after this Abbott, L. J. & Parker, S. (2000), ‘Auditor selection
date the switch was unavoidable and there and audit committee characteristics’, Auditing:
would be very little surprise to the switch A Journal of Practice & Theory, Vol. 19, Fall,
pp. 47–66.
announcement. Ashbaugh, H., LaFond, R. & Mayhew, B. W. (2003),
31. We use the equally-weighted market index ‘Do nonaudit services compromise auditor
to calculate CMAR. However, the results are independence?’ The Accounting Review, Vol. 78, July,
similar when we use the value-weighted pp. 611–39.
market index. Atiase, R. K. (1985), ‘Predisclosure information, firm
32. We manually check newswires and major capitalization and security price behavior around
earnings announcements’, Journal of Accounting
newspapers for the ten day period (-7, 2) Research, Vol. 23, Spring, pp. 21–36.
around the 8-K filing date for concurrent Baber, W. R., Kumar, K. R. & Verghese, T. (1995),
disclosures. We identify 192 such cases out of ‘Client security price reactions to the Laventhol
the sample of 635 firms. To ensure that these and Horwath bankruptcy’, Journal of Accounting
concurrent disclosures are not driving the Research, Vol. 33, Autumn, pp. 385–96.
results reported above, we rerun our tests after Balsam, S., Krishnan, J. & Yang, J. S. (2003), ‘Auditor
deleting these 192 firms. Our conclusions are industry specialization and earnings quality’,
Auditing: A Journal of Practice & Theory, Vol. 22,
unchanged. September, pp. 71–97.
33. One of the effects may be the differential Barton, J. (2005), ‘Who cares about auditor
pre-disclosure information environment reputation?’ Contemporary Accounting Research, Vol.
(Atiase, 1985) that is captured by LMV. As a 22, Fall, pp. 549–86.
result, investors in larger firms may have Beatty, R. P. (1989), ‘Auditor reputation and the
anticipated the auditor switch and therefore the pricing of initial public offerings’, The Accounting
Review, Vol. 64, October, pp. 693–709.
coefficient on LMV could be negative. Bédard, J., Chtourou, S. M. & Courteau, L. (2004), ‘The
34. As discussed above, switches, especially the effect of audit committee expertise, independence,
later ones, were probably anticipated by and activity on aggressive earnings management’,
investors. One example is Delta Airlines, Auditing: A Journal of Practice & Theory, Vol. 23,
whose actual switch in March was September, pp. 13–35.
predated by several weeks of discussion in Beltran, L. (2002), ‘Andersen faces a fight for its life:
trial begins Monday as settlement hopes fade;
the financial press. Consequently the returns Duncan expected to testify’, CNN Money, available
observed in reaction to the switch at: http://money.cnn.com/2002/05/03/news/
announcement do not reflect the full benefit andersen_walkup/
of the switch. Bewley, K., Chung, J. & McCracken, S. (2008), ‘An
35. In untabulated analysis we find that auditor examination of auditor choice using evidence from
change announcements by non-Andersen Andersen’s demise’, International Journal of
clients during our sample period were met by Auditing, Vol. 12, No. 2, pp. 89–110.
Blouin, J., Grein, B. & Rountree, B. (2007), ‘An analysis
insignificant negative market reactions and of forced auditor change: the case of former Arthur
that the difference in return between Andersen Andersen clients’, The Accounting Review, Vol. 82,
and non-Andersen changes was positive and May, pp. 621–50.
significant. Bradshaw, M. T., Richardson, S. A. & Sloan, R. G.
36. We also run regressions with DELAY and (2001), ‘Do analysts and auditors use information
dummies for different time periods replacing in accruals?’ Journal of Accounting Research, Vol. 39,
June, pp. 45–74.
ABNDELAY. The overall conclusions are Brannigan, M. & Opdyke, J. D. (2002), ‘Delta Air looks
unaltered. for a new auditor as it considers replacing
37. We acknowledge that, unlike the situation faced Andersen’, Wall Street Journal, January 29.
by Andersen’s clients, under the mandatory Bravin, J. (2005), ‘Justices overturn criminal verdict
auditor rotation regime the parties involved in Andersen case; court ruling could inhibit
would know ahead of time that the auditor government’s broader push against white-
collar crime’, The Wall Street Journal, June 1,
change will occur. Nevertheless, we argue p. A.1.
that variations in switching costs would Callen, J. L. & Morel, M. (2003), ‘The Enron-Andersen
differentially affect clients under a mandatory debacle: do equity markets react to auditor
rotation regime as well. reputation?’ Finance Letters, Vol. 1, No. 5.

Carcello, J. V. & Neal, T. L. (2000), ‘Audit committee Ettredge, M. & Greenberg, R. (1990), ‘Determinants
composition and auditor reporting’, The Accounting of fee cutting on initial audit engagements’,
Review, Vol. 75, October, pp. 453–67. Journal of Accounting Research, Vol. 28, Spring,
Carcello, J. V. & Neal, T. L. (2003), ‘Audit committee pp. 198–210.
characteristics and auditor dismissals following Fama, E. & French, K. (1993), ‘Common risk factors in
“New” going-concern reports’, The Accounting the returns on stocks and bonds’, Journal of Financial
Review, Vol. 78, January, pp. 95–117. Economics, Vol. 33, No. 1, pp. 3–57.
Carhart, M. M. (1997), ‘On persistence in mutual fund Ferguson, A., Francis, J. R. & Stokes, D. J. (2003), ‘The
performance’, Journal of Finance, Vol. 52, No. 1, pp. effects of firm-wide and office-level industry
57–82. expertise on audit pricing’, The Accounting Review,
Chaney, P. K. & Philipich, K. L. (2002), ‘Shredded Vol. 78, April, pp. 429–48.
reputation: the cost of audit failure’, Journal Francis, J. R. & Wilson, E. R. (1988), ‘Auditor changes:
of Accounting Research, Vol. 40, September, A joint test of theory relating to agency costs and
pp. 1221–45. auditor differentiation’, The Accounting Review, Vol.
Chen, K. Y. & Zhou, J. (2007), ‘Audit committee, board 63, October, pp. 663–82.
characteristics, and auditor switch decisions by Frankel, R. M., Johnson, M. F. & Nelson, K. K. (2002),
Andersen’s clients’, Contemporary Accounting ‘The relation between auditors’ fees for nonaudit
Research, Vol. 24, Winter, pp. 1085–117. services and earnings quality’, The Accounting
Coles, J. L., Daniel, N. & Naveen, L. (2008), ‘Boards: Review, Vol. 77, Supplement, pp. 71–105.
Does one size fit all?’ Journal of Financial Economics, Government Accountability Office (GAO) (2003),
Vol. 87, February, pp. 329–56. Public Accounting firms: Required Study on the
Cox, D. (1972), ‘Regression models and life tables’, Potential Effects of Mandatory Audit Firm Rotation.
Journal of the Royal Statistical Society, Vol. 24, pp. Washington DC: GAO.
187–201. Hackenbrack, K. E. & Hogan, C. E. (2002), ‘Market
Davis, L. R., Soo, B. & Trompeter, G. (2009), ‘Auditor response to earnings surprises conditional on
tenure and the ability to meet or beat earnings reasons for an auditor change’, Contemporary
forecasts’, Contemporary Accounting Research, Vol. Accounting Research, Vol. 19, Summer, pp. 195–
26, Summer, pp. 517–48. 223.
DeAngelo, L. E. (1981), ‘Auditor size and audit Hribar, P. & Collins, D. W. (2002), ‘Errors in estimating
quality’, Journal of Accounting and Economics, Vol. 3, accruals: Implications for empirical research’,
December, pp. 182–89. Journal of Accounting Research, Vol. 40, March, pp.
Dechow, P. M., Sloan, R. G. & Sweeney, A. P. (1996), 105–34.
‘Causes and consequences of earnings Jensen, M. C. (1993), ‘The modern industrial
manipulation: An analysis of firms subject to revolution, exit, and the failure of internal control
enforcement actions by the SEC’, Contemporary systems’, The Journal of Finance, Vol. 48, July, pp.
Accounting Research, Vol. 13, Spring, pp. 1–36. 831–80.
DeFond, M. L. & Jiambalvo, J. (1994), ‘Debt covenant Johnson, W. B. & Lys, T. (1990), ‘The market for audit
violation and manipulation of accruals’, Journal of services’, Journal of Accounting and Economics, Vol.
Accounting and Economics, Vol. 17, January, pp. 145– 12, January, pp. 281–308.
76. Jones, J. (1991), ‘Earnings management during import
DeFond, M. L., Hann, R. N. & Hu, X. (2005), ‘Does relief investigations’, Journal of Accounting Research,
the market value financial expertise on audit Vol. 29, Fall, pp. 193–228.
committees of boards of directors?’ Journal of Klein, A. (2002), ‘Audit committee, board of director
Accounting Research, Vol. 43, May, pp. 153–93. characteristics, and earnings management’, Journal
Doogar, R., Sougiannis, T. & Xie, H. (2003), of Accounting and Economics, Vol. 33, August, pp.
‘The impairment of auditor credibility: stock 375–400.
market evidence from the Enron-Andersen saga’. Krishnamurthy, S., Zhou, J. & Zhou, N. (2006),
Working Paper, University of Illinois at ‘Auditor reputation, auditor independence, and the
Urbana-Champaign. stock-market impact of Andersen’s indictment on
Dopuch, N., Holthausen, R. W. & Leftwich, R. W. its client firms’, Contemporary Accounting Research,
(1987), ‘Predicting audit qualifications with Vol. 23, Summer, pp. 465–90.
financial and market variables’, The Accounting Krishnan, G. V. (2005), ‘Did Houston clients of Arthur
Review, Vol. 62, July, pp. 431–54. Andersen recognize publicly available bad news in
Easton, P. D. & Zmijewski, M. E. (1993), ‘SEC form a timely fashion?’ Contemporary Accounting Research,
10-K/10-Q reports and annual reports to Vol. 22, Spring, pp. 165–93.
shareholders: reporting lags and squared market Krishnan, G. & Visvanathan, K. (2008), ‘Does the SOX
model prediction errors’, Journal of Accounting definition of an accounting expert matter? The
Research, Vol. 31, Spring, pp. 113–29. association between audit committee directors’
Eichenwald, K. (2002), ‘Andersen guilty in effort to accounting expertise and accounting conservatism’,
block inquiry on Enron’, The New York Times, June Contemporary Accounting Research, Vol. 25, Fall, pp.
16, Section 1, p. 1. 827–57.

Krishnan, J. & Krishnan, J. (1997), ‘Litigation risk and Monograph No. 13, The Canadian Certified General
auditor resignations’, The Accounting Review, Vol. 72, Accountants’ Research Foundation.
October, pp. 539–60. Williams, D. D. (1988), ‘The potential determinants of
Krishnan, J., Raghunandan, K. & Yang, J. S. (2007), auditor change’, Journal of Business Finance &
‘Were former Andersen clients treated more Accounting, Vol. 15, Summer, pp. 243–61.
leniently than other clients? Evidence from Xie, B., Davidson III, W. N. & DaDalt, P. J. (2003),
going-concern modified audit opinions’, Accounting ‘Earnings management and corporate governance:
Horizons, Vol. 21, December, pp. 423–35. The roles of the board and the audit committee’,
Myers, J. N., Myers, L. A. & Omer, T. C. (2003), Journal of Corporate Finance, Vol. 9, June,
‘Exploring the term of the auditor-client pp. 295–314.
relationship and the quality of earnings: a case for Yang, J. S. & Krishnan, J. (2005), ‘Audit
mandatory rotation?’ The Accounting Review, Vol. 78, committees and quarterly earnings management’,
July, pp. 779–99. International Journal of Auditing, Vol. 9, November,
Nagy, A. L. (2005), ‘Mandatory audit firm turnover, pp. 201–19.
financial reporting quality, and client bargaining Yermack, D. (1996), ‘Higher market valuation of
power: the case of Arthur Andersen’, Accounting companies with a small board of directors’, Journal
Horizons, Vol. 19, June, pp. 51–68. of Financial Economics, Vol. 40, February, pp. 185–211.
Nelson, K. K., Price, R. A. & Rountree, B. R.
(2008), ‘The market reaction to Arthur
Andersen’s role in the Enron scandal: Loss of
reputation or confounding effects?’ Journal of AUTHOR PROFILES
Accounting and Economics, Vol. 46, December,
pp. 279–93. Sharad Asthana is an Associate Professor in the
Nichols, D. R. & Smith, D. B. (1983), ‘Auditor Department of Accounting, College of Business,
credibility and auditor changes’, Journal of University of Texas-San Antonio. His research
Accounting Research, Vol. 21, Autumn, pp. 534–44. areas are capital markets, the market for audit
Raghunandan, K. & Rama, D. V. (2003), ‘Audit services, executive compensation, and pension
committee composition and shareholder actions:
Evidence from voting on auditor ratification’, accounting.
Auditing: A Journal of Practice & Theory, Vol. 22, Steven Balsam is a Professor of Accounting and
September, pp. 253–63. Merves Senior Research Fellow at the Fox School
Raghunandan, K., Read, W. J. & Rama, D. V. (2001), of Business and Management at Temple University.
‘Audit committee composition, “gray directors,” In addition to writing a book for World@Work,
and interaction with internal auditing’, Accounting Executive Compensation: An Introduction to Practice
Horizons, Vol. 15, June, pp. 105–18.
Reynolds, J. K., Deis Jr., D. R. & Francis, J. R. (2004),
and Theory, and an AICPA self study course,
‘Professional service fees and auditor objectivity’, ‘Accounting for Stock Options and Other
Auditing: A Journal of Practice & Theory, Vol. 23, Stock-Based Compensation’, he has published
March, pp. 29–52. articles in academic journals including The
Sankaraguruswamy, S. & Whisenant, J. S. (2004), Accounting Review, Journal of Accounting Research,
‘An empirical analysis of voluntarily supplied Journal of Accounting and Economics, Contemporary
client–auditor realignment reasons’, Auditing: A Accounting Research, Journal of the American Taxation
Journal of Practice & Theory, Vol. 23, Spring, pp.
107–21. Association, Journal of Accounting and Public Policy,
Schwartz, K. B. & Menon, K. (1985), ‘Auditor switches Journal of Accounting, Auditing and Finance, and
by failing firms’, The Accounting Review, Vol. 60, Accounting Horizons; and practitioner journals
April, pp. 248–61. including the Journal of Accountancy. Professor
Schwartz, K. B. & Soo, B. S. (1996), ‘Evidence of Balsam is also a member of the editorial boards
regulatory noncompliance with SEC disclosure of the Journal of Accounting and Public Policy and
rules on auditor changes’, The Accounting Review,
Vol. 71, October, pp. 555–72. International Journal of Accounting.
Securities and Exchange Commission (SEC) (2002), Jagan Krishnan is a Professor of Accounting
Disclosure Required by Sections 406 and 407 of the and Merves Senior Research Fellow at Temple
Sarbanes-Oxley Act of 2002. Release Nos. 33-8177; University’s Fox School of Business. He has over 20
34-47235; File No. S7-40-02. years of teaching and professional experience in
Shu, S. Z. (2000), ‘Auditor resignations: clientele auditing and consulting. His research interests are
effects and legal liability’, Journal of Accounting and
Economics, Vol. 29, April, pp. 173–205.
in the areas of audit quality, corporate governance,
Simunic, D. A. & Stein, M. T. (1987), Product auditor litigation, and fraudulent financial
Differentiation in Auditing: Auditor Choice in the reporting. He serves on the editorial boards of
Market for Unseasoned New Issues. Research Accounting Horizons, Auditing: A Journal of Practice

& Theory, and Current Issues in Auditing. He has Accounting, Auditing, and Finance, International
published articles in academic journals including Journal of Auditing, and Accounting Horizons. Prior
The Accounting Review, Contemporary Accounting to entering academia, he worked in the audit and
Research, Auditing: A Journal of Practice & Theory, consulting divisions of affiliates of Ernst & Young
Journal of Accounting and Public Policy, Journal of and KPMG.

Asthana Et Al (2010)

Uploaded by

Copyright:

Available Formats

Asthana Et Al (2010)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Asthana Et Al (2010)

Uploaded by

Copyright:

Available Formats

International Journal of Auditing doi:10.1111/j.1099-1123.2010.00417.

Corporate Governance, Audit Firm

Sharad C. Asthana,1 Steven Balsam2 and Jagan Krishnan2

The financial scandal surrounding the collapse of Enron

Key words: Auditor reputation, auditor change, Arthur Andersen,

SUMMARY auditor change by former Andersen clients. After

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

Table 1: Sample selection and distribution

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

Table 2: Cumulative abnormal returns

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

Table 3: Firm characteristics by timing of switch

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd

Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

Table 5: Market reaction around disclosure of auditor change

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

variable. Given the models are comparable (i.e., CONCLUSIONS

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

30. We confine our market test to the 635 firms REFERENCES

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

© 2010 Blackwell Publishing Ltd Int. J. Audit. 14: 274–293 (2010)

You might also like