blouin et al - 2005 - the ultimate form of mandatory auditor rotation - the case of former arthur andersen clients

The Ultimate Form of Mandatory Auditor Rotation:
The Case of Former Arthur Andersen Clients

Jennifer Blouin
University of Pennsylvania

Barbara Grein
Drexel University

Brian Rountree
Rice University

This Draft: February 2005



Abstract: The collapse of Arthur Andersen provides a unique quasi-experimental setting to study
the implications of mandatory auditor rotation for public firms. Consistent with the extant
literature on mandatory auditor rotation, we hypothesize the selection of a new auditor is a
function of agency and switching costs. Using a unique dataset identifying whether or not
former Andersen clients followed their audit team to a new auditor, we are able to find support
for both hypotheses. Further, we find firms with greater performance adjusted discretionary
accruals were more likely to follow their Andersen audit team to a new auditor, consistent with
these firms trying to mitigate the costs of switching auditors. However, tests of accruals in the
subsequent year reveal the most aggressive firms who followed Andersen staff were no longer
aggressive in the following year relative to the group who chose new non-Andersen auditors.
This is inconsistent with mandatory auditor rotation directly improving financial reporting since
we expect the non followers to exhibit the greatest degree of accrual correction. However, our
results must be interpreted with caution since the rotation described in the paper is quite atypical.
Further, whether the benefits outweigh the costs is still difficult to determine since we cannot

empirically quantify these aspects.

Keywords: mandatory rotation, audit quality, earnings quality, Arthur Andersen

Data Availability: Data are available from public sources

* We would like to thank Kevin Raedy, Stefanie Tate, and workshop participants at Drexel
University and the University of Massachusetts at Lowell for constructive criticisms and
suggestions.

1
I. INTRODUCTION
Mandatory auditor rotation policies have been considered since at least the early 1900s.
Zeff (2003) documents intra-firm correspondence at E.I. du Pont de Nemours & Company
(Dupont) dating back to 1922 concerning the company’s policy of switching auditors every year.
The primary force underlying the auditor rotation policy was to insure the independence of the
public accountant. Similar reasoning has resurfaced after the recent business scandals (e.g.
Enron, WorldCom, Tyco) as part of the national debate on mandatory auditor rotation and the
quality of financial statements.
1

The primary tradeoff involved in the debate on mandatory auditor rotation involves
agency benefits relative to switching costs. Switching costs have been defined as the start-up
costs incurred by both client and auditor for a new audit engagement, as well as the increased
risk of audit failure in new audits (GAO 2003). Commonly cited benefits of auditor rotation are
an increase in auditor independence, which may lead to improved audit and financial statement
quality. The relative magnitude of the costs and benefits of switching auditors is central to this
debate. We use the unique setting created by the collapse of Arthur Andersen (AA) to examine
these costs that are inherently captured by a firm’s selection of a new auditor. Specifically, we
examine which costs/benefits explain a client’s decision to follow their former AA audit team

(follow firms) or to choose an entirely new audit firm (not-follow firms). We also examine the
relative quality of financial reporting before and after the change in auditor for the follow and
not-follow firms.
Typically, a change in auditor involves two actions: dismissal/resignation of the current
audit firm and the selection of a new auditor. Most prior research on auditor changes has been


1
A summary of the mandatory auditor rotation debate appears in the GAO report entitled “Public Accounting Firms:
Required Study on the Potential Effects of Mandatory Audit Firm Rotation.”

2
unable to examine the two actions separately and therefore has focused on the joint action (see
for example Nichols and Smith 1983; Francis and Wilson 1988; Shu 2002 and Landsman et al.
2005).
2
After AA’s criminal indictment on June 15, 2002 and subsequent agreement to stop
practicing in front of the SEC, all of their public clients were forced to select a new auditor.
3

Therefore, our sample of former AA clients is homogeneous in the decision to dismiss their
auditors, enabling us to create more direct tests of the costs involved in the selection of a new
auditor than has been possible in past studies.
We hypothesize the selection of a new auditor involves three primary concerns: 1)
switching costs, 2) agency costs, and 3) implicit insurance provided by the auditor. We hold the
latter constant by considering only switches to one of the remaining Big4 audit firms (Deloitte,
PriceWaterhouseCoopers, KPMG, or Ernst & Young).
4
We characterize the selection of a new
auditor as being either: 1) completely independent of any existing relationships with AA’s audit

personnel or 2) based on the prospective employment of the AA audit team. For example, in
Casella Waste Systems’ Form 8-K filing on June 13, 2002, the firm reports:
As recommended by the audit committee, the Board of Directors on May 20, 2002, decided to no
longer engage its independent accountants, Arthur Andersen LLP, and engaged KPMG LLP
(“KPMG”) to serve as the Company’s independent accountants for the fiscal year ending April 30,
2003 and to audit the Company’s financial statements for the fiscal year ended April 30, 2002.
The Audit Committee’s recommendation to engage KPMG was based on the assumption that
certain individuals from Arthur Andersen’s Boston, Mass. office, including the team auditing the
Company, would join KPMG. That event did not occur. As a result, the Audit Committee
subsequently reconsidered its recommendation and, as recommended by the Audit Committee, the
Board of Directors on June 13, 2002 decided to no longer engage KPMG, and engaged
PricewaterhouseCoopers LLP (“PWC”) to serve as the Company’s independent accountant for the

2
Schwartz and Menon (1985) is a notable exception that examines factors associated with 35 firms that changed
auditors because of bankruptcy related issues.
3
Arguably, the demise of Andersen occurred on March 14, 2002, the date of Andersen’s indictment, while the
criminal conviction simply represented the ceremonial nail in the coffin. Supporting this position, the Wall Street
Journal stated “In the 212-year history of the U.S. financial markets, no major financial-services firm has ever
survived a criminal indictment.” “Called to Account: Indictment of Andersen In Shredding Case Puts Its Future in
Question,” Wall Street Journal, March 15, 2002 (page A1).
4
This assumes that the relative implicit insurance provided by the remaining Big4 auditors is in fact reasonably
equal. This is consistent with prior literature that examined implicit insurance (i.e. Menon and Williams 1994),
which utilizes a BigN/non-BigN designation to test for differences in insurance values.

3
fiscal year ending April 30, 2003 and to audit the Company’s financial statement for the fiscal
year ended April 30, 2002.

As it turned out, AA’s Boston office actually became part of PWC rather than KPMG. We argue
that firms like Casella Waste Systems did not switch audit teams, but instead simply transferred
their existing audit relationship to a new firm. Since other firms clearly severed ties with their
former AA audit team, we have identified an interesting quasi-experimental setting to study the
cost/benefit relationship underlying the selection of a new auditor under a mandatory rotation
regime.
We find firms with greater agency issues and/or monitoring concerns, as measured by
earnings transparency, geographic diversity and the presence of an outside blockholder, were
more likely to severe ties with Andersen completely and select an entirely new auditor. At the
same time, using an indicator whether AA was the industry expert and performance adjusted
discretionary accruals as proxies for switching costs, we find firms with greater switching costs
were more likely to follow their former AA audit team to the new auditor. The unique research
design allows us to simultaneously determine if switching and agency costs played a role in the
selection of a new auditor in a setting incrementally more powerful than previous studies using
measures such as abnormal returns and duration.
The final set of tests provide direct evidence concerning the hypothesis posed and tested
in Myers et al. (2003) on the quality of earnings before and after mandatory auditor rotation. As
discussed above, we find that firms that followed AA had more aggressive performance matched
discretionary accruals relative to the non-follow firms in the final year audited by AA. Utilizing
this discretionary accrual measure as an indicator of earnings quality, results are consistent with
more aggressive firms attempting to follow AA in an effort to minimize the costs associated with
said behavior. However, when we model the discretionary accruals using the framework

4
outlined in Myers et al. (2003), we find firms in the lowest quintile of accruals (i.e. the most
conservative) in the final year of AA continue to have lower discretionary accruals on average in
the following year regardless of the follow decision. This finding suggests the mandatory
rotation did nothing to improve the reporting for firms in this particular tail. Further, the follow
firms in the highest quintile of discretionary accruals in their final year with AA exhibit

reversion in the following year under a new auditor whereas the corresponding not-follow firms
do not. Given followers did not fully switch firms per se, we expect the not-follow group to
exhibit the greatest degree of reversion in aggressive behavior if mandatory rotation is effective
in improving financial reporting. This is not the case and therefore casts some doubt that
reporting quality would be greatly influenced under a mandatory rotation regime. However,
these results need to be interpreted with caution since this is not the typical mandatory rotation
regime entertained by the literature and rule making bodies. We discuss these limitations in
section 4 below.
The rest of the paper is organized as follows. Section 2 reviews the related literature on
mandatory auditor rotation and develops hypotheses. Sample selection and research design are
outlined in section 3. Section 4 provides results and robustness checks. Section 5 concludes.

II. H
YPOTHESIS DEVELOPMENT
The GAO’s November 2003 report on mandatory auditor rotation states that the majority
of Tier 1 public accounting firms and Fortune 1000 public companies “believe that the costs of
mandatory audit firm rotation are likely to exceed the benefits.” Costs identified by the GAO
include the risk of audit failure in the early years of an audit engagement, audit firm competition
issues, increased initial year audit costs, auditor selection costs and support costs. The report

5
goes on to explain that the “benefits of mandatory audit firm rotation are harder to predict and
quantify …” (GAO 2003, 8). The prior literature on auditor changes has not addressed this issue
because of the general lack of mandatory auditor rotation regimes.
5
Instead, research has
focused on auditor changes that result from the dismissal or resignation of the incumbent auditor,
which is obviously quite different than mandatory rotation. However, the collapse of Arthur
Andersen creates the opportunity to study the costs and benefits related to mandatory auditor
rotation.

We acknowledge that the collapse of Andersen is not the standard mandatory auditor
rotation setting. Nevertheless, it is arguably the closest setting available to date with enough data
to properly address this important issue. The unexpected and rapid collapse of Arthur Andersen
provides us an opportunity to examine a group of firms that switched auditors for the same
reason: their former audit firm was forced to stop practicing. We use this mandatory rotation to
examine a firm’s choice of new auditor. Specifically, we examine which costs explain a client’s
decision to follow their former AA audit team or to choose an entirely new audit firm. Given the
decision to change auditors has been uniformly mandated, prior research on auditor change and
the debate on auditor rotation suggest three potential costs involved in the selection of a new
auditor - switching, agency, and implicit insurance. We hold the latter constant by only
examining switches to the remaining Big4 auditors, allowing us to focus on switching and
agency costs.
6

We also utilize the setting to test the implications of a mandatory auditor rotation on the
financial reporting characteristics of firms. Specifically, we analyze firms’ performance adjusted
discretionary accrual behavior surrounding the collapse of AA. Myers et al. (2003) document

5
Exceptions include Austia, Brazil, Italy and Singapore which currently have mandatory rotation policies. Also,
Spain and Canada had mandatory policies that were ended in 1995 and 1991 respectively (GAO 2003).
6
See footnote 4 above.

6
more aggressive accrual behavior for firms with shorter auditor tenure. They interpret their
findings as being inconsistent with mandatory auditor rotation improving financial reporting.
However, the authors recognize that they are not analyzing mandatory auditor rotation and that
their results are simply suggestive. On the contrary, the current setting is a more direct form of
mandatory auditor rotation and therefore has the potential to be informative for this debate.

S
WITCHING COSTS
We define switching costs as the start-up costs incurred by both the client and auditor for
a new audit engagement. These include the (1) costs incurred by the auditor to learn about the
company’s operations, systems, financial reporting practices, and accounting issues, (2) costs
incurred by the client to aid the auditor in understanding its operations, and (3) costs incurred by
a client in selecting a new auditor (GAO 2003). Further, there is an increased risk of audit
failure. AICPA (1978), Palmrose (1986, 1991), Geiger and Raghunandan (2002) and Myers et
al. (2003) all find evidence consistent with a greater likelihood of audit failure in early years of
an audit engagement.
Standard value maximizing behavior suggests that firms will seek to minimize switching
costs, ceteris paribus. We hypothesize that companies may try to minimize the cost of switching
auditors by following their AA audit team who already possess client and industry specific
knowledge or more succinctly:
H
1
: The greater the switching costs, the more likely a former AA client will follow its
AA audit team to a new auditor, ceteris paribus.
7



7
The maintained assumption throughout is that, ceteris paribus, following AA has lower switching costs than not-
following. Education of the audit team about the operations of the firm along with obtaining comfort with the
reported results is an expensive and time consuming proposition. Following AA would almost certainly reduce
these costs even if the audit team is not maintained because at minimum they would be available for consultation.

7

AGENCY COSTS
Consistent with Jensen and Meckling (1976), we define agency costs as monitoring
expenditures by the principal, bonding expenditures by the agent, and loss in welfare
experienced by the principal due to the agent not acting in the principal’s best interest. Auditing
is widely believed to be a means of reducing agency costs through the monitoring of the agent by
an independent third party auditor (Jensen and Meckling 1976 and Watts and Zimmerman 1983,
among others). Further the greater the agency costs, the greater is the demand for high quality
audits (DeAngelo 1981; Dopuch and Simunic 1982).
8

The decision to change auditors is frequently cast in terms of mitigating agency costs
and/or changes in audit quality (Nichols and Smith 1983; Francis and Wilson 1988; Johnson and
Lys 1990; DeFond 1992). In our setting, it may be that agency conflicts at the firm are
unchanged, while the perceived quality of the AA audit has suddenly declined. The results in
Chaney and Philipich (2002) and Krishnamurthy et al. (2003), which document negative market
reactions for Andersen clients after negative news concerning their auditor, suggest that investors
may have perceived audit quality issues to be systematic at AA. Further, duration analyses
examining cross sectional differences in former AA clients support the notion that firms were
concerned about the perceived quality of AA’s audits by illustrating clients with greater agency
conflicts dismissed AA sooner (Chang et al. 2003, Barton 2004). If firm management perceived
audit quality issues and/or is concerned with investors’ perceptions of audit quality, then we
hypothesize that:


Consistent with this notion, the GAO found that Tier 1 public accounting firms (firms with 10 or more public
company clients that were members of the AICPA’s self-regulatory program for audit quality) “generally saw more
potential value in having access to the previous audit team and its audit documentation than in performing additional
audit procedures and verification of the public company’s data during the initial years of the auditor’s tenure” (GAO
2003, 2133).
8

Consistent with DeAngelo (1981) and DeFond (1992), we define audit quality as the probability that an audit firm
will detect and report “material breaches in the accounting system.”

8
H
2
: The greater the agency conflicts, the more likely a former AA client will not follow
its AA audit team to a new auditor, ceteris paribus.
It is important to note that a firm may have conflicting costs (i.e. high agency and switching
costs), which biases against finding any systematic relation between the decision to follow and
our measures of the underlying costs.

III.
RESEARCH DESIGN AND SAMPLE SELECTION
We model the decision to follow AA personnel as a function of variables aimed at capturing
the degree of switching and agency costs, along with industry fixed effects to allow for
differences in mean follow rates across industries:
εγγγ
γγγγ
γγγγγα
++++
++++
+++++=
∑
RETURNLOSSROA
INSIDERBLOCKLEVERAGEACCRUAL
COMPLEXCYTRANSPARENSIZETENUREEXPERTFOLLOW
II
121110
9876

54321

where all variables are measured as of the final year audited by AA and are defined as follows
(Compustat data items in parentheses):

FOLLOW
= 1 if the client followed AA, and 0 otherwise;
EXPERT
= 1 if AA had at least 5% more clients in a particular industry and
state than the next closest competitor, and 0 otherwise;
TENURE
= number of years audited by AA per Compustat;
SIZE
= natural logarithm of total assets (#6);
TRANSPARENCY
= decile rank of absolute value of residual from regression of annual
returns on annual earnings (#18), changes in annual earnings, both
scaled by total assets (#6) and SIZE.
COMPLEX
=
∑
=

























N
i
i
i
TotalSales
Segment
Segment
TotalSales
LN
1

where TotalSales is company sales revenue for 2001 and Segment
i

represents the sales for a specific geographic segment of the
business per Compustat (Bushman et al. 2002).
ACCRUAL
= performance adjusted discretionary total accruals following
Kothari et al. (2004);
LEVERAGE
= ratio of debt (#9 + #34) to total assets (#6);
BLOCK
= 1 if an outside blockholder per Spectrum holds at least 5% of the

9
outstanding shares, and 0 otherwise;
INSIDER
= 1 if an insider per Spectrum holds at least 5% of the outstanding
shares, and 0 otherwise;
ROA
= return on assets, defined as net income before extraordinary items
(#18) divided by ending total assets (#6);
LOSS
= 1 if ROA < 0, and 0 otherwise;
RETURN
= abnormal market model return for the ±1 days surrounding AA’s
indictment date, using CRSP’s value weighted index as a proxy for
the market.
I
denotes industry as defined in Barth et al. (1998).

We utilize logistic regression to isolate the determinants of auditor choice. The empirical
specification of the dependent variable, FOLLOW, is a 1 when firms are categorized as following
AA, and 0 otherwise. In constructing our sample, we identified firms that were audited as of

fiscal year 2000 or 2001 by AA and changed auditors after November 8, 2001.
9
Next, we hand
collected information concerning the acquisition of AA offices by other auditors from a variety
of sources including audit firm press releases, AA client Form 8-Ks relating to the choice of a
new auditor, and representatives from two of the remaining Big4 audit firms. Through this
process we were able to classify 561 former AA clients as either following AA personnel to a
new auditor or completely severing ties with their AA audit team.
10
Of these firms, 425 have the
necessary financial statement information to perform our baseline tests concerning the decision


9
AA received a subpoena from the SEC as Enron’s auditor on November 8, 2001. The following highlights other
key dates in Arthur Andersen’s collapse. On March 15, 2002, the grand jury indictment of AA was unsealed. AA
signed and announced a Memorandum of Understanding with Deloitte and Touche for the sale of its tax practice on
April 4, 2002, following through on plans to reduce its business to just the core audit practice. The criminal trial of
AA began on May 6, 2002, the same day that AA agreed to settle a lawsuit with the Baptist Foundation of Arizona
for $217 million. The first of many office sale announcements, was also made on May 6, 2002 – Ernst & Young
acquired the Detroit, Toledo, Ann Arbor and Grand Rapids offices of AA. Finally, AA was convicted of one count
of obstructing justice on June 15, 2002. As a result of the conviction, AA agreed to cease practicing before the SEC
by August 31, 2002.
10
For example, KPMG acquired AA’s Philadelphia office. If an AA client whose audit opinion was signed
Philadelphia chose KPMG as their new auditor, we assume they followed their AA audit team. If that same client
chose Ernst & Young, we assume that they did not follow their AA audit team.

10
to follow AA or not and represent switches that range from February 12, 2002 to August 2,

2002.
11
Table 1 provides a summary of the sample selection process.
Our methodology is novel as it allows us to distinguish between the switching costs and
agency costs hypotheses in a systematic fashion that neither returns nor duration can necessarily
replicate. For instance, in an abnormal returns analysis, the researcher may be unable to
disentangle switching costs, agency costs, and insurance costs since the magnitudes of the
returns may be equivalent for firms facing different cost issues.
12
In contrast, our research
design provides the opportunity to isolate these alternatives even if the magnitudes of their
influence are equivalent since, presumably, they will lead to different auditor selection decisions.
Although there are strengths and weaknesses to any methodology, we believe our setting
provides us with a unique opportunity to study the selection of auditors from a relatively
homogenous set of potential auditors.

S
WITCHING COSTS
The regressors in the model are measures of switching and agency costs motivated by prior
research. EXPERT, TENURE, SIZE, TRANSPARENCY, COMPLEX and ACCRUAL all relate to
switching costs and, with the exception of TENURE and TRANSPARENCY, all are predicted to
have positive coefficients suggesting the greater the switching costs the more likely the firm is to
follow AA. Note, SIZE, TRANSPARENCY and COMPLEX are also related to agency costs,
which we describe below in section 3.2.


11
We perform robustness tests by restricting the characterization of following AA to periods in which public
documents indicated the office switch occurred prior to the firm’s announcement of the switch. The results remain
qualitatively and statistically unchanged.

12
For instance, firm A may have had a -2% abnormal return on the indictment date because this was the market’s
assessment of the costs of switching auditors. Firm B may have also experienced the same return, but because of
agency related issues while firm C experienced the same return due to a loss of insurance and agency costs. Since
these firms will presumably vary in regards to measures of switching, agency and insurance costs the ability to
isolate these alternatives is hindered.

11
The model includes a measure of industry expertise, EXPERT. Expertise reduces the start-
up and switching costs for clients opting to hire those auditors. If AA was the industry expert,
then switching costs may be reduced by following the expert to the new audit firm, hence the
predicted positive coefficient. Similarly, we predict a positive coefficient on SIZE, because the
costs of changing auditors are expected to be higher for larger clients (DeAngelo 1981).
13

TENURE is an auditor related variable that corresponds to switching costs, but its
direction cannot be specified. For instance, DeAngelo (1981) suggests there may be a
relationship specific investment between auditor and client where, in order to recover start-up
costs and switching costs, the two firms are better off maintaining their relationship, at least in
the early years. This suggests firms with shorter TENURE will be more likely to follow AA.
However, firms with extended TENURE may also find it costly to switch since they have
developed relations with their auditor over an extended period of time making a sign prediction
ambiguous.
Next, we expect switching costs to be decreasing in the financial transparency of the firm.
Following prior research (Easton and Harris 1991, Bushman et al. 2004, Barth and Landsman
2004), we measure financial reporting transparency as the degree to which a firm’s accounting
summary measures correlate with its economic value. The variable TRANSPARENCY is the
decile rank (in descending order) of the absolute value of the residual from a cross sectional
regression of annual returns on ROA, changes in earnings, SIZE and industry fixed effects.
Observations in the highest decile are those with the highest transparency, while those in the

lowest decile are those with the lowest transparency. We predict a negative coefficient for


13
An alternative interpretation of a positive association would be that SIZE is a proxy for audit fee potential
consistent with Simunic (1980) and therefore simply represents the effort of former AA partners to maintain their
most lucrative clients.

12
TRANSPARENCY because we expect firms with greater transparency to find it easier to switch
auditors.
14

Related to financial transparency is the firm’s valuation complexity. Bushman et al.
(2002) utilize COMPLEX as a proxy for valuation complexity in analyzing the market-wide
effects from the Enron bankruptcy. We consider COMPLEX as a measure of overall valuation
complexity/firm transparency based on the firm’s geographic segments, where the higher the
value the less transparent or more complex and therefore more difficult to audit.
15
Under the
switching cost hypothesis, we predict companies with higher values of COMPLEX would be
more likely to follow AA.
16

The final measure of switching costs is ACCRUAL. Bradshaw et al. (2001) finds that
auditor changes are less likely for high accrual firms suggesting that it is more costly for these
firms to voluntarily change auditors. In the current context, these same firms may attempt to
reduce the costs of switching auditors by following AA resulting in a positive prediction for the
ACCRUAL coefficient. Alternatively, Defond and Subramanyan (1998) find firms changing
auditors have negative discretionary accruals on average and attribute the change to overly

conservative accounting required by the incumbent auditor. Firms with more negative accruals
may find it less costly to change auditors thereby leading to the same positive coefficient
prediction.




14
Inferences are unaltered if we utilize the actual residual value rather than the decile rank.
15
COMPLEX captures more than just the number of geographic segments. The correlation with the number of
segments is positive and significant, but is only 0.32. Further when the number of geographic segments is added to
our cross-sectional analysis, it is not significant and does not alter the significance of COMPLEX.
16
SIZE may also act as a proxy for client complexity and geographic constraints which we expect to be positively
correlated with start up costs associated with switching auditors.

13
AGENCY COSTS
Under our agency cost hypothesis, the greater the agency costs the greater the demand for
a credible audit and the less likely a client will follow their AA audit team to the new auditor.
Given the fact that sample firms were forced to switch auditors, firms with the most severe
agency issues may have opted to completely severe ties with their former auditor in an effort to
minimize the reputational damage related to their association with AA. We consider two aspects
related to agency issues – degree of conflict and monitoring by outside parties.
As previously noted, SIZE, TRANSPARENCY and COMPLEX are related to both
switching and agency costs. However, their predicted signs change under the agency costs
hypothesis. Barton (2004) uses firm size as a proxy for reputation costs from the AA collapse.
He finds that larger AA clients switched to a new auditor earlier than smaller firms and argues
that this result is attributable to larger firms being subject to greater reputation costs. In addition

to reputation costs, SIZE may also measure the diffusion of ownership and related agency costs.
If agency costs dominate the decision to switch auditors, we expect SIZE to be negatively related
to the likelihood of following the AA team.
Similarly, we expect TRANSPARENCY and COMPLEX to change signs under the agency
hypothesis to positive and negative, respectively. The inability to perfectly observe the actions
of managers by outside parties increases agency costs (Jensen and Meckling 1976).
TRANSPARENCY and COMPLEX capture income statement transparency and the overall firm
transparency/complexity, respectively. As such, they capture the degree of difficulty outside
parties have in monitoring management. Firms with lower values of TRANSPARENCY and
higher values of COMPLEX are less transparent and more difficult to monitor, which leads to a
greater demand for severing ties with AA.

14
LEVERAGE and BLOCK are alternative measures of the degree of outside monitoring that
might lead to an increased desire to completely change auditors in an effort to mitigate concerns
about the quality of the audit. DeFond (1992) argues that companies with greater leverage tend
to switch to higher quality audit firms because of the monitoring performed by bondholders. If
debt holders viewed the demise of AA as indicative of quality related problems, then we predict
the greater the LEVERAGE the less likely firms will be to follow AA. We also consider the
presence of a large blockholder as indicative of monitoring concerns (see Francis and Wilson
1988). We predict that firms with outside monitoring will be less likely to follow AA in an
effort to increase the credibility of their audit.
INSIDER captures the degree of conflict between insiders and outsiders. Jensen and
Meckling (1976) show that higher management ownership leads to greater alignment of interests
with outside owners and, hence, lower agency conflicts. The results in prior research related to
insider ownership and auditor changes have been mixed. Francis and Wilson (1988) and
Palmrose (1984) find no significant relation between insider ownership and the quality of the
successor auditor, while Simunic and Stein (1987) find a negative association and Eichenseher
and Shields (1989) find a positive association.
17

If low insider ownership (INSIDER) is
indicative of greater agency problems then we predict these firms will be less likely to follow
AA.
18

In addition to industry fixed affect, the remaining variables ROA, LOSS, and RETURN
are utilized as controls. Landsman et al. (2005) and Schwartz and Menon (1985) find that firms


17
In related research, Barton (2004) finds that firms with smaller managerial ownership were more likely to dismiss
AA sooner.
18
In unreported results, we also utilize the Kaplan and Zingales (1997) measure of the need for external financing as
a measure of the degree of conflict between insiders and outsiders, as well as equity volatility. However, neither
measure was ever significant, nor were any of the other inferences altered, therefore we excluded these variables to
make the model more parsimonious.

15
with poor financial performance are more likely to change auditors. In our context, this suggests
that poorly performing firms may be less likely to follow AA, but classifying this prediction as
related to agency or switching costs is difficult. Chaney and Philipich (2002) and Krishnamurthy
et al. (2003) find that clients experienced negative abnormal returns on AA related event dates
and conclude the reactions are related to agency concerns. We include RETURN in our analysis,
but our research design allows it to capture either switching or agency costs thereby making the
prediction of its coefficient’s sign ambiguous. Figure 1 summarizes our sign predictions under
the two hypotheses.

IV. Results
DATA AND UNIVARIATE TESTS

Our sample consists of 425 former AA clients that selected one of the remaining Big4
auditors after the collapse of AA. There are a total of 236 firms classified as following their AA
audit team, and 189 choosing not to follow. In unreported tabulations, similar breakdowns were
found incorporating non-Big4 auditors, but given the relatively small reduction in sample size
and the increased homogeneity of the sample we restrict our attention to the Big4 sample.
19
The
industry composition for the sample is illustrated in table 1, panel B.
Table 2 documents descriptive statistics for the firms that did not follow (panel A) and
firms that are designated as following their AA audit team (panel B). Neither following nor not-
following firms appears to have performed very well in the final year audited by AA as indicated


19
We have relatively little information concerning AA personnel switches to non-Big4 auditors, which reduces our
ability to generalize to this population. Furthermore, the extant literature suggests that switches to non-Big4
auditors occur for significantly different reasons than upward or lateral movements (Johnson and Lys 1990).
Although Landsman et al. (2005) illustrate downward and lateral changes involving BigN auditors are influenced by
similar characteristics, we focus on the Big4 sample in order to avoid concerns about downward switches biasing
our results. Nevertheless, results are unchanged when firms selecting non-Big4 auditors are included.

16
by mean ROAs of -0.16 and -0.09 respectively, with no significant difference between the
samples at conventional levels. The next variable of interest is RETURN, which provides
evidence consistent with the Chaney and Philipich (2002) and Krishnamurthy et al. (2003)
findings that AA firms on average experienced negative abnormal returns. Unreported statistics
show that both the mean and median RETURN for the not-follow sample are significantly less
than zero, whereas only the median RETURN is significantly less than zero for the follow
sample. This suggests that the market values of follow firms were not as influenced by the AA
collapse, but tests of differences in either means or medians fail to confirm this result.

The variables that are significantly different in cross sample comparisons are
TRANSPARENCY, COMPLEX, EXPERT and ACCRUAL. The first two measures relate to the
transparency/complexity of the firm measuring the ability of earnings to explain returns and
geographic diversity in corporate sales respectively. Results from univariate tests suggest that
firms that chose not to follow AA were less transparent than firms that followed AA. If
TRANSPARENCY (COMPLEX) represents the learning costs of new auditors then it would be
reasonable to expect AA followers to be less (more) transparent (complex) than those that did
not follow. This does not appear to be the case and presents the first evidence that agency costs
played a role in the auditor choice decision in this instance.
EXPERT and ACCRUAL results illustrate that firms chose to follow AA when AA was
the designated expert within their industry and region and when they had higher performance
adjusted discretionary accruals than their not-follow counterparts. The EXPERT result is
indicative of clients trying to minimize switching costs by following the expert to their new
auditors. The ACCRUAL results are consistent with the those in Defond and Subramanyan
(1998), who find clients switching auditors have negative discretionary accruals on average. The

17
mean and median for the not-follow firms are significantly negative, whereas the mean and
median for the follow firms are positive, but not significantly different from zero. This
illustrates that the not-follow firms were relatively conservative in their financial reporting and is
consistent with these firms facing lower switching costs. However, these are only univariate
comparisons and therefore are only suggestive.
Table 3 documents the pearson and spearman correlations among the variables in the
regression described above. In general there are a variety of significant correlations, but none
that would indicate multicollinearity problems.
20
Based on the results described above for table
2, it is not surprising to find significant correlations between FOLLOW and EXPERT,
TRANSPARENCY, COMPLEX and ACCRUAL. All are in the same direction as the univariate
tests in table 2 and no new significant correlations are noted.

Table 4 presents regression results for three different models (Model 1-Model 3). Model
1 presents our primary findings, while Models 2 and 3 provide robustness tests based on different
categorizations of the dependent variable. It is likely that some clients in our sample chose an
auditor before knowing where their former AA audit team ended up. To allow for this
possibility, in Model 2 we re-coded any company that changed auditors prior to May 1, 2002 as a
not-follow client. This is an admittedly arbitrary date, but represents a week before the first
publicly available information concerning an AA audit office takeover by another auditing firm
in our sample. As a further robustness test, in Model 3 we exclude all observations that changed
auditors prior to May 1, 2002.
The results are consistent across the different models suggesting that we have
characterized the choice of auditor with some degree of precision. Furthermore, the pseudo R-


20
Although the correlations do not indicate multicollinearity issues, we estimated standard diagnostics within the
regression framework anyway. No causes for concern were found.

18
squares of the models, which range from 0.18 to 0.21, suggest that our model has reasonable
explanatory power.
21
Repeating the analyses with only industry fixed effects, the largest R-
square is 0.04 implying that the majority of the explanatory power is attributable to our
constructs measuring switching and agency costs.
Given the consistency across the models we restrict the discussion of the results to Model
1. The first item to note is that none of our control variables, ROA, LOSS or RETURN are
significantly different from zero. The fact that RETURN is insignificant is consistent with our
earlier conjecture that the market reaction to AA’s indictment may have been indicative of both
switching and agency costs, which reduces the power of returns tests to isolate the factors
underlying the reaction.

22
This is not consistent with Chaney and Philipich (2002) and
Krishnamurthy et al. (2003) who argue the returns on AA related events are solely because of
agency issues.
The variables that are significantly different from zero and consistent with the agency
costs hypothesis are TRANSPARENCY, COMPLEX and BLOCK. We discuss each in turn
followed by a discussion of the insignificance of SIZE, which is contrary to findings in prior AA
studies using alternative research methodologies. Finally, we turn to the results for EXPERT and
ACCRUAL, which are consistent with our switching costs hypothesis.
The variable TRANSPARENCY (COMPLEX) is positive (negative) and significant
confirming our univariate results. Firms that were less transparent/more complex viewed
following their AA audit team as simply too costly implying the agency costs outweighed the


21
We acknowledge that there is still considerable cross-sectional variation left unexplained. We have not included
what may or may not be a significant component of the decision, audit fees. We do not have access to all fee
information and therefore have not included audit fees in our model. However, a survey of CEOs of the Inc.
Magazine 500 found that a CPA firm’s proposed audit fees ranked fourth in criteria in selecting a new auditor
(Addams and Alfred 2002).
22
A second possibility is the market’s reaction is primarily related to insurance concerns and therefore, RETURN
has little power in distinguishing between follow and not-follow firms.

19
switching costs. In contrast to the no result for RETURN, this solidifies the arguments made by
Chaney and Philipich (2002) and Krishnamurthy et al. (2003) concerning the tainted nature of
AA’s reputation. At first glance it may be surprising that COMPLEX was not positive given the
significant costs associated with educating an audit team about various geographic segments.
However, there is no guarantee that the other offices that aided in the audit of the various

geographic segments also followed AA, thereby potentially limiting the cost savings from
following AA.
23

Similar to the results on TRANSPARENCY and COMPLEX, the coefficient on BLOCK
is negative and significant, indicating monitoring by outside blockholders led firms to switch
from AA. This is again consistent with the view that AA’s reputation had been tainted and
firms’ with the greatest agency issues deciding to minimize the fallout by choosing an entirely
new auditor.
The coefficient on SIZE is insignificant, which is inconsistent with both hypotheses.
Recall, we expected SIZE to be positive (negative) and significant if the switching (agency) cost
hypothesis holds. Since larger firms are more complex, it is more costly to educate the audit
team about the operations of the firm. Nevertheless, these switching costs did not subsume the
agency costs in the auditor choice decision and therefore resulted in an insignificant coefficient.
This is not consistent with the agency explanation offered in Barton (2004) who interprets the
negative relation between firm size and duration as evidence that firms with better reputations
acted to protect those reputations by switching earlier.
Turning to the switching costs results, the coefficients on EXPERT and ACCRUAL
confirm our univariate findings and illustrate switching costs were indeed a consideration in


23
Recall that many of AA’s international offices transferred their businesses to other Big4 firms immediately, rather
than waiting to determine whether AA would be convicted.

20
selecting a new auditor. The findings indicate if AA was the designated expert then clients were
more likely to follow. Further, the greater the performance adjusted discretionary accruals the
more likely firms were to follow their former AA team to a new auditor. This last finding for
ACCRUAL has the potential to be informative about the implications of mandatory auditor

rotation for financial reporting and deserves further attention. We address this issue in section
4.2 below.
In unreported results, we analyzed the standardized coefficients from model 1 in an
attempt to assess the relative importance of the variables consistent with agency costs versus
those related to switching costs. Overall, there is some support for agency concerns playing a
larger role in the follow decision compared to switching costs with the relative influence of
agency variables being 1.33 times larger than switching cost variables. However, we place
limited weight on these results since they are subject to criticisms concerning standardized logit
coefficients (for instance, see Soofi et al. 2000).

M
ANDATORY ROTATION AND ACCRUALS
Given the results in table 4 concerning the increased (decreased) probability of following
AA the more aggressive (conservative) the financial reporting, we directly analyze the
implications of changing auditors on the financial reporting behavior of AA firms. We adopt
Myers et al. [2003] model of discretionary accruals and augment it with our FOLLOW variable
and indicators for extreme ACCRUAL quintiles as follows:
εβββ
ββββ
βββα
++++
++++
+++=
∑
CASHFLOWowthIndustryGrSIZE
AGETENUREAGGRESSIVEFOLLOWAGGRESSIVE
VECONSERVATIFOLLOWVECONSERVATIFOLLOWACCRUAL
II
1098
7654

321
*
*


21
where the variables are defined as follows (Compustat data items in parentheses):
ACCRUAL
= performance adjusted discretionary total accruals
following Kothari et al. (2004);
FOLLOW
= 1 if the client followed AA, and 0 otherwise;
CONSERVATIVE
= 1 if ACCRUAL is in the lowest quintile during last year
audited by AA, 0 otherwise.
FOLLOW*CONSERVATIVE
= interaction of FOLLOW and LOW.
AGGRESSIVE
= 1 if ACCRUAL is in the highest quintile during last year
audited by AA, 0 otherwise.
FOLLOW*AGGRESSIVE
= interaction of FOLLOW and HIGH.
TENURE
= number of years audited by AA per Compustat;
AGE
= number of years for which total assets (#6) was reported in
Compustat since 1980.
SIZE
= natural logarithm of total assets (#6);
IndustryGrowth

=
∑∑
==
−
N
i
N
i
titi
SalesSales
11
1,,
/ by industry.
CASHFLOW
= cash flow from operations (#308) divided by ending total
assets (#6).
I
denotes industry as defined in Barth et al. (1998).

Results are reported in table 5 for the year prior to AA’s collapse (year t-1), the final year audited
by AA (year t) and the first year audited by the new auditor (year t+1). An important distinction
between our analysis and Myers et al.’s (2003) is that we performance adjust discretionary
accruals (our dependent variable), whereas Myers et al. (2003) does not. Given our much
smaller sample size and our control/treatment research design, we believe performance adjusted
discretionary accruals is the most appropriate measure of aggressive behavior in this context.
24

Further, we include two indicator variables, CONSERVATIVE and AGGRESSIVE that distinguish
firms in the lowest and highest quintiles of ACCRUAL as of the last year audited by AA. We
expect these extreme accrual portfolios to exhibit the most significant changes surrounding the

rotation of auditors if financial reporting quality was suspect and subsequently improved by the
rotation as suggested by proponents of mandatory auditor rotation (SEC 1994).


24
We estimated all analyses using non-performance adjusted discretionary accruals and the results are unchanged.
However, the interpretation of aggressive behavior is made easier by performance adjusting along with reducing
concerns about bias induced by extreme performance.

22
The insignificance of the FOLLOW variable suggests that the middle three quintiles of
the ACCRUAL variable are not significantly different on average than the corresponding group
for the not-follow firms. As indicated by the coefficient on CONSERVATIVE, not-follow firms
in the extreme negative ACCRUAL quintile had persistently lower discretionary accruals than the
remainder of the sample in all three years analyzed. More importantly, the FOLLOW firms do
not appear to behave differently after the change in auditors relative to the not-follow firms as
witnessed by the insignificance of the coefficient on FOLLOW*CONSERVATIVE.
Turning to the AGGRESSIVE observations, we notice these firms also exhibit some
persistence with not-follow firms being different on average in both the final year with AA (year
t) and the first year audited by the new auditor (year t+1). This suggests that the mandatory
auditor rotation did not serve to reign in this aggressive behavior. The FOLLOW firms on the
other hand are no longer significantly different on average than the middle three quintiles
implying that their aggressive behavior in the final year of AA was not repeated under the new
audit firm.
25

This is an intriguing result that at first may seem counterintuitive. For instance, it may be
reasonable to expect firms following AA to exhibit relatively more aggressive behavior, since a
primary motivation of the follow decision was presumably to maintain relations with the auditor
that originally opined on the aggressive accounting. However, given the unusual circumstances

surrounding this particular auditor rotation, the pattern of behavior exhibited in table 5 may not
be unexpected for a couple of reasons. First, AA partners moving to a new auditor may be more
likely to reign in aggressive behavior given their reduction in wealth and other disutilities from
said behavior while at AA. Second, audit firms taking on AA clients and personnel may have


25
The incremental coefficient for firms that followed and were in the high accrual quintile is
AGGRESSIVE+FOLLOW*AGGRESSIVE. Reported statistics are included in the bottom of table 5.

23
subjected the companies to increased levels of scrutiny because of the Enron and Worldcom
fiascos.
As a robustness check, we collected a control sample of firms that did not switch BigN
auditors over the three year period analyzed. We selected two match firms for each AA firm in
the sample from the same industry and ROA decile. Matches were determined by ranking firms
on ACCRUAL and selecting the firms just above and below the AA firm in question. In
unreported analyses, the results hold with only the aggressive followers exhibiting improved
behavior in the year of the new auditor. These tests support the validity of the results presented
in table 5.
Overall, we find evidence consistent with firms choosing to follow (not-follow) AA if
they had aggressive (conservative) financial reporting. We find no evidence that performance
adjusted discretionary accruals improved for the most conservative firms for both follow and
not-follow clients. Further, there is no evidence that aggressive behavior was curbed in firms
that selected an entirely new auditor. Whereas firms that were aggressive and followed AA were
less aggressive after following AA to a new auditor. Combined, this evidence does not support
the contention that mandatory auditor rotation would necessarily improve financial reporting
confirming conclusions made in Myers et al. (2003).
However, there are reasons to believe a more typical mandatory rotation may result in
outcomes different from those we find in our setting. For instance, under a typical mandatory

rotation auditors would have limited terms, which may reduce independence problems and
improve financial reporting. In the current setting, the auditors were rotated in a mandatory
fashion, but the term of the new auditor was not limited, which potentially reduces the benefits
from enhanced independence. Nevertheless, we believe the results presented are of interest to

24
standard setters, other rule making bodies, academics and practitioners in considering the costs
and benefits of mandatory auditor rotation.

ROBUSTNESS TESTS AND CAVEATS
To test the robustness of our presented results, we estimated all the models using ranks
for all non-indicator variables and obtain virtually identical results reducing concerns about the
influence of outliers. Further, we included alternative measures of agency costs such as the need
to raise external financing as suggested by Kaplan and Zingales (1997), volatility, and
institutional holdings, but none were significantly different from zero nor are any inferences
altered by their inclusion. Finally, findings do not change when AA office fixed effects are
included in our regressions.
Although the results are quite robust, we were unable to obtain information concerning
where exact personnel went along with their client lists. Instead, we relied on entire office
switches to categorize firms, which is undoubtedly noisy. In a related matter, we were unable to
accurately categorize some large AA offices such as New York and Chicago (AA’s
headquarters), which means our results may not be generalizable to all AA clients. We feel it is
important to provide these caveats, but at the same time we have no reason to believe they are
more of a concern in this setting than other settings utilized in accounting research.

V. CONCLUSION
The AA collapse presents a rare opportunity to study the determinants of auditor
selection. Ordinarily, researchers are limited to switching decisions that are created by an
auditor resignation or client dismissal, both of which are contaminated events. In the current

25