firth et al - 2012 - rotate back or not after mandatory audit partner rotation [mapr]

Rotate back or not after mandatory audit partner rotation?
Michael A. Firth
a,
⇑
, Oliver M. Rui
b
,XiWu
c
a
Department of Finance and Insurance, Lingnan University, Hong Kong
b
China Europe International Business School (CEIBS), Shanghai, China
c
School of Accountancy, Central University of Finance and Economics, China
abstract
Many countries have implemented rules that require an audit part-
ner to rotate off the audit of a specific client after a certain period of
time in the belief that rotation will improve independence and will
allow for a fresh look at the audit. The rules are either silent on
whether or when a partner can rotate back or else they specify a
cooling-off period after which the rotated-off partner can resume
the audit. Using archival data from China, a country with a 2-year
cooling-off period, this paper explores the determinants of
whether the audit partner rotates back or not when the cooling-
off period expires, and whether audit quality is weakened by the
audit partner rotation-back practice. We find that the audit partner
rotation-back practice can be explained by factors relating to
switching costs, agency conflicts, client desirability, and the audit
partner’s capacity constraint considerations. Interestingly, we find
that clients suffering greater audit adjustments immediately prior
to the expiration of the cooling-off period are more likely to be

associated with subsequent audit partner rotation-back. Further-
more, we find that rotation-back partners tend to treat former cli-
ents more favorably than non-rotation-back cases using modified
audit opinions as our proxy for audit quality. Overall, our findings
offer preliminary explanations for and shed light on the conse-
quences of rotation-back practice arising from mandatory audit
partner rotation requirements and lend support to regulatory con-
cerns on rotation-back practice among audit partners.
Ó 2012 Elsevier Inc. All rights reserved.
0278-4254/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
/>⇑
Corresponding author.
E-mail addresses: mafi (M.A. Firth), (O.M. Rui), (X. Wu).
J. Account. Public Policy 31 (2012) 356–373
Contents lists available at SciVerse ScienceDirect
J. Account. Public Policy
journal homepage: www.elsevier.com/locate/jaccpubpol
1. Introduction
Mandatory audit partner rotation is now required in many jurisdictions.
1
Rotation is seen as a po-
tential means of enhancing auditor independence and audit quality by reducing partner–client familiar-
ity and bringing in fresh perspectives.
2
However, the benefits of rotation could be lost if the previously
rotated-off audit partner rotates back to the client. In 2003, the U.S. Securities and Exchange Commission
(SEC, 2003) expressed its concern about audit partner rotation-back practice in proposing a rule designed
to strengthen auditor independence requirements:
While the [Sarbanes–Oxley] Act specified that these two partners [the lead and concurring part-
ners] were subject to rotation after five years, the Act is silent with regard to the time out period.

One approach to the partner rotation rules could have been to preclude the partner from returning
to the audit client after he or she rotates off to that engagement. The Commission is adopting
rules to require the lead and concurring partners to rotate after five years and, upon rotation, be
subject to a five-year ‘‘time out’’ period. Because of the importance of achieving a fresh look to
the independence of the audit function, we believe that a five-year time out period is appropriate
for these two partners.
Prior to the implementation of the SEC, 2003 rule, the American Institute of Certified Public
Accountants (AICPA) adopted a 2-year cooling-off professional practice standard (SEC, 2003, footnote
123). After debating various approaches including the permanent barring of audit partners from rotat-
ing back to a former client, the SEC settled on a 5-year cooling-off period for lead and concurring audit
partners, effective from May 6, 2003. Many professional entities including the AICPA argued for a
shorter cooling-off period (SEC, 2003, footnote 126). To strike a balance, the SEC (2003) requires that
partners subject to rotation requirements other than the lead and concurring partner rotate after no
more than 7 years and be subject to a 2-year time-out. Note that the names of the lead and concurring
audit partners are not disclosed. The cooling-off period is shorter than 5 years in some other jurisdic-
tions where the mandatory audit partner rotation requirement is in place. For example, a 2-year cool-
ing-off period is required in both Australia (Hamilton et al., 2005, p. 3, footnote 4) and China (China
Securities Regulatory Commission (CSRC) 2003).
Motivated by the regulatory concern about audit partner rotation-back, this study capitalizes on
the natural laboratory setting of China, a country where audit partner rotation and rotation back rules
apply, to explore what determines whether the audit partner rotates back when the cooling-off period
expires and whether audit quality is compromised when the rotated-off partner rotates back. Focusing
our investigation on China, a jurisdiction which requires that two certified public accountants sign the
published audit report, allows us to identify audit partners and their rotation status for a given client.
To examine the determinants of audit partner rotation-back, we test four hypotheses, switching
costs, agency conflicts, client desirability, and the audit partner’s capacity constraint, that have been
proposed in the literature on auditor selection and the auditor–client relationship. Utilizing data from
the mandatory audit partner rotation regime in China, we find that all of these four hypotheses can
help explain the partner rotation-back practice. Specifically, audit partners are more likely to rotate
back if there is a greater degree of partner–client familiarity and are less likely to rotate back if the

client is larger, more complex, in the high-risk category, or if the client offers a limited tenure for
future audit work. More interestingly, we find that if a client suffers greater audit adjustments to
pre-audit earnings immediately prior to the expiration of the cooling-off period, the previously
rotated-off partner is more likely to rotate back, suggesting an opportunistic switching cost consider-
ation by the client. To investigate the consequences of audit partner rotation-back, we follow prior
China-related studies (e.g., DeFond et al., 2000; Chen et al., 2010; Chan and Wu, 2011) and use the
1
Based on an international survey by the GAO (2003, Appendix V), these jurisdictions include France, Germany, Japan,
Singapore, Spain, the United Kingdom, and the United States. Australia, the Chinese mainland, and Taiwan also require audit
partner rotation.
2
There is a growing literature that examines the efficacy of audit partner rotation either from the partner-tenure perspective
(Chi and Huang 2005; Carey and Simnett 2006; Chen et al. 2008; Manry et al. 2008; Fargher et al. 2008) or from the perspective of
the immediate (i.e., the first-year) effect of mandatory partner rotation (Hamilton et al. 2005; Chi et al. 2009).
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
357
auditor’s propensity to issue modified opinions as a proxy for audit quality. We find that rotation-back
partners tend to treat former clients more favorably than non-rotation-back cases in the first post-
cooling-off year.
This study makes a number of contributions to the auditing literature. First, we are among the first
to offer explanations for the audit partner rotation-back/non-rotation-back practice. Recent literature
(e.g., Chi et al., 2009) has focused on the immediate effect of mandatory audit partner rotation. We
extend the investigation of the rotation policy to a longer window (i.e., when the mandatory cool-
ing-off period expires). Second, this study enriches our understanding of the auditor–client relation-
ship and (individual-level) auditor selection. Recent studies (Blouin et al., 2007; Chen et al., 2009)
have explored why some clients maintain the relationship with former audit partners in a forced audit
firm change setting. We add to this line of literature by utilizing another specific setting (i.e., partner
rotation-back vs. non-rotation-back), which allows us to extend our understanding of why some cli-
ents voluntarily resume the relationships with former audit partners after a mandatory cooling-off
period. Third, our findings shed light on the consequences of audit partner rotation-back, and lend

support to regulatory concerns about the rotation-back practice among audit partners. Our findings
should have implications for the policy debate on the appropriate duration of the cooling-off period.
The rest of this paper is structured as follows. Section 2 presents theoretical discussions and
hypotheses development of the determinants and the effects of audit partner rotation-back on audit
quality. Section 3 describes the institutional background in China and sample selection. Section 4 pre-
sents the research design and results of a determinant model on audit partner rotation-back. Section 5
examines the audit quality associated with the rotation-back practice and Section 6 concludes the
paper.
2. Hypotheses development
Given a specified (mandatory) cooling-off period, why do some audit partners rotate back when the
cooling-off period expires? The rotation-back practice is a voluntary manifestation of auditor–client
relationships and (individual-level) auditor selection. Such a practice involves three parties: the client,
the audit partner, and the audit firm. We base our theoretical analysis on the auditor selection and
auditor–client relationship literature. We analyze the issue from both the demand and supply sides.
The auditor–client relationship literature suggests three potential costs incurred by the client in
auditor selection, i.e., switching costs, agency costs, and implicit insurance costs
3
(Blouin et al.,
2007). The switching costs can be defined as start-up costs incurred by the client for a new audit engage-
ment (Blouin et al., 2007, p.624), which commonly include costs to educate the auditor about the com-
pany’s environment, business, and financial reporting issues.
4
An audit partner rotation invariably leads
to switching costs for the client as they have to educate the newly assigned partners. If the rotated off
partner retakes charge of the audit after the cooling off period expires, additional switching costs will
be incurred. The magnitude of these additional switching costs depends on several factors. If the rotated
off partner had developed an extensive knowledge of the client prior to rotation then the switching back
costs may be quite low if the partner retains this knowledge and experience. The cost of switching back
to the former audit partner has to be compared to any lingering switching costs of the new auditors (i.e.,
if the new auditors take more than 2 years to gain a thorough understanding of the client). The greater

the switching costs the client is likely to incur for its annual audit, the less likely the client is willing to
accept a previously rotated-off partner to rotate back. If the rotated-off partner is more experienced and
3
The implicit insurance costs hypothesis suggests that various audit firms provide different values of implicit insurance (Menon
and Williams 1994), and audit firm changes may incur costs when switching from a larger audit firm to a smaller firm. However, in
our within-firm audit partner analysis, the audit partner rotation itself does not change the implicit insurance the audit firm
provides to the client. Therefore, insurance costs are not considered in our subsequent analysis.
4
As elaborated in Blouin et al. (2007), switching costs also include costs incurred by the client in selecting a new auditor, and an
increased risk of audit failure. The former costs are more applicable to the case of audit firm switching, but less applicable to the
within-firm partner rotation. We argue that the latter costs can arise because of the different knowledge bases of the new and the
former audit partner.
358 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373
familiar with the client than the newly introduced auditors, he/she may rotate back if this will save
money for the client. Based on these cost considerations, our first hypothesis is as follows:
H1. The greater the switching cost, the less likely a previously rotated-off partner is to rotate back
when the mandatory cooling-off period expires, ceteris paribus.
Agency costs are another major type of cost consideration in auditor selection. Jensen and Meckling
(1976) defined 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 princi-
pal’s best interest’’. A high-quality auditor can reduce agency costs through the competent and inde-
pendent monitoring of the agent. Mandatory audit partner rotation has been widely implemented to
mitigate agency costs and to improve audit quality (GAO, 2003), whereas the rotation-back practice is
deemed by some regulators as compromising auditor independence (SEC, 2003). Therefore, a previ-
ously rotated-off partner is likely to be associated with lower perceived audit quality when he/she ro-
tates back. If agency costs are a major type of cost consideration in the client’s selection of individual
auditors, the client is less likely to accept a rotation-back practice, other things being equal. Jennings
et al. (2006) and Kaplan and Mauldin (2008) suggest that the strength of corporate governance plays a
role in information users’ perceptions of the effectiveness of mandatory auditor rotation in their
experimental studies. Strong corporate governance may place constraints on the close ties between

client management and the audit partner, whereas weak corporate governance would do little to cur-
tail the negative aspects of auditor–client familiarity. Therefore, we develop our second set of hypoth-
eses as follows:
H2a. The greater the agency conflict, the less likely a previously rotated-off partner is to rotate back
when the mandatory cooling-off period expires, ceteris paribus.
H2b. The stronger the corporate governance, the less likely a previously rotated-off partner is to
rotate back when the mandatory cooling-off period expires, ceteris paribus.
Besides the client (demand-side) considerations in the process of auditor selection, there are also
supply-side factors considered by the auditor. Prior literature on auditor-initiated audit firm changes
(e.g., Shu, 2000) and on audit firm portfolio management (e.g., Johnstone and Bedard, 2004) suggest
that auditors normally avoid audit clients that fall into the high-risk category (see Bedard et al.,
2008 for a review). Similarly, when the rotated-off partner considers whether to rotate back after
the cooling-off period expires he/she is less likely to do so if the client is deemed as a highly risky
one during the cooling-off period. Therefore, we propose our third hypothesis as follows:
H3. The more risky the client during the cooling-off period, the less likely a previously rotated-off
partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus.
Audit firms could also play a role in the audit partner rotation-back practice. As cited in SEC (2003),
some respondents expressed a concern about the lack of audit partners, relative to the audit firm’s
business capacity, to implement the rotation policy. This concern suggests that an auditor’s capacity
constraint can be a deterrent to the rotation-back decision. If an audit firm has fewer qualified prac-
titioners relative to the firm’s service jobs, each auditor, including the previously rotated-off auditor,
will be fully occupied by the assigned tasks and will have less capacity to rotate back.
5
Therefore, we
state our fourth hypothesis as follows:
H4. The more business revenues each of an audit firm’s practitioners earns, the less likely a previously
rotated-off partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus.
5
Besides the auditor capacity hypothesis, an alternative explanation is related to client desirability from the standpoint of an
individual audit partner. That is, when an audit firm provides more (less) business services to its partners, each partner is on

average economically less (more) dependent on a particular client, thus leading the rotated-off partner to be less (more) inclined to
rotate back.
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
359
Underlying the debate on whether and when the previously rotated-off partners are allowed to ro-
tate back (SEC, 2003), the regulator is concerned about an impaired audit quality associated with the
audit partner rotation-back. On the one hand, auditors may have incentives to maintain their own rep-
utation and audit quality (e.g., Reynolds and Francis, 2001). On the other hand, recent audit failures in
Enron and WorldCom suggest that auditors still are subject to client pressures which lead to compro-
mised audit quality. It is therefore an empirical question as to whether the audit partner rotation-back
is associated with lower audit quality when compared with the non-rotation-back practice. We thus
develop our fifth hypothesis in null form as follows:
H5. The audit partner rotation-back is associated with the same audit quality as compared to that for
the non-rotation-back practice, ceteris paribus.
3. Chinese institutional background and sample selection
To enhance auditor independence and bring in a fresh perspective, the CSRC and the Ministry of
Finance (MOF) of China jointly issued a mandatory audit partner rotation policy on October 8, 2003
which took effect from the 2003 annual audit. Specifically, the CSRC (2003) prohibits a signing auditor
from providing audit services for the same listed company for more than five consecutive years and
bars the signing auditor from resuming audit services for the entity concerned for at least 2 years after
he or she rotates off.
6
Audit reports in China record both the name of the audit firm and the names of the signing audi-
tors.
7
This allows us to use the financial auditing history of each listed Chinese company to identify
whether a signing auditor has served for five consecutive years or more. For companies that have been
listed for less than 5 years, we also check the names of the IPO signing auditors because regulations spec-
ify that the tenure for an IPO audit is three consecutive years.
8

Our sample period begins in 2003, which is the earliest year in which the regulations were in force.
The MOF made systematic changes to China’s accounting and auditing standards in 2006, and 39 (48)
newly enacted or revised accounting (auditing) standards became effective in 2007. Since these two
sets of standards are fundamental for the auditors to make accounting judgments and conduct audit
procedures, we end our sample period in 2006 to avoid the possible confounding effect of different
standard regimes on the auditing process and outcomes.
Pre-MR period
Mandatory Partner Rotation
(MR) period
Post-MR period
(Partner rotating-back or
continuous cooling-off)
PreMR(…) PreMR(-1) MR(1) MR(2) PostMR(1) PostMR(…)
Fig. 1. The timeline of mandatory partner rotation in China.
6
There is also one exception whereby if both of the signing auditors have provided audit services for the same entity for the
same period of five consecutive years, one of them is allowed to extend his or her term as a signing auditor of the entity concerned
by a maximum of 1 year.
7
According to the relevant Chinese regulations (e.g., The Ministry of Finance’s Notification on Issues Regarding Certified Public
Accountants’ Signing and Stamping on the Audit Report, effective from July 2, 2001), and in normal situations, one of the two signing
auditors must serve as the lead auditor in charge of field work and the other must be at least a deputy executive of the CPA firm
and serve as the reviewer of the engagement. These two signing auditors are required to assume the same legal liabilities as each
other (unless one can prove the contrary). We use the term ‘‘audit partner’’ to describe the signing auditor even if the audit firm is
not a partnership (e.g., if it is a limited liability company).
8
The CSRC (2003) requires that for companies that have recently been listed or are awaiting a listing, signing auditors must not
provide audit services for the same initial public offering (IPO) entity for more than two consecutive years after the IPO. This 2-year
period follows on from the 3 years for which audited financial statements must be included in the IPO applicant’s filing, thus giving
partner tenure of 5 years. This requirement is consistent with the (US) SEC Office of the Chief Accountant’s interpretation that the

signing auditor’s term under rotation requirements includes the number of years for which audited financial statements are
included in the IPO filing (SEC 2004, ‘‘Audit Partner and Partner Rotation’’ section, question 3).
360 M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
To help illustrate our sample construction, we show in Fig. 1 the mandatory rotation timeline,
which is partitioned into the pre-rotation period (PreMR), the mandatory partner rotation period
(MR), and the post-mandatory-rotation period (PostMR). For partners required to rotate in 2003,
the first mandatory rotation or cooling-off year (MR(1)) is 2003, the second cooling-off year
(MR(2)) is 2004, and the first post-cooling-off year (PostMR(1)) is 2005. In a similar fashion, partners
who are required to rotate in 2004 have a cooling-off period of 2004 and 2005 and the first post-cool-
ing-off period is 2006. In the PostMR period, the former audit partner may resume the audit (i.e., part-
ner rotation-back) or may not (i.e., continuous cooling-off).
Table 1 presents the sample selection process (Panel A) and composition of rotation-back status
(Panel B). We first identify 382 A-share listed Chinese companies whose audit partner was subject
to mandatory rotation in 2003 or 2004. We eliminate 61 observations where both audit partners
should have rotated off in MR(1) but one of them used the exemption article to remain on the audit
for one additional year. We then exclude 22 observations where the audit partners should have ro-
tated off for 2 years but rotated back in MR(2) (i.e., 1 year earlier), as they represent suspected cases
of superficial partner rotation (Bamber and Bamber, 2009).
9
Another five observations are excluded be-
cause the other audit partner should have rotated off in MR(2) or PostMR(1) due to meeting the 5-year
rotation threshold but did not do so. 44 companies changed their audit firms in MR(2) or PostMR(1), and
therefore are excluded from the sample. Two companies are further excluded because they were de-
listed in PostMR(1). Our final sample consists of 248 companies with an observable status of partner
rotation-back or non-rotation-back in the first post-cooling-off year. Panel B shows that for 115
(46.4%) out of 248 companies, the previously rotated-off audit partners rotate back when the 2-year cool-
ing-off period expires (RB clients hereafter), whereas for the other 133 companies (53.6%), the previously
rotated-off audit partners do not rotate back (NRB clients). Untabulated statistics show that RB clients
and NRB clients have similar industry profiles. The frequency in each industry is representative of the
overall distribution.

Table 1
Sample selection and composition.
MR(1) = 2003 MR(1) = 2004 Total
N
Panel A: Sample selection
Companies that underwent mandatory partner rotation in MR(1) 162 220 382
Less
Both audit partners should have rotated off in MR(1) but one of them used the
exemption article to remain
17 44 61
Audit partners should have rotated off for 2 years but rotated back in MR(2) 9 13 22
The other partner should have rotated off in MR(2) but did not 2 1 3
The other partner should have rotated off in PostMR(1) but did not 0 2 2
Audit firm changed in MR(2) 7 17 24
Audit firm changed in PostMR(1) 7 13 20
The client was delisted in PostMR(1) 1 1 2
Final sample 119 129 248
Rotation-back Continuous cooling-off Total N
Number Percent Number Percent Number Percent
Panel B: Sample composition based on rotation-back status
PostMR(1) 115 46.4 133 53.6 248 100
MR(1) = the first mandatory rotation (cooling-off) year
MR(2) = the second mandatory rotation (cooling-off) year
PostMR(1) = the first post-mandatory rotation year
9
These companies clearly violated the specified cooling-off requirement and we treat 1-year rotations off as cases of superficial
rotation.
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
361
4. Determinants of audit partner rotation back: design and results

4.1. Empirical design
To test the four hypotheses (switching costs, agency conflicts, client desirability, and audit partner’s
capacity consideration) relating to the possible determinants of audit partner rotation-back vs. non-
rotation-back practice, we construct the following probit model:
probitðRotBack ¼ 1Þ¼b
0
þ b
1
LTA þ b
2
SqSubs þ b
3
IPOPtn þ b
4
PtnTenr þ b
5
jAudAdjj
MR2
þ b
6
StaOwn þ b
7
BODSize þ b
8
IndDir þ b
9
HighRisk
þ b
10
Re

v
nPerCPA þ Yr2006 þ IndDum þ
e
ð1Þ
We set the dependent variable, RotBack, as a dummy variable coded 1 if an observation belongs to the
group of clients whose audit partners rotated back when the 2-year cooling-off period expired. As
shown in Table 2, we identify a number of experimental variables to test our hypotheses.
Simunic (1980) suggests that client size and complexity have the most significant effects on the
cost of an audit. Moreover, client size and complexity are also used as proxies for agency concerns,
as larger companies are subject to greater reputation costs and more complex companies are more
Table 2
Variables used in audit partner rotation-back determinant model: hypotheses and sign predictions.
Dep. Var. = RotBack Hypotheses
Switching costs Agency conflicts Client desirability Capacity consideration
Experimental variables
LTA ÀÀ
SqSubs ÀÀ
IPOPtn + À
PtnTenr + À
|AudAdj|
MR2
+ À
StaOwn ÀÀ
BODSize +
IndDir À
HighRisk À
RevnPerCPA ÀÀ
Control variables
Yr2006 ?? ? ?
IndDum ?? ? ?

Variable definitions:
RotBack: a dummy variable coded 1 for a rotation-back client and 0 otherwise;
LTA: the natural logarithm of total assets (in 10 thousand RMB Yuan);
SqSubs: the square root of the number of consolidated subsidiaries;
IPOPtn: a dummy variable coded 1 if the rotated-off audit partner served as the initial public offering (IPO) auditor for the client
and 0 otherwise;
PtnTenr: the number of years for which the audit partner has served as the signing auditor for the client when he or she has to
rotate off in line with the mandatory requirement;
|AudAdj|
MR2
: the absolute value of AudAdj (audit adjustment to pre-audit earnings) made in the second mandatory rotation
(cooling-off) year, where AudAdj is equal to (post-audit earnings À pre-audit earnings)/total assets;
StaOwn: a dummy variable coded 1 if the client is ultimately controlled by the government or a state-owned enterprise and 0
otherwise;
BODSize: the natural log of the number of directors on the board;
IndDir: the proportion of independent directors on the board;
HighRisk: a dummy variable coded 1 if the client suffers a loss, receives a modified audit opinion, or is subject to a regulatory
sanction in any of the 3 years prior to the first post-cooling-off year and 0 otherwise;
RevnPerCPA: the natural log (total business revenues of the audit firm divided by the number of certified public accountants of
the firm);
Yr2006: a dummy variable coded 1 for a 2006 observation and 0 for a 2005 observation;
IndDum: industry dummy variables based on the CSRC classification.
362 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373
difficult to monitor (Blouin et al., 2007). Both the switching cost and agency cost considerations pre-
dict that client size and complexity are negatively associated with the likelihood of audit partner rota-
tion-back. We use LTA, defined as the natural logarithm of total assets, as a proxy for client size and
SqSubs, denoted as the square root of the number of consolidated subsidiaries, as a proxy for client
complexity.
10
We expect the coefficients on LTA and SqSubs to be negative.

We also use IPOPtn and PtnTenr as proxies for switching costs. IPOPtn is coded 1 if the rotated-off
audit partner is the IPO signing auditor and 0 otherwise. PtnTenr is the audit partner tenure since the
IPO, i.e., the number of years for which the audit partner has served as the signing auditor for the client
when he or she has to rotate off in line with the mandatory requirement. Both variables capture the
rotated-off partner’s client-specific experience. Therefore, if the rotated-off partner has served the cli-
ent as the IPO signing auditor, or has audited the client for a longer time, the switching costs associ-
ated with rotation-back should be lower and we expect a positive sign for both IPOPtn and PtnTenr.On
the other hand, IPOPtn and PtnTenr also measure a close partner–client relationship. An IPO audit is
associated with the long process prior to a successful IPO in which the audit partner and the client
work closely together to float the company on the stock exchange. The financial packaging that is com-
mon during Chinese IPOs (Aharony et al., 2000) also suggests that the engagement audit partners may
well be familiar with and acquiescent to the client’s earnings management or accounting preferences.
As to the post-IPO partner tenure, for the first year of rotation (2003), the affected partners had served
for at least 3 years (which may include the 3 years of the IPO period), although some had served for up
to 12 years. As familiarity can be a threat to the independence of the previously rotated-off partner,
11
there is a greater demand for a high-quality audit. Therefore, from the concerns about agency conflicts,
IPOPtn and PtnTenr are expected to be negatively associated with audit partner rotation-back.
Our final measure of switching costs is |AudAdj|
MR2
, which is the absolute value of audit adjustments
to pre-audit earnings made in the second mandatory rotation (cooling-off) year. AudAdj is defined as
(post-audit earnings À pre-audit earnings)/total assets). DeFond and Subramanyam (1998) suggests
that clients change auditors due to overly conservative auditing treatment. Following this logic, if
the newly assigned auditors treat the client in a conservative manner, the client is more inclined to
have the previously rotated-off partner back when the cooling-off period expires. Therefore, we expect
the conservatism of auditing treatment in the second cooling-off year is positively associated with
audit partner rotation-back. Audit adjustments to pre-audit earnings represent how strict auditors
are in curbing clients’ earnings management or misstatements attempts (e.g., Wright and Wright,
1997; Nelson, 2009, p.16) and can thus be used as a direct measure of auditor conservatism. Specifi-

cally, pre-audit earnings represent the manager’s reporting preference, whereas the absolute value
of audit adjustments (i.e., the divergence between pre- and post-audit earnings) can be viewed as
the amount of misstatement corrected by the auditor. The larger the absolute value of audit adjust-
ments, the higher the level of the auditor’s effort and independence in terms of withstanding the pres-
sure from the manager and making corrections to the manager’s preferred earnings, which include both
income-increasing and income-decreasing misstatements. We obtain proprietary data on audit adjust-
ments from the Chinese Institute of Certified Public Accountants (CICPA). The CICPA has required audit
firms to file pre-audit and post-audit earnings for publicly listed companies since 2001.
12
The coeffi-
cient on |AudAdj|
MR2
is expected to be positive from the switching costs perspective. However, a greater
magnitude of absolute audit adjustments also suggests a higher level of agency conflicts, which may lead
to a greater demand for monitoring and a lower likelihood of audit partner rotation-back. Thus the coef-
ficient on |AudAdj|
MR2
is expected to be negative based on agency cost arguments.
We use StaOwn as a proxy for agency conflict. It is coded 1 if the client is controlled by a state-
owned entity including a government agency and 0 otherwise. Prior China-related studies (e.g., Wang
et al., 2008) suggest agency conflicts between the State shareholders and minority shareholders. Such
an agency concern increases the demand for high-quality monitoring and decreases the likelihood of
10
We do not use the number of segments in our research design due to the lack of consistent disclosure practices among most
companies in our sample period. A formal Chinese accounting standard on segment reporting took effect from January 1, 2007.
11
The SEC (2003) is concerned that if a lengthy partner-client relationship has been established before the mandatory partner
rotation, the rotated-off audit partner may be more likely to rotate back (and the client may request a rotation back).
12
We only have access to the CICPA audit adjustment data for the 2001–2005 annual audits.

M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
363
an audit partner rotation-back. The unique institutional background in China also makes the StaOwn
variable a useful proxy for client desirability. The government sometimes forces listed companies con-
trolled by the state to change auditor to be consistent with the non-listed parent company’s auditor
choice (SASAC, 2004).
13
Thus, audit firm tenure (and thus partner tenure) quite often is less assured
Table 3
Descriptive statistics of main variables used in audit partner rotation-back determinant model.
Continuous
variables
RB group (RotBack =1)
(n = 115)
NRB group (RotBack =0)
(n = 133)
Test of differences
Mean (Median) Mean (Median) t-statistic (Mann–Whitney
Z)
LTA 12.118 12.455 À2.853
***
(12.053) (12.402) (À2.582
***
)
SqSubs 2.588 3.044 À2.144
**
(2.449) (2.828) (À1.734
*
)
PtnTenr 5.191 5.380 À0.631

(5) (5) (À0.204)
|AudAdj|
MR2
#
0.0117 0.0068 2.270
**
(0.0044) (0.0026) (1.998
**
)
BODSize 2.251 2.217 1.270
(2.197) (2.197) (1.516)
IndDir 0.332 0.345 À1.689
*
(0.333) (0.333) (À1.383)
RevnPerCPA
##
3.798 3.961 À2.613
***
(3.752) (3.797) (À1.909
*
)
Categorical
variables
Number (Percent) Number (Percent) Pearson Chi-square
IPOPtn 67 65 2.184
(58.3) (48.9)
StaOwn 67 101 8.821
***
(58.3) (75.9)
HighRisk 25 39 1.853

(21.7) (29.3)
Variable definitions:
RotBack: a dummy variable coded 1 for a rotation-back client and 0 otherwise;
LTA: the natural logarithm of total assets (in 10 thousand RMB Yuan);
SqSubs: the square root of the number of consolidated subsidiaries;
PtnTenr: the number of years for which the audit partner has served as the signing auditor for the client when he or she has to
rotate off in line with the mandatory requirement;
|AudAdj|
MR2
: the absolute value of AudAdj (audit adjustment to pre-audit earnings) made in the second mandatory rotation
(cooling-off) year, where AudAdj is equal to (post-audit earnings À pre-audit earnings)/total assets;
BODSize: the natural log of the number of directors on the board;
IndDir: the proportion of independent directors on the board;
RevnPerCPA: the natural log (total business revenues of the audit firm divided by the number of certified public accountants of
the firm);
IPOPtn: a dummy variable coded 1 if the rotated-off audit partner served as the initial public offering (IPO) auditor for the client
and 0 otherwise;
StaOwn: a dummy variable coded 1 if the client is ultimately controlled by the government or a state-owned enterprise and 0
otherwise;
HighRisk: a dummy variable coded 1 if the client suffers a loss, receives a modified audit opinion, or is subject to a regulatory
sanction in any of the 3 years prior to the first post-cooling-off year and 0 otherwise.
*
Significance at the 0.10 level (two-tailed).
**
Significance at the 0.05 level (two-tailed).
***
Significance at the 0.01 level (two-tailed).
#
The number of observations in RB (NRB) group with available values for |AudAdj|
MR2

is 111 (124).
##
The number of observations in RB (NRB) group with available values for RevnPerCPA is 111 (127).
13
In 2004, the State-owned Assets Supervision and Administration Commission (SASAC) mandated that a list of major Chinese
state-owned enterprises (SOEs) rotate their audit firms with new firms. This policy forced a few listed companies (as subsidiaries of
their parent SOEs) to rotate their audit firms. In 2005, the SASAC further required that each SOE under its jurisdiction rotate its
audit firm after serving as auditor for five consecutive years.
364 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373
for these companies. This may make rotated-off partners less willing to rotate back as they face the
uncertainty of losing the audit again if there is a switch of audit firm due to state ownership.
14
Therefore,
both agency conflict and client desirability considerations suggest a negative sign for StaOwn.
As discussed earlier, audit partner rotation back could arguably be perceived as one form of
very close auditor–client ties. Companies with stronger corporate governance may thus be less
willing to accept a switch back to the rotated-off partner. We use BODSize and IndDir as proxies
for the corporate governance of a company. BODSize is defined as the natural log of the number
of directors on the board, while IndDir is denoted as the proportion of independent directors on
the board. Yermack (1996) and Vafeas (2000) suggest that larger boards tend to be less effective
than smaller boards. This is possibly due to larger boards being fragmented and less likely to reach
agreements among board members. Since 2003, the Chinese regulators have formally required
companies to have independent non-executive directors, whose primary aim is to serve the inter-
ests of the minority shareholders (Firth et al., 2007). We expect a positive (negative) sign for BOD-
Size (IndDir).
We use HighRisk as a proxy for client desirability. If a client falls into the high-risk category during
the cooling-off period, the rotated-off audit partner is less motivated to rotate back. We classify the
client as highly risky (HighRisk = 1) if the client suffers a loss, receives a modified audit opinion, or
is subject to a regulatory sanction in any of the 3 years prior to the first post-cooling-off year. We ex-
pect the coefficient on HighRisk to be negative.

We use RevnPerCPA to capture the potential impact of audit firm characteristics on the capacity of
the rotated-off partner to rotate back.
15
It is defined as the natural log (total business revenues of the
audit firm divided by the number of certified public accountants of the firm). The more service revenues
the audit firm can provide for each of its auditors to earn in the first post-cooling-off year, the more likely
the workload capacity of an individual auditor (including the rotated-off auditor) will be saturated.
Moreover, as the audit firm has provided each auditor with more economic incentives from other
sources, an individual auditor is less incentivized to rotate back. Therefore, from both the view of auditor
capacity constraint and that of client desirability, we expect RevnPerCPA to be negatively associated with
audit partner rotation-back.
Finally, we include Yr2006 and IndDum to control for year and industry fixed effects. Yr2006 is
coded 1 for a 2006 observation and 0 for a 2005 observation, while IndDum is industry dummy vari-
ables based on the CSRC first-digit industry classification (using the manufacturing industry as the
base group).
4.2. Descriptive statistics and univariate analysis
Table 3 provides descriptive statistics for the main variables used in model (1) for the RB and NRB
clients, respectively. Among the RB (NRB) group, the rotated-off partners serve as the IPO signing audi-
tor for 58.3% (48.9%) of the companies. 75.9% of the companies within the NRB group are controlled by
the State, whereas a significantly smaller percent (58.3%) within the RB group are State-controlled.
29.3% (21.7%) of companies within the NRB (RB) group fall into the high-risk category. The mean |Au-
dAdj|
MR2
is 0.0117 for the RB group, which is significantly greater than the magnitude (i.e., 0.0068) for
the NRB group (p < 0.05).
16
This univariate result is consistent with the client trying to avoid stricter
auditing treatments, which we treat as one type of switching cost, but not with the agency concern argu-
ment. The mean RevnPerCPA is 3.798 for the RB group, which is significantly smaller (p < 0.01) than the
level for the NRB group (3.961). This result is consistent with the audit partner’s capacity and client

desirability considerations. Our univariate results also show that NRB clients are significantly larger
14
Feedback from our interviews with CPA firms corroborates that state-controlled companies are more likely to switch audit
firms for regulatory reasons.
15
Since 2002, the CICPA began to publicize the total service revenues and the total number of certified public accountants for
each of the Chinese top-100 audit firms based on total service revenues.
16
The mean raw value of audit adjustments (AudAdj
MR2
)isÀ0.0071 (À0.0056) for the RB (NRB) clients. This suggests that
auditors on average make downward audit adjustments to pre-audit earnings in the second cooling-off year, which is consistent
with the notion that listed clients generally attempt to overstate earnings.
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
365
Table 4
Probit regression results of audit partner rotation-back determinant model.
Dep. Var. = RotBack Expected sign Coefficient
(z-statistic)
(1) (2) (3) (4)
LTA ÀÀ0.311 À0.165 À0.263 À0.128
(À3.02)
***
(À1.50)
*
(À2.32)
***
(À1.07)
SqSubs ÀÀ0.074 À0.110 À0.069 À0.104
(À1.22) (À1.68)

**
(À1.11) (À1.54)
*
IPOPtn ± 0.515 0.425 0.491 0.419
(2.30)
**
(1.79)
*
(2.18)
**
(1.75)
*
PtnTenr ± 0.125 0.146 0.136 0.169
(2.12)
**
(2.31)
**
(2.29)
**
(2.69)
***
|AudAdj|
MR2
± 17.122 17.912
(2.78)
***
(2.66)
***
StaOwn ÀÀ0.498 À0.591 À0.520 À0.630
(À2.65)

***
(À3.02)
***
(À2.69)
***
(À3.16)
***
BODSize + 0.926 0.781 0.953 0.708
(2.14)
**
(1.71)
**
(2.15)
**
(1.53)
*
IndDir ÀÀ1.444 À3.112 À1.414 À2.699
(À0.98) (À1.58)
*
(À0.94) (À1.47)
*
HighRisk ÀÀ0.572 À0.792 À0.620 À0.847
(À2.69)
***
(À3.36)
***
(À2.92)
***
(À3.57)
***

RevnPerCPA ÀÀ0.403 À0.529
(À1.92)
**
(À2.06)
**
Yr2006 ? À0.073 À0.104 À0.041 À0.101
(À0.43) (À0.57) (À0.23) (À0.54)
IndDum ? Yes Yes Yes Yes
Constant ? 1.946 1.136 2.804 2.627
(1.39) (0.78) (1.83)
*
(1.49)
Model Chi-square 35.76
***
40.96
***
37.87
***
42.16
***
Pseudo R
2
0.12 0.15 0.13 0.16
N 248 235 238 225
Correct classification 69.4% 68.5% 68.5% 67.6%
Variable definitions:
RotBack: dummy variable coded 1 for a rotation-back client and 0 otherwise;
LTA: the natural logarithm of total assets (in 10 thousand RMB Yuan);
SqSubs: the square root of the number of consolidated subsidiaries;
IPOPtn: a dummy variable coded 1 if the rotated-off audit partner served as the initial public offering (IPO) auditor for the client

and 0 otherwise;
PtnTenr: the number of years for which the audit partner has served as the signing auditor for the client when he or she has to
rotate off in line with the mandatory requirement;
|AudAdj|
MR2
: the absolute value of AudAdj (audit adjustment to pre-audit earnings) made in the second mandatory rotation
(cooling-off) year, where AudAdj is equal to (post-audit earnings À pre-audit earnings)/total assets;
StaOwn: a dummy variable coded 1 if the client is ultimately controlled by the government or a state-owned enterprise and 0
otherwise;
BODSize: the natural log of the number of directors on the board;
IndDir: the proportion of independent directors on the board;
HighRisk: a dummy variable coded 1 if the client suffers a loss, receives a modified audit opinion, or is subject to a regulatory
sanction in any of the 3 years prior to the first post-cooling-off year and 0 otherwise;
RevnPerCPA: the natural log (total business revenues of the audit firm divided by the number of certified public accountants of
the firm);
Yr2006: a dummy variable coded 1 for a 2006 observation and 0 for a 2005 observation;
IndDum: industry dummy variables based on the CSRC classification.
As there are missing values for |AudAdj|
MR2
and RevnPerCPA, we first regress model (1) without these two variables in column (1)
(n = 248). We then add |AudAdj|
MR2
in column (2) (n = 235), and add RevnPerCPA in column (3) (n = 238), and include both
variables in column (4) (n = 225).
*
Significance at the 0.10 level (two-tailed for variables with bi-directional signs, and one-tailed for variables with one
directional sign).
**
Significance at the 0.05 level (two-tailed for variables with bi-directional signs, and one-tailed for variables with one
directional sign).

***
Significance at the 0.01 level (two-tailed for variables with bi-directional signs, and one-tailed for variables with one
directional sign).
366 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373
and more complex than their RB counterparts. This is consistent with both the switching cost and agency
cost considerations.
4.3. Regression results
Table 4 presents the probit regression results of model (1). As there are missing values for |Au-
dAdj|
MR2
and RevnPerCPA, we first regress model (1) without these two variables in column (1)
(n = 248). We then add |AudAdj|
MR2
in column (2) (n = 235), and add RevnPerCPA in column (3)
(n = 238), and include both variables in column (4) (n = 225). In all four regressions, the chi-square
is significant (p < 0.01) and the model has an overall correct classification rate ranging between
67.6% and 69.4%. The regression results are reported using standard errors corrected for heteroskedas-
ticity (White, 1980). We present the two-tailed significance level for variables with a prediction of
bi-directional signs, and the one-tailed significance level for variables with one predicted directional
sign.
The regressions yield a number of interesting results. First, the coefficients on LTA are consistently
negative in the four regressions and three of them are significant at the 10% level or better. We also
find that the coefficients on SqSubs are consistently negative in four regressions, and two are signifi-
cant at the 10% level or better in the one-tailed test. The results suggest that larger clients and more
complex clients are less likely to be associated with audit partner rotation-back, which is consistent
with both the switching cost hypothesis (H1) as well as the agency cost hypothesis (H2a).
Second, the coefficients on IPOPtn are consistently positive and significant at the 10% level or better
(two-tailed) in all four regressions. Moreover, the coefficients on PtnTenr are significantly positive
(p < 0.05 or better, two-tailed) in the four regressions. The results for partner-client familiarity are also
economically meaningful. For example, based on the full sample regression (n = 248, column (1)), a

partner who was also the IPO partner (IPOPtn = 1) is 20.2% more likely to rotate back at the end of
the cooling-off period than a partner who was not the signing partner for the IPO audit (IPOPtn = 0).
Furthermore, one additional year of partner tenure (PtnTenr) increases the probability of rotating back
by 5.0%. These results indicate that the rotated-off partner is more likely to rotate back if he was the
client’s IPO signing auditor or has had long auditor tenure since the IPO, suggesting that the switching
cost considerations dominate the agency concerns. The evidence is consistent with H1, but not with
H2a.
Third, the coefficients on |AudAdj|
MR2
are significantly positive (p < 0.01, two-tailed) in columns (2)
and (4). This indicates that the stricter the auditors treat the client in making earnings adjustments in
the second cooling-off year, the more likely the previously rotated-off partner is to rotate back in the
first post-cooling-off year. The evidence shows that the opportunistic switching cost considerations
dominate the agency conflict concerns, and support H1 but not H2a.
Fourth, the coefficients on StaOwn are significantly negative (p < 0.01, one-tailed) in all four regres-
sions. This indicates that an audit partner is less likely to rotate back if the client is controlled by the
state. The results are consistent with both the agency conflict concerns (H2a) as well as the client
desirability considerations (H3). Two other variables related to agency concerns (BODSize and IndDir)
also show the expected directional signs. Specifically, BODSize has a significantly positive coefficient
(p < 0.05 in three regressions, and <0.10 in one regression, one-tailed). This suggests that the ro-
tated-off partner is more likely to rotate back to clients when there is a large board of directors. IndDir
has a negative coefficient (p < 0.10 in two among four regressions, one-tailed), indicating that the pro-
portion of independent directors on the board has some marginal effect in deterring the audit partner
rotation-back. The results on BODSize and IndDir lend some support to H2b.
Fifth, the coefficients on HighRisk are significantly negative (p < 0.01, one-tailed) in the four regres-
sions. This indicates that a rotated-off partner is less likely to rotate back if the client falls into the
high-risk category during the cooling-off period. The results are consistent with the client desirability
considerations (H3).
Finally, the coefficients on RevnPerCPA are significantly negative (p < 0.05, one-tailed) in columns
(3) and (4). This suggests that the more service revenues per auditor an audit firm has, the less likely

the previously rotated-off partner is to rotate back in the first post-cooling-off year. The evidence is
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
367
consistent with both the audit partner capacity constraint (H4) as well as the client desirability con-
siderations (H3).
In robustness checks, we consider the possible effect of absolute audit adjustments made in the
first mandatory rotation year (|AudAdj|
MR1
) on the partner rotation-back practice. We find the coeffi-
cient on |AudAdj|
MR1
is positive but not significant (p > 0.79, two-tailed), whereas the coefficient on
|AudAdj|
MR2
is still significantly positive (p < 0.01, two-tailed). These results suggest that the manda-
tory 2-year cooling-off requirement makes the rotation-back decision more sensitive to the auditor
conservatism level immediately prior to the audit partner rotation-back decision making.
In addition to the proportion of independent directors considered here, there are other corporate
governance mechanisms in China, such as the supervisory board and the audit committee (Firth
et al., 2007). During our sample period, the Chinese regulators have encouraged, but not required, a
listed firm to establish an audit committee. In our sensitivity tests, we consider the effect of the super-
visory board size and an audit committee on the incidence of audit partner rotation back. Neither of
these two mechanisms shows significant results, whereas our main results are qualitatively unaffected.
Lastly, to control for potentially omitted correlated variables, we also include the financial leverage,
profitability (operating income divided by total assets), and listing age of the client in the model. None
of coefficients on these variables is significantly different from zero. Our main results on experimental
variables are qualitatively unchanged.
Overall, our empirical analysis documents that the audit partner rotation-back practice can be ex-
plained by the factors relating to switching costs, agency conflicts, client desirability, and an audit
partner’s capacity constraint considerations. Our results are in general consistent with Blouin et al.

(2007) as they find that both the client’s switching cost and agency conflict considerations can explain
the maintenance of partner–client relationships.
17
Some of our findings suggest that audit partner rota-
tion-back/non-rotation-back can be attributed to rational decisions (e.g., client size, complexity, and an
audit partner’s working capacity) or an auditor’s risk-averse behavior (e.g., client risk profile). However,
we also find that audit partners are more likely to rotate back if there is a greater degree of partner-client
familiarity or if a client suffers greater audit adjustments to pre-audit earnings. These opportunistic
incentives associated with audit partner rotation-back, which are also consistent with prior studies
(Blouin et al., 2007; Chen et al., 2009) that find that clients with greater magnitudes of earnings manage-
ment are more likely to follow their former audit partners, should concern the regulator.
5. Audit quality associated with audit partner rotation-back
5.1. Research design
In this section we further examine whether there is a compromise in audit quality when the ro-
tated-off partner rotates back. To empirically test H5, we make a direct comparison of the audit quality
in the first post-cooling-off year of RB and NRB clients.
Modified audit opinions are an important manifestation of different audit outcomes and variation
in audit quality (Francis, 2004). Prior China-related studies have commonly used the auditor’s propen-
sity to issue modified opinions as a proxy for audit quality (DeFond et al., 2000; Chan et al., 2006,
2010; Wang et al., 2008; Chan and Wu, 2011). Following this line of research, we use a modified audit
opinion (MAO) as our proxy for audit quality.
18
We estimate the following logistic regression model:
logitpðMAO ¼ 1Þ¼b
0
þ b
1
RotBack þ b
2
LTA þ b

3
LEV þ b
4
RECV þ b
5
OPROA þ b
6
Loss
þ b
7
PreMAO þ b
8
LocTop10 þ b
9
StaOwn þ b
10
ListAge þ b
11
IMR
þ Yr2006 þ IndDum þ
e
ð2Þ
where the variables are defined as follows:
MAO is a dummy variable coded 1 for a modified audit opinion and 0 otherwise; RotBack is a dum-
my variable coded 1 for a rotation-back client and 0 otherwise; LTA is the natural logarithm of total
17
However, the auditor’s supply-side considerations are less discussed in Blouin et al. (2007) due to their specific research
setting (i.e., forced audit firm change).
18
MAOs include unqualified opinions with explanatory notes, qualified, disclaimer and adverse opinions.

368 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373
Table 5
Modified audit opinions associated with audit partner rotation-back.
Dep. Var. = MAO Coefficient
(z-statistic)
First-stage regression based
on the full sample (Table 4 Column 1)
First-stage regression based on a complete
set of variables (Table 4 column 4)
Experiment variable
RotBack À2.101 À2.344
(À2.59)
***
(À2.66)
***
Control variables
LTA 0.298 0.321
(0.75) (0.68)
LEV À0.982 À1.409
(À0.36) (À0.48)
RECV 5.687 3.933
(1.27) (0.93)
OPROA À7.912 À6.581
(À1.00) (À0.77)
Loss 3.509 3.705
(1.71)
*
(1.75)
*
MB À2.196 À2.123

(À1.34) (À0.98)
RET À3.021 À2.419
(À1.92)
*
(À1.58)
PreMAO 7.072 6.764
(4.78)
***
(5.23)
***
LocTop10 À2.488 À2.106
(À2.97)
***
(À2.51)
**
StaOwn À3.233 À3.063
(À2.00)
**
(À1.80)
*
ListAge À0.047 À0.027
(À0.34) (À0.20)
IMR 2.159 1.849
(1.87)
*
(2.08)
**
Yr2006 0.300 0.023
(0.24) (0.02)
IndDum Yes Yes

Constant À4.248 À4.289
(À0.63) (À0.57)
Model Chi-square 55.98
***
73.86
***
Pseudo-R
2
0.652 0.652
N 248 225
Variable definitions:
MAO: dummy variable coded 1 for a modified audit opinion and 0 otherwise;
RotBack: a dummy variable coded 1 for a rotation-back client and 0 otherwise;
LTA: the natural logarithm of total assets (in 10 thousand RMB Yuan);
LEV: total liabilities divided by total assets;
RECV: accounts receivable divided by total assets;
OPROA: operating income divided by total assets;
Loss: a dummy variable coded 1 if the firm reports a net loss and 0 otherwise;
MB: market value to total assets ratio;
RET: the company’s annual stock return;
PreMAO: a dummy variable coded 1 if the client received a modified audit opinion in the previous year and 0 otherwise;
LocTop10: a dummy variable coded 1 if the audit firm is one of the top 10 domestic auditors based on the total assets of listed
clients and 0 otherwise;
StaOwn: a dummy variable coded 1 if the client is ultimately controlled by the government or a state-owned enterprise and 0
otherwise;
ListAge: the number of years since the initial public offering (IPO);
IMR: the inverse Mills ratio computed from the first-stage rotation-back determinant model;
Yr2006: a dummy variable coded 1 for a 2006 observation and 0 for a 2005 observation;
IndDum: industry dummy variables based on the CSRC classification.
*

Significance at the 0.10 level (two-tailed).
**
Significance at the 0.05 level (two-tailed).
***
Significance at the 0.01 level (two-tailed).
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
369
assets (in 10 thousand RMB Yuan); LEV is the total liabilities divided by total assets; RECV is the ac-
counts receivable divided by total assets; OPROA is the operating income divided by total assets; Loss
is a dummy variable coded 1 if the firm reports a net loss and 0 otherwise; MB is the market value to
total assets ratio; RET is the company’s annual stock return; PreMAO is a dummy variable coded 1 if
the client received a modified audit opinion in the previous year and 0 otherwise; LocTop10 is a dum-
my variable coded 1 if the audit firm is one of the top 10 domestic auditors based on the total assets of
listed clients and 0 otherwise; StaOwn is a dummy variable coded 1 if the client is ultimately con-
trolled by the government or a state-owned enterprise and 0 otherwise; ListAge is the number of years
since the initial public offering (IPO); IMR is the inverse Mills ratio computed from the first-stage rota-
tion-back determinant model; Yr2006 is a dummy variable coded 1 for a 2006 observation and 0 for a
2005 observation and IndDum are industry dummy variables based on the CSRC first-digit industry
classification (using the manufacturing industry as the base group).
Our experimental variable in model (2) is RotBack, which is coded 1 for a rotation-back client. We
include control variables based on prior China-related studies (e.g., DeFond et al., 2000; Chen et al.,
2001, 2010; Wang et al., 2008). Client size (LTA) is expected to be negatively associated with the fre-
quency of receiving a modified audit opinion because larger companies tend to be financially health-
ier. We use the total liabilities to total assets ratio (LEV) and accounts receivable to total assets ratio
(RECV) as the proxies for financial conditions and risks. We expect that LEV and RECV are positively
associated with the frequency of receiving modified audit opinions. A more profitable company is less
likely to be in financial distress, so we expect OPROA, defined as operating income divided by total as-
sets, to have a negative coefficient and the variable LOSS to have a positive coefficient. We also include
the market-to-book ratio (MB) to capture a client’s growth opportunities (Rajan and Zingales, 1995),
and the company’s annual stock return (RET) to capture how stock market variables affect audit opin-

ions (Dopuch et al., 1987). To control for the effect of receiving a modified audit opinion in the prior
year, we include PreMAO in the model (Monroe and Teh, 1993). We follow DeFond et al. (2000) and
include a large local audit firm dummy variable LocTop10 which is coded 1 if the audit firm is one
of the top 10 domestic auditors based on the total assets of listed clients and 0 otherwise.
19
Given that
government-controlled companies are more likely to be economically and/or politically protected by
central or local governments from business risks and operating uncertainties, auditors may be less likely
to issue modified audit opinions to these clients (Chan et al., 2006; Wang et al., 2008). Therefore, we in-
clude StaOwn in our model and expect its coefficient to be negative. We also include the variable ListAge
to control for the listing age of a company (Chen et al., 2001).
We acknowledge that an endogeneity issue arises as the rotation-back practice is not randomly
determined (as shown in Section 4). For example, audit partners are less likely to rotate back to highly
risky clients, which would decrease the likelihood of the issuance of an MAO. To control for the poten-
tial selection bias, we adopt a Heckman two-stage approach (Heckman, 1979). The inverse Mills ratio
(IMR) is computed from the first-stage rotating-back determinant model. In the second stage, we in-
clude IMR as an additional control variable in estimating model (2). Finally, we include both year and
industry dummies to control for year- and industry-fixed effects.
5.2. Empirical results
Among our sample (n = 248), 13 (5.2%) companies receive a modified audit opinion in the first post-
cooling-off year. For the RB clients, 6.1% received a prior-year MAO, whereas 4.3% receive a current-year
MAO. For the NRB clients, 4.5% received a prior-year MAO, whereas 6.0% receive a current-year MAO.
Table 5 reports the multivariate regression results of the modified audit opinion model. We obtain
qualitatively similar results when the first-stage regression results are based on any of the four
columns of Table 4, and we present results with the first-stage regression based on the full sample
(n = 248) and based on a reduced sample (n = 225, with all the variables in model (1) included). Table 5
19
Prior research generally finds that brand name audit firms play a greater monitoring role in conducting an audit (Becker et al.
1998; Francis et al. 1999). However, the Chinese auditing market is characterized by a low level of Big 4 auditor participation. In
our sample used for model (2) (n = 248), there are only 12 observations audited by Big 4, and none of them receive modified audit

opinions. Therefore, we do not include a Big 4 auditor dummy in model (2). Untabulated statistics show that these 12 companies
are significantly larger, less leveraged, and more profitable than companies audited by non-Big 4 auditors.
370 M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
shows that the coefficient on RotBack is significantly negative (p < 0.01, two-tailed) in both regres-
sions. This suggests that rotating-back audit partners are less likely to issue a modified audit opinion
than non-rotating-back cases in the first post-cooling-off year, after controlling for the client charac-
teristics and the self-selection bias (IMR is significant at the 10% or 5% level). This is also consistent
with our previous finding that clients suffering stricter audit adjustments in the second cooling-off
year are more likely to rotate back. The results in Table 5 support the regulatory concerns about
the impaired audit quality associated with audit partner rotation-back.
With respect to control variables, we find that auditors are more likely to issue modified opinions
to clients who received a modified audit opinion in the prior year (PreMAO) and the loss-making cli-
ents (Loss). LocTop10 shows a significantly negative coefficient in both regressions (p < 0.01). This sug-
gests that large local audit firms are less likely to issue MAOs than smaller local audit firms. However,
untabulated descriptive statistics suggest that large local audit firms have similar client size, financial
leverage, and profitability to smaller local firms. This raises questions about the audit quality of large
local audit firms. The coefficients on StaOwn are negative in both regressions and significant at the 10%
or 5% level, suggesting that auditors are less likely to issue a MAO to clients controlled by the State.
This is consistent with the results reported by Wang et al. (2008). RET shows a negative sign in both
regressions and significant at the 10% level in one regression. This indicates that companies with high
stock returns are less likely to receive a modified audit opinion.
6. Conclusion
Mandatory audit partner rotation is now required in many jurisdictions. However, regulators have
expressed concerns about the practice whereby audit partners rotate back to their former clients. A
short cooling-off period may be insufficient to sever the close ties between the rotated-off partner
and the client. Drawing on prior auditor selection and auditor–client relationship literatures, and uti-
lizing data from the mandatory audit partner rotation regime in China, we test a number of hypotheses
(i.e., switching costs, agency conflicts, client desirability, and audit partner’s capacity constraint) that
likely explain the audit partner rotation-back or non-rotation-back practice when a 2-year cooling-
off period expires. We find that all of these four hypotheses can help explain the partner rotation-back

practice. Specifically, audit partners are more likely to rotate back if there is a greater degree of partner–
client familiarity and are less likely to rotate back if the client is larger, more complex, in the high-risk
category, or if the client offers a limited tenure for future audit work. Moreinterestingly, we find that if a
client suffers stricter audit adjustments to pre-audit earnings immediately prior to the expiration of the
cooling-off period, the previously rotated-off partner is more likely to rotate back, suggesting an oppor-
tunistic switching cost consideration by the client. Consistent with this finding, we also document that
partners who rotate back to their former clients treat them more favorably than non-rotating-back
cases using modified audit opinions as the proxy for audit quality. Overall, our evidence is consistent
with the regulatory concerns about the incentives and consequences of audit partner rotation-back.
Our findings have implications for regulators in China and other jurisdictions where mandatory
audit partner rotation requirements are in place. As discussed earlier, the SEC (2003) has imposed a
longer cooling-off period (5 years) than was previously the case. Based on our evidence, a short cool-
ing-off period is associated with opportunistic rotation-back incentives and impaired audit quality in
the first rotation-back year. Therefore, a short cooling-off period seems to be less justified than a long-
er one, because the longer an audit partner is rotated off, the greater the switching-back costs are, and
the close ties between the rotated-off audit partner and the client are severed to a greater extent. A
long cooling-off period may also increase the incentives of the newly introduced audit partner to
expend more effort on the client from as early as the beginning of the cooling-off period. Regulators
in jurisdictions with relatively short cooling-off periods may thus wish to consider implementing
longer cooling-off periods.
20
20
A caveat in our discussion about whether a longer cooling-off period is better than a shorter one is that we are not able to
directly and empirically compare different regimes with different specified cooling-off periods. Therefore, we cannot draw any
strong inferences regarding whether the audit quality for a client whose audit partner rotates back will improve as the cooling-off
period increases beyond 2 years.
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373
371
The results of this study are based on archival data from China. Given that non-Big 4 auditors are
the main players in the Chinese auditing market, our evidence is more generalizable to markets where

Big 4 auditors are less dominant.
21
We are aware that there are other notable differences in the insti-
tutional background and auditing regime for auditor rotations between various jurisdictions (e.g., a part-
ner-visible auditing regime in China and Australia vs. a partner-invisible auditing regime in most
markets including the United States; a single leading partner regime vs. a two leading partner regime).
Readers should thus exercise caution in generalizing our evidence to other jurisdictions with mandatory
audit partner rotation practices. Moreover, we have the same difficulty in distinguishing between the
leading engagement partner from the reviewing partner as that encountered in prior related studies
(e.g., Chen et al., 2008; Chi et al., 2009). Nevertheless, our study supplements the growing literature that
focuses on partner-level auditor independence issues (e.g., Carey and Simnett, 2006; Blouin et al., 2007;
Chen et al., 2008; Chi et al., 2009). More specifically, we contribute to the auditor tenure and mandatory
rotation debate and related literature by employing a wider lens to examine audit partner rotation
arrangements within audit firms and showing that partner rotation-back arrangements reinforce the
regulatory concern about close partner–client relationships.
Acknowledgements
We thank Martin Loeb (the editor), two anonymous reviewers, Charles J.P. Chen, Marcus Doxey
(discussant at 2011 AAA annual meeting), Roger Simnett, Xijia Su, and participants at the 17th annual
International Symposium on Audit Research (June 2011, Canada) for helpful comments and sugges-
tions. We thank the Chinese Institute of Certified Public Accountants (CICPA) for proprietary data
items. Michael Firth and Oliver Rui acknowledge support from the Research Grants Council of the
HKSAR (LU340308) and Oliver Rui acknowledges support from the Research Grants Council of the
HKSAR (CU450009). Xi Wu acknowledges financial support from the National Natural Science Foun-
dation of China (No. 70602020), the Program for Innovation Research at the Central University of Fi-
nance and Economics (CUFE), the New Century Scholar Program of China’s Ministry of Education
(NCET-11-0754), the Beijing Municipal Commission of Education ‘‘Joint Construction Project’’, and
the ‘‘Project 211’’ Fund of CUFE. Part of this study was carried out when Xi Wu worked for the Chinese
Institute of Certified Public Accountants (CICPA). However, the views expressed in this paper do not
reflect the official views of the CICPA.
References

Aharony, J., Lee, C-W.J., Wong, T.J., 2000. Financial packaging of IPO firms in China. Journal of Accounting Research 38 (1), 103–
126.
Bamber, E.M., Bamber, L.S., 2009. Discussion of Discretionary accruals, audit-firm tenure and audit-partner tenure: empirical
evidence from Taiwan. Contemporary Accounting Research 26 (2), 393–402.
Becker, C.L., DeFond, M.L., Jiambalvo, J., Subramanyam, K.R., 1998. The effect of audit quality on earnings management.
Contemporary Accounting Research 15 (1), 1–24.
Bedard, J.C., Deis, D.R., Curtis, M.B., Jenkins, J.G., 2008. Risk monitoring and control in audit firms: a research synthesis. Auditing:
A Journal of Practice & Theory 27 (1), 187–218.
Blouin, J., Grein, B., Roundtree, B., 2007. An analysis of forced auditor change: the case of former Arthur Andersen clients. The
Accounting Review 82 (3), 621–650.
Carey, P., Simnett, R., 2006. Audit partner tenure and audit quality. The Accounting Review 81 (3), 653–676.
Chan, K.H., Wu, D., 2011. Aggregate quasi rents and auditor independence: evidence from audit firm mergers in China.
Contemporary Accounting Research 28 (1), 175–213.
Chan, K.H., Lin, K.Z., Mo, P., 2006. A political-economic analysis of auditor reporting and auditor switches. Review of Accounting
Studies 11 (1), 21–48.
21
For example, our (untabulated) supplementary analysis shows that in the first mandatory partner rotation year, Chinese listed
companies are more likely to receive modified audit opinions as compared to the year prior to the mandatory rotation, suggesting
an immediate positive effect of mandatory partner rotation on audit quality in China Mainland. This finding is not consistent with
the results in Chi et al. (2009) who fail to find a significant improvement in audit quality after mandatory partner rotation in
Taiwan. However, we note that China Mainland is very different from Taiwan in terms of its audit market structure, where the
latter’s audit market is much more dominated by Big 4 auditors. The differences in audit market structure make it difficult to
reconcile our findings with those of Chi et al. (2009).
372 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373
Chen, C.J.P., Chen, S., Su, X., 2001. Profitability regulation, earnings management, and modified audit opinions: evidence from
China. Auditing: A Journal of Practice & Theory 20 (2), 9–30.
Chen, C.J.P., Su, X., Wu, X., 2009. Forced audit firm change, continued partner-client relationship, and financial reporting quality.
Auditing: A Journal of Practice & Theory 28 (2), 227–246.
Chen, C Y., Lin, C J., Lin, Y C., 2008. Audit partner tenure, audit firm tenure, and discretionary accruals: does long auditor
tenure impair earnings quality? Contemporary Accounting Research 25 (2), 415–445.

Chen, S., Sun, S.Y.J., Wu, D., 2010. Client importance, institutional improvements, and audit quality in China: an office and
individual auditor level analysis. The Accounting Review 85 (1), 127–158.
Chi, W., Huang, H., 2005. Discretionary accruals, audit-firm tenure and audit-partner tenure: empirical evidence from Taiwan.
Journal of Contemporary Accounting and Economics 1 (1), 65–92.
Chi, W., Huang, H., Liao, Y., Xie, H., 2009. Mandatory audit partner rotation, audit quality, and market perception: evidence from
Taiwan. Contemporary Accounting Research 26 (2), 359–391.
China Securities Regulatory Commission (CSRC), 2003. Q&As to the Regulation of Mandatory Signing-Auditor Rotation.
DeFond, M., Subramanyam, K.R., 1998. Auditor changes and discretionary accruals. Journal of Accounting and Economics 25 (1),
35–68.
DeFond, M.L., Wong, T.J., Li, S., 2000. The impact of improved auditor independence on audit market concentration in China.
Journal of Accounting and Economics 28 (3), 269–305.
Dopuch, N., Holthausem, R., Leftwich, R., 1987. Predicting audit qualifications with financial and market variables. The
Accounting Review 62 (3), 431–454.
Fargher, N., Lee, H Y., Mande, V., 2008. The effect of audit partner tenure on client managers’ accounting discretion. Managerial
Auditing Journal 23 (2), 161–186.
Firth, M., Fung, P.M.Y., Rui, O.M., 2007. Ownership, two-tier board structure, and the informativeness of earnings: evidence from
China. Journal of Accounting and Public Policy 26 (4), 463–496.
Francis, J.R., 2004. What do we know about audit quality? The British Accounting Review 36 (4), 345–368.
Francis, J.R., Maydew, E.L., Sparks, H.C., 1999. The role of Big 6 auditors in the credible reporting of accruals. Auditing: A Journal
of Practice & Theory 18 (2), 17–34.
General Accounting Office (GAO), 2003. Public accounting firms: required study on the potential effects of mandatory audit firm
rotation. Report to the Senate Committee on Banking, Housing, and Urban Affairs and the House Committee on Financial
Services, Washington, DC.
Hamilton, J., Ruddock, C., Stokes, D., Taylor, S.L. 2005. Audit Partner Rotation, Earnings Quality and Earnings Conservatism.
Working Paper. Available at SSRN. < />Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47 (1), 153–161.
Jennings, M.M., Pany, K.J., Reckers, P.M.J., 2006. Strong corporate governance and audit firm rotation: effects on judges’
independence perceptions and litigation judgments. Accounting Horizons 20 (3), 253–270.
Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: managerial behavior, agency costs, and ownership structure. Journal of
Financial Economics 3 (October), 305–360.
Johnstone, K.M., Bedard, J.C., 2004. Audit firm portfolio management decisions. Journal of Accounting Research 42 (4), 659–690.

Kaplan, S.E., Mauldin, E.G., 2008. Auditor rotation and the appearance of independence: evidence from non-professional
investors. Journal of Accounting and Public Policy 27 (2), 177–192.
Manry, D.L., Mock, T.J., Turner, J.L., 2008. Does increased audit partner tenure reduce audit quality? Journal of Accounting,
Auditing and Finance 23 (4), 553–572.
Menon, K., Williams, D., 1994. The insurance hypothesis and market prices. The Accounting Review 69 (2), 327–342.
Monroe, G.S., Teh, S.T., 1993. Predicting uncertainty and audit qualifications in Australia using publicly available information.
Accounting & Finance 33 (2), 79–106.
Nelson, M.W., 2009. A model and literature review of professional skepticism in auditing. Auditing: A Journal of Practice &
Theory 28 (2), 1–34.
Rajan, R.G., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. The Journal of
Finance 50 (5), 1421–1460.
Reynolds, J.K., Francis, J., 2001. Does size matter? The influence of large clients on office-level auditor reporting decisions.
Journal of Accounting and Economics 30 (3), 375–400.
Securities and Exchange Commission (SEC), January 28, 2003. Strengthening the Commission’s Requirements Regarding Auditor
Independence. Final Rule Rel. No. 33-8183, File No.: S7-49-02. < />Securities and Exchange Commission (SEC), December 13, 2004. Office of the Chief Accountant: Application of the Commission’s
Rules on Auditor Independence – Frequently Asked Questions. < />121304.htm>.
Shu, S.Z., 2000. Auditor resignations: clientele effects and legal liability. Journal of Accounting and Economics 29 (2), 173–205.
Simunic, D., 1980. The pricing of audit services: theory and evidence. Journal of Accounting Research 18 (1), 161–190.
State-owned Assets Supervision and Administration Commission of the State Council of China (SASAC), February 5, 2004. Rules
on Financial Audit for State-Owned Enterprises. Guozifa Pingjia File No. 173.
Vafeas, N., 2000. Board structure and the informativeness of earnings. Journal of Accounting and Public Policy 19 (2), 139–160.
Wang, Q., Wong, T.J., Xia, L.J., 2008. State ownership, the institutional environment, and auditor choice: evidence from China.
Journal of Accounting and Economics 46 (1), 112–134.
White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity.
Econometrica 48 (4), 817–838.
Wright, A., Wright, S., 1997. An examination of factors affecting the decision to waive audit adjustments. Journal of Accounting,
Auditing and Finance 12 (1), 15–36.
Yermack, D., 1996. Higher market valuation of companies with a small board of directors. Journal of Financial Economics 40 (2),
185–211.
M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373

373