yazawa - 2014 - does long audit partner tenure decrease earnings quality - evidence from japan [mapr]

1


Does Long Audit Partner Tenure Decrease Earnings Quality?
Evidence from Japan


Kenichi Yazawa
1

School of Business Administration,
Aoyama Gakuin University

Current Version
March 2014


Acknowledgement: I would like to thank Roger Simnett, Gary Monroe, Peter Roebuck, Elizabeth Carson,
and other participants in the 18th annual International Symposium on Audit Research for their valuable
feedback on earlier versions of this paper. I also would like to Tsuyoshi Numagami, Hironori Fukukawa,
Mikiharu Noma, Soo-Joon Chae, Yuan Zhang and other participants in the Hitotsubashi G-COE Research
Workshop on Innovation and Management.


1
Address: 4-4-25 Sibuya, Shibuya-ku, Tokyo 150-8366, Japan
Tel./Fax: +81 (3) 3409-6272 Email address:
2

Does Long Audit Partner Tenure Decrease Earnings Quality? Evidence from Japan
Abstract:

This study examines the relationship between audit partner tenure and audit quality by
using data on 7,567 Japanese listed companies from 1999 to 2006. First, I find that the
absolute and positive values of discretionary accruals decrease significantly with increased
length of audit partner tenure. These findings are not consistent with the argument that
earnings quality decreases with audit partner tenure, especially when income-increasing
earnings management are the primary concern. Second, I find that positive discretionary
accruals decrease significantly with audit partner tenure only after the tenure exceeds seven
years. Given that half of the companies in the sample for this study have more than 50
years of business experience, these findings indicate that relatively long-term engagement
is required to effectively audit a company with a longer history. Third, I find that the
impact of tenure on earnings quality disappeared after the introduction of a periodic partner
rotation rule in Japan in 2003. This is consistent with the argument that the purpose of audit
partner rotation is to break up personal relationships between auditors and their clients and
introduce fresh approaches to audits. In addition, the results of my sensitivity checks reveal
the same results as those of the main tests. These findings contribute to the literature on
auditor tenure and earnings quality.

JEL Classification: G14; M41; M42

Key Words: Audit partner tenure; Audit partner rotation; Audit quality; Earnings
management

Data Availability: The data used in this paper are publicly available from the sources
identified herein.



3

Does Long Audit Partner Tenure Decrease Earnings Quality? Evidence from Japan

1. Introduction
In many countries around the world, periodic audit partner rotation is mandatory for
businesses. An implicit assumption in a policy of mandatory audit partner rotation is that
long relationships between audit partners and clients could lead to overfamiliarity and
reduce earnings quality. However, this assumption has not been systematically tested in the
literature, due to the lack of audit partner information in US and major country’s audit
reports. In addition, some existing empirical research, which investigate firms in countries
that practice the audit partner to sign the audit reports, has not provided consistent evidence
as to whether long audit partner tenure reduces earnings quality (Carey and Simnett, 2006;
Chen, Lin, and Lin, 2008; Manry, Mock, and Turner 2008; Chi et al., 2009).
In addition, the maximum audit-partner tenure varies considerably in each country. In the
1970s, the U.S. was the first to require audit-partner rotation after seven years of tenure for
Securities and Exchange Commission (SEC)-registered clients. The Sarbanes–Oxley Act of
2002 (SOX) further strengthened this requirement by mandating a five-year rotation for the
lead and concurring partners (SOX Sec.203). By the end of June 2008, all 27 Member
States of the European Union (EU) were required to enact the revised 8th Directive’s
(2006) requirements into national law. One important detail of the Directive is key
4

audit-partner(s) rotation for seven years (2006/43/EC Article 42 .2), but each Member State
has the discretion to decide how long the tenure with the audit client should be. For
example, the UK rule requires rotation of the lead audit-engagement partner after five years
and of other engagement partners after seven years. Germany requires seven years, France
requires six years, and Greece requires four years. These facts suggest that regulation of
audit partner rotation is not well underpinned by empirical findings.
Regardless of the effectiveness of the audit partner rotation rule has not verified
sufficiently, the European Union (EU) is currently scheduling the introduction of a new
rotation rule, which mandates that European listed companies tender their audit firms (not
audit partners) every 10 years and rotate auditors every 20 years. The rule will be subject to

the approval of the European Parliament, which will be determined in April 2014.
2
This
rule is based on the assumption that the negative impact caused by familiarity between
auditors and clients is not effectively eliminated by changes of audit partner (EC green
paper 2010).
3
In contrast, in the US, Public Company Accounting Oversight Board
(PCAOB) Chairman James Doty told the Securities and Exchange Commission (SEC) that
“we don’t have an active project or work going on within the board to move forward on a

2
Economia, 17 December 2013, Journal of Accountancy, 24 January 2014.
3
An EC green paper (2010) says that “situations where a company has appointed the same audit firm for
decades seem incompatible with desirable standards of independence. Even when ‘key audit partners’ are
regularly rotated as currently mandated by the Directive, the threat of familiarity persists.”
5

term limit for auditors.”
4
There are political factors behind the different approaches of the
EU and the US, but the differences have primarily resulted from the fact that no clear
explanation of the economic consequences of auditor rotation exist, even for audit partners.
In other words, the issue of audit partner rotation is still an open one.
As mentioned above, the amount of existing empirical evidence on this topic is very
limited. The reason for this is that many countries do not require the signature of an audit
partner in audit reports, although countries such as Taiwan and Australia are exceptions.
Japan is unique in that it requires that the signature of the audit partner who is in charge
appear in audit reports. In addition, Japan introduced an audit partner rotation rule similar

to those of the EU and US in 2003.
5
These features make the Japanese audit market a
suitable one for examining the relationship between audit partner tenure and audit quality,
as well as the effects of audit partner rotation rules.
In this study, I examine the relationship between audit partner tenure and audit quality by
using data on 7,567 Japanese listed companies from 1999 to 2006, in order to verify the
impact of long audit partner tenure on earnings quality and the economic consequence of
the introduction of the periodic partner rotation rule in Japan. First, I find that the absolute

4
CFO journal, 6 February 2014.
5
As explained later, Japanese rules require all listed firms to rotate audit-engagement partners after seven
years, and require clients of large audit firms to rotate lead engagement partners after five years. A “large
audit firm” is defined as an audit firm that has more than 100 clients. The large audit firms are the Japanese
Big 4 audit firms, and included Azusa, Shin-Nihon, Tomastu, and ChuoAoyama at the time of this study.
6

and positive values of discretionary accruals decrease significantly with increases in the
length of audit partner tenure. These findings are contrary to the argument that earnings
quality decreases with the length of audit partner tenure, especially when
income-increasing earnings management attempts are the primary concern. Second, I find
that positive discretionary accruals decrease significantly with audit partner tenure only
after the tenure in question exceeds seven years. As half of the businesses in the sample
used in this study have more than 50 years of business history, these findings indicate that
relatively long-term engagement is required to effectively audit a company with a longer
history. Third, I find that the impact of tenure on earnings quality disappeared after the
introduction of the periodic rotation rule in Japan in 2003. This is consistent with the
argument that the purpose of audit partner rotation is to break up personal relationships

between audit partners and their clients, and to add fresh perspective to audits. In addition,
these results are robust, as revealed by the results of sensitivity checks.
The findings of this study contribute to the literature on auditor tenure and earnings
quality by providing evidence at the audit partner level. Interestingly, the finding of this
study are essentially the same as those of Chen, Lin, and Lin (2008) and Chi et al. (2009),
whose studies examine Taiwanese companies, while these are not consistent with the
results of Carey and Simnett (2006) and Hamilton et al. (2005), whose studies examine
Australian companies, While doing so can prevent the generalization of findings, it is
7

important to consider the specific business practices of countries with regard to how they
affect the impact of audit tenure on earnings quality. Japan and Taiwan belong to the Asian
economic zone and have a deep historical relationship with one another. It is possible that
the fact that Japanese companies prefer stable long-term relationships affects relationships
between clients and audit partners. If regulators regard the tenures of audit firms as
influencing earnings quality but consider the tenures of audit partners to have no effects, a
reasonable explanation for this approach is needed.
The rest of this paper is organized as follows. In Section 2, I discuss the institutional
background of audit partner rotation in Japan and propose hypotheses. In Section 3, I
explain the methodology used to test these hypotheses. In Section 4, I describe the sample
used in this study and discuss the results. Finally, I provide a summary and conclusion in
Section 5.

2. Background and Hypotheses
Audit market and institutional background in Japan
Japan has roughly 3,500 listed companies and Tokyo Stock Exchange is the third-largest
capital market in the world after NYSE and NASDAQ in the US. A securities and
exchange law like that of the United States was introduced in Japan in the 1950s, and the
capital market developed over the next several decades. As in the audit markets of the US
8

and the EU, the innumerable small audit firms that initially operated in Japan have
integrated into the four largest audit firms in the country (the Big 4) over a period of
several decades. By the 2000s, the Big 4 in Japan held a market share of 80% of all
accounts. The Japanese Big 4 have partnerships with the global Big 4 accounting firms, and
use audit standards that are in accordance with the International Standards on Auditing
(ISAs), which are issued by the International Auditing Assurance Standards Board
(IAASB) of the International Federation of Accountants (IFAC).
Japan introduced a periodic audit partner rotation rule similar to those of the EU and US
in 2003 by the amendment of the Certified Public Accountant Law.
6
However, the rule did
not come into force until March 31, 2010, because partners’ tenure was not counted
retroactively. To address this problem, the Big 4 audit firms and the Japanese Institution of
Certified Public Accountants (JICPA) announced on September 13 2005 that all Big 4
firms would rotate audit partners over seven years until March 2007 as a means of
self-regulation.
Figure 1 shows transitions in audit partner tenure in the Japanese Big 4 and small to
medium-sized audit firms (non-Big 4) from 1999 to 2006. The audit partner tenure for the
non-Big 4 firms has remained stable at roughly 17 years. On the other hand, the audit
partner tenure for the Big 4 firms was stable at 12 years until 2002, began to decrease in

6
Certified Public Accountant Act, 24(3)
9

2003, and had reduced to six years by 2006, which is half the tenure of only four years
priorago. As mentioned above, this is due to the fact that because the Big 4 have introduced
a partner rotation rulesystem as a means of self-regulation from in 2003. In this study, in
order to analyze the relationship between long tenure and the audit quality, the sample is

has been corrected from 1999 on. In addition, in order to analyze how this relationship
between them has changed with the introduction of rotation rules, the period for analysis
period has been extended until 2006. Therefore, the sample period covers eight years, from
1999 to 2006, which is divided into the two periods of 1999-2002 and 2003-2006.
[Insert Figure 1 about here]

Theory and prior research
Audit quality is defined as the joint probability that an auditor will both detect material
statements and prevent or report the misstatements (DeAngelo, 1981). According to this
definition, audit quality is a function of the auditor’s ability to detect material
misstatements and report those detected misstatements. To detect material misstatements,
auditors must have specific knowledge and skills (auditor expertise). To report detected
misstatements, auditors must be independent in both fact and appearance (auditor
independence).
The debate on auditor rotation has always centered on a key question. What makes audits
10

more effective: a fresh pair of eyes (independence effect) or deep knowledge of the ins and
outs of a complex company (expertise effect)? With an increase in audit-partner tenure, the
relationship between the auditor and her or his client might introduce a so-called
“familiarity threat” to auditor independence (IFAC Code of Ethics ED 2008, para.100.13).
7

This threat could result in a decline in the audit partner’s ability to judge the company’s
performance accurately. On the other hand, audit quality could be considered to increase
with auditor tenure, because of the accumulation of client-specific information and
knowledge. Thus, the net effect of auditor tenure on earnings quality is determined by the
magnitude of the independence and expertise effects. However, only a few studies examine
the impact of audit partner tenure and rotation on the earnings quality. The reason for this
is that audit reports in most major countries record only the auditing firm’s signature, and

not the audit-engagement partner’s name. Therefore, prior research has focused on a
limited number of countries in which this is not the case, such as Australia, Germany, and
Taiwan. In addition, the findings of that research have been mixed.
Carey and Simnett (2006) investigate the relationship between audit partner tenure and
earnings quality by using 1,021 Australian companies listed in 1995. They find no evidence
of a relationship between audit partner tenure and abnormal working capital accruals. Chen,

7
The IFAC Code of Ethics ED states that a “familiarity threat occurs when, by virtue of a close relationship
with an assurance client, its directors, officers or employees, a firm or a member of the assurance team
becomes too sympathetic to the client’s interests (IFAC Code of Ethics ED 2008: 18).”
11

Lin, and Lin (2008) use a sample of 5,213 firm-years between 1990 and 2001 from
Taiwanese companies. They find that the absolute and positive values of discretionary
accruals decrease significantly with audit partner tenure. These findings are not consistent
with the argument that earnings quality decreases with audit partner tenure. By focusing on
a specific client, Manry, Mock, and Turner (2008) examine whether there is a relationship
between audit partner tenure and reduced earnings quality. They find evidence that for
small companies, regardless of their level of engagement risk, earnings quality may
actually increase as audit partner tenure increases.
Hamilton et al. (2005) use a sample of 3,621 firm-years between 1998 and 2003 from
Australian-domiciled companies, and report that among these companies, audit-partner
changes that are most likely to reflect partner rotation (when the company is not due for a
switch of audit firm) are associated with lower signed discretionary accruals. They
conclude that this result is consistent with the more conservative reporting following an
auditor rotation, and furthermore, find that this explanation is supported by evidence that
suggests that a significant increase in the asymmetrically timed recognition of economic
losses occurs when firms change audit partners. Chi et al. (2009) investigate whether the
mandatory audit-partner rotation regime implemented in 2004 in Taiwan has had a positive

influence on audit quality. They find that, in 2004, audit quality was higher for companies
that were subject to the mandatory rotation regime than for firms that were not. The authors
12

use absolute and signed discretionary accruals and abnormal working capital accruals as
proxies for audit quality. Lindscheid, Pott, and Wartin (2009) use a sample of 457
firm-years from German companies and investigate the effects of both audit-engagement
and review-partner switches. They find evidence that earnings increase more in the case of
audit review-partner switches. However, they do not find an effect on any accrual measures
when audit-engagement partners change.

Hypotheses
As described above, the net effect of audit partner tenure on earnings quality can be
theoretically determined depending on the magnitude of both the expertise effect and the
entrenchment effect. However, consistent results regarding this matter have not been
obtained by prior researchers. As such, it is difficult to construct a hypothesis only from
empirical and theoretical perspectives. Therefore, in this study, I create hypotheses by also
applying an institutional point of view. This approach is reasonable because one of the
purposes of this study is to investigate the economic consequences of regulation. The
implicit assumption of the audit partner rotation regulations that have been introduced in
developed countries is that the long tenure of an audit partner has a negative effect on audit
quality. Thus, a basic hypothesis can be created, which states that earnings quality
decreases with audit partner tenure (H1).
13

The maximum tenure of audit partners is set in the periodic audit partner rotation rules of
major countries. This assumes that the positive effects of accumulating knowledge and
experience will be greater than the negative effects of familiarity until a certain number of
years have passed, after which the negative effects of familiarity become greater than the
positive effects. In other words, it can be said that audit partner tenure and earnings quality

do not have a linear relationship to each other, and may be in a non-linear relationship that
is convex at a certain level. Thus, a second hypothesis can be put forward, which states that
earnings quality increases (decreases) with audit partner tenure up to (beyond) a certain
level (H2).
The number of years needed for a positive relationship to become a negative one must
also be determined. Rotation rules in many countries set the turning point in this
relationship at five or seven years. As mentioned above, Japanese rule requires five year
rotation for lead audit partners. It is similar to the US, which requires a five-year rotation
for the lead audit partners. In addition, the revised 8th Directive’s of EU (2006) requires
seven year rotation for the key audit partners. Based on the points discussed above, this
study sets the turning point at five years. On the other hand, Chen et al. (2008), who
analyze the audit partner tenures of Taiwanese companies, define long tenures as those
lasting more than five or 10 years. So, it needs to be checked for sensitivity.
As mentioned above, periodic partner rotation requirements were introduced in Japan in
14

2003, as a means of self-regulation. New audit partners not only make independent
judgments, but can also improve audit quality with fresh perspectives. If longer audit
partner tenures compromise auditor independence, auditors with long tenures may fail to
prevent clients from pursuing aggressive accounting policies. If so, audit partner rotation
will enhance auditor independence, and could control client aggressiveness. Thus, the third
hypothesis in this study is that the negative effects of tenure are (not) observed before
(after) the introduction of rotation rules (H3).

3. Research Design
Main regression model
I examine the impact of audit-engagement partner tenure on audit quality by using the
regression model below:
𝐴𝐴 = 𝛽
!

+ 𝛽
!
𝑇𝑁𝑅𝐸 + 𝛽
!
𝐴𝑆𝑆𝐸𝑇 + 𝛽
!
𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽
!
𝐶𝐹𝑂 + 𝛽
!
𝐴𝐺𝐸 + 𝛽
!
𝐶𝐻𝐺_𝐹𝑆𝑇
+ 𝛽
!
𝐶𝐻𝐺_𝐿𝑆𝑇 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦&𝑌𝑒𝑎𝑟_𝑒𝑓𝑓𝑒𝑐𝑡 + 𝜀
!
(1)


Audit quality
Prior research has examined the impact of audit-related factors on audit quality by testing
whether the audit-related factors are systematically associated with earnings quality, after
controlling for other factors that may affect earnings quality. Based on prior research, I use
15

a cross-sectional version of a modified Jones (1995) model (Dechow, Sloan, and Sweeney,
1995) to estimate discretionary accruals. I control for the company’s lagged rate of return
on assets (ROA) and compute discretionary accruals (DA) as follows:
𝐴𝐴 = 𝑇𝐴 − 𝑎 + 𝛽

!
𝛥𝑅𝐸𝑉 − 𝛥𝑅𝐸𝐶 + 𝛽
!
𝑃𝑃𝐸 + 𝛽
!
𝑅𝑂𝐴
!!!
!!(2)
where
TA
=
Earnings before extraordinary items - net cash flow from operations
ΔREV
=
Change in net sales
ΔREC
=
Change in net accounts receivable
PPE
=
Net property, plant, and equipment
ROA
t-1

=
Net profit
t-1
divided by total assets
t-2

TA, REV, and PPE are scaled by lagged total assets

The coefficients α, β
1
, β
2
, and β
3
, are those gained by estimating the following model by
industry-year, as is consistent with DeFond and Jiambalvo (1994) and Subramanyam
(1996). Prior research has found that the discretionary accruals of Big 4 audit firms’ clients
are smaller than those of non-Big 4 clients (Becker et al., 1998; Francis and Krishnan,
1999; Francis and Wang, 2008). This means that Big 4 auditors employ conservative
accounting policies, and so are likely to detect income-increasing earnings management
and request appropriate adjustments from clients (Nelson et al., 2002). In Japan, Yazawa
(2010) has examined the relationship between auditor size and discretionary accruals by
using data on 7,127 firm years among Japanese listed companies from 2004 to 2007. He
finds that the discretionary accruals of Big 4 auditors are smaller than those of non-Big 4
16

auditors. This study analyzes the relationship between partner tenure and discretionary
accruals by using absolute and signed value in a sub-sample dividing the negative and
positive values of DA. In any case, It is possible to interpret that where DA is small, the
quality of earnings is high.

Tenure of audit partner
In Japan, two or three partners typically sign each audit report, which means that several
audit partners influence and have responsibility for the audit. Based on the idea that length

of tenure would have a negative effect on the audit quality, it can be assumed that an audit
partner with the longest tenure has the greatest impact over the audit judgment. Thus, this
study defines TNRE as the natural logarithm of the longest audit partner's tenure. If a
longer tenure compromises auditor independence, then an auditor with long tenure may
lose the ability to combat the aggressive accounting policies of a client. As a result,
discretionary accruals and auditor’s tenure are shown to have a positive correlation. In
addition, this study examines the effect of other alternative audit partner tenure measures in
its sensitivity check.

Control variables
Based on prior research, I include several control variables for other known determinants
17

of discretionary accruals in my model. First, I control for the size of a company (ASSET),
which determines earnings-management behavior in the sense that larger companies face
higher political costs (Watts and Zimmerman, 1986) and higher litigation risks (Lang and
Lundholm, 1993). Additionally, larger companies tend to report larger and more stable
accruals (Dechow and Dichev, 2002). I also control for the growth rate of a company’s
sales (GROWTH), which is likely to increase the absolute value of accruals. Dechow
(1994) also suggests that accruals and cash flow are negatively correlated, so I include cash
flow from operations (CFO) in my controls. Company age (AGE) is used to control for the
level of business experience of a company.
This study also controls for the first year after the audit partner with top tenure in an
audit team is replaced (CHG_FST) and the last year before the audit partner with top tenure
in an audit team is replaced (CHG_LST). In the first year, a successor audit partner may not
make decisions properly, because of a shortage of client-specific knowledge. In this case,
DA should be higher. In the last year of tenure, a partner change will occur in the
subsequent year. For a given length of partner tenure, if a partner change is more likely
when the company is facing more risk, then DA should be higher in cases of partner
change.

[Insert Table 1 about here]

18

4. Results
Data
The sample in this study is composed of data on listed companies in Japan from 1999 to
2006. As mentioned earlier, the reason that sample period has been set until 2006 is to
verify the influence of long tenure in the before and after the introduction of rotation rule.
Since the Big 4 auditing firms introduced the partner rotation rule as a means of
self-regulation in 2003, the sample has been divided into two sub-samples, which are for
the periods before the introduction of the rule (b2003) and after its introduction (a2003).
Our data source for accounting variables is the NIKKEI NEEDS–Financial Quest by
Nikkei Media Marketing. The name of the audit firm and signature of each audit partner
were obtained from the audit reports of each company. Information on the sample selection
is shown in Table 2. The initial sample included data on 15,636 Japanese listed companies
from 1999 to 2006. Companies that use US-GAAP, that conduct M&A, and for which
adequate financial data were not available were then deleted, leaving 13,245 firm-year
observations. In order to control for the effect that audit firm tenure has on the tenure of
audit partners, this study analyses companies that did not change their audit firm for more
than 10 years. After companies that are audited by non-Big 4 auditors, that are audited by
two audit firms in joint audits, or for which adequate auditor data is not available were all
eliminated, a final sample was left that consists of 7,567 firm-year observations.
19

[Insert Table 2 about here]

Descriptive statistics and correlations
Descriptive statistics on the variables in this study are presented in Table 3. The mean
(median) of absolute DA equals 0.30% (0.22%) of the total assets. The mean and median of

audit partner tenure are 2.211 (about 9 years) and 2.303 (about 10 years), respectively. The
tenures of audit partners in Japan are relatively long, compared to those of audit partners in
other countries. For example, Carey and Simnett (2006) report that the average tenure of
audit partners in Australia is 3.75 years, while Chen et al. (2008) report that the average
(median) tenure of audit partners in Taiwan is six years (five years). Note that the above
means and medians do not indicate that the relationships between clients and auditors are
significantly longer in Japan than in other countries. The mean value of AGE is 4.023 (56
years), which is double the mean for Taiwanese companies (22 years). This suggests that
the average company age in this study is relatively old, and that the years of audit
partnership are longer as a result. In the first year, 30.5% of audit partners changed
(CHG_FST), and in the last year (CHG_LST) 30.5% changed also, meaning that over 30%
of the total number of longest-standing audit partners changed during the eight years
addressed in this study.
[Insert Table 3 about here]
20

 Table 4 presents the correlations of variables. Pearson (Spearman) correlations are
shown above (below) the diagonal. TNRE is positively associated with ASSET and AGE,
and negatively associated with GROWTH, CFO, and CHG_FST. The positive correlation
between TNRE, ASSET, and AGE suggests that the accumulation of client-specific
knowledge is required for companies that are large and have long histories. The highest
correlation coefficient in the sample is 29%, for CHG_FST and TNRE, which is not at a
level that suggests multicollinearity.
[Insert Table 4 about here]

Univariate analysis
First, this study constructs 10 portfolios related to audit partner tenure, and examines the
distribution of DA in each portfolio. The results are shown in Figure 2. There is no clear
pattern in the mean (median) of DA by portfolio rankings of TNRE, but the pattern is clear
for extreme values of positive DA. The 90th percentile of DA is the highest for portfolio 2

and it becomes less positive as the portfolio ranking increases. On the other hand, the 10th
percentile of DA is almost constant in each portfolio. In addition, this trend is clearer in the
period prior to the introduction of the partner rotation rule, and cannot be observed after its
introduction, as can be seen in Panels A and B of Figure 3. In sum, the results suggest that
companies are less likely to record extreme values of income-increasing discretionary
21

accruals when their audit partner-client relationship is longer. The results in Figures 2 and
3 are based on univariate analysis, without controlling for any confounding factors.

Relationship between audit partner tenure and discretionary accruals
Table 5 shows the results of regressing discretionary accruals on audit partner tenure and
the control variables. I compute the f-statistics and z-statistics by using robust standard
errors adjusted for clustering by company. In other words, I treat the observations as
independent across companies, but not necessarily independent within companies (Rogers,
1993). The first column in Table 5 shows the OLS regression results when the dependent
variable is absolute discretionary accruals. The coefficient of TNRE equals -0.0014 (t =
-1.95). The significantly negative coefficient of TNRE is consistent with the pattern of less
extreme discretionary accruals over the length of audit partner tenure, as illustrated in
Figures 2 and 3. The coefficients of ASSET and AGE are significantly negative, and the
coefficient of GROWTH is significantly positive. These results are consistent with the
findings of prior research (e.g., Johnson et al., 2002; Myers et al., 2003; Chen, Lin, and Lin,
2008). The second column in Table 5 shows that when the dependent variable is signed
discretionary accruals, TNRE is negatively associated with signed DA (coefficient =
-0.0016, t = -2.41). The coefficients of CFO, AGE, and CHG_FST are significantly
negative and the coefficients of ASSET and GROWTH are positively associated with
22

signed discretionary accruals.
The third and fourth columns in Table 5 show the regression results when the sample is

divided into two sub-samples, which have positive and negative discretionary accruals.
Because the dependent variables are truncated at zero, we use truncated regression instead
of OLS regression to avoid biased coefficient estimates (Greene, 2000). For the regression
of positive DA, the coefficient of TNRE is -0.0028 (z = -2.3). For the regression of
negative DA, the coefficient of TNRE is -0.0010 (z = -0.65). These results suggest that
companies are less likely to record extreme values of positive (but not negative)
discretionary accruals as the length of audit partner tenure increases. Interestingly, these
findings are almost completely consistent with the results of Chen, Lin, and Lin (2008).
The reason will be considered later in this study. In summary, the results presented in Table
5 suggest that earnings quality improves with extended audit partner tenure, which is not
consistent with Hypothesis 1.
[Insert Table 5 about here]

Length of partner tenure
While this study finds that the magnitude of discretionary accruals decreases as audit
partner tenure grows longer, I was also interested in whether this relationship is conditional
on the length of partner tenure or not. Since five or seven years is the limit of partner tenure
23

imposed in most countries where partner rotation is mandatory, this study first investigates
whether the relationship between audit partner tenure and discretionary accruals differs
before and after tenure exceeds five or seven years. This question is important, because if
the magnitude of discretionary accruals decreases with partner tenure before but not after
tenure exceeds five or seven years, audit partner rotation requirements could be justified by
arguing that rotation does not impair earnings quality but helps to improve the appearance
of auditor independence (at certain costs). On the other hand, if decreases in the magnitude
of discretionary accruals occur only after tenure exceeds five years, it would be difficult to
justify partner rotation based on the argument that longer audit partner tenure impairs
earnings quality.
To investigate the above question, this study divided audit partner tenure into two

categories: tenure up to five (seven) years and tenure in excess of five years, and ran
piecewise linear regressions. The results, which are presented in Table 6, show that all
discretionary accruals are not negatively associated with audit partner tenure before tenure
exceeds five and seven years. After partner tenure exceeds five and seven years (but not
before then), only negative accruals decrease as the length of tenure increases. Thus, if the
main argument for audit partner rotation is that income-increasing earnings management
becomes more prevalent after partner tenure exceeds five or seven years, it is not supported
by the regression results in this study.
24

[Insert Table 6 about here]
One argument in support of mandatory audit partner rotation is that audit quality will be
compromised when the partner becomes excessively familiar with the client. The existence
of a U-shaped relationship between partner tenure and discretionary accruals would mean
that earnings quality is lowered when audit partner tenure is either very short or very long.
To investigate whether such a relationship exists, we drop TNRE in the regression model
and add two dummy variables, SHORTTNRE and LONGTNRE, which are equal to 1 if
TNRE<=2 (short tenure)
8
and TNRE > 5 or 7 (long tenure), respectively, and which are
equal to 0 otherwise. The regression in this new model uses the medium-tenure group (3-5
or 10) as its basis, and tests whether short and long tenures are associated with larger
discretionary accruals, after controlling for confounding factors. Panel A of Table 7 shows
that LNGTNRE is negatively associated with only positive DA. In addition, in Panel B of
Table 7, the coefficients of LONGTNRE are all negative and significant with the exception
of the case of negative DA. Taken together, the results in Table 7 show that earnings
quality is impaired when audit partner tenure is longer.
[Insert Table 6 about here]

Conditions before and after the enforcement of the rotation rule

8
Carey and Simnett(2006) defines short tenure as tenure less than 2 years.
25

Since 2003, companies in Japan have been required to periodically change audit partners
as a means of self-regulation. Changing audit partners is considered to eliminate the
negative personal relationships between clients and audit partners. Panel A of Table 8
displays results concerning the impact of audit partner tenure on discretionary accruals in a
sub-sample of companies before the enforcement of self-regulation. It shows that TNRE
has significant negative associations with all types of discretionary accruals, which means
that discretionary accruals decrease with the length of audit partner tenure. On the other
hand, TNRE has no significant relationship with any type of discretionary accruals in the
sample from after the enforcement of the 2003 rule. These results suggest that the
relationships between audit partners and their clients have disappeared with the
enforcement of the rotation rule.
Table 8 shows the results of a re-estimation of the model, which adds the interaction
variable to the independent variables. AFTER*TNRE denotes audit partner tenure in the
period after the enforcement of the 2003 rotation rule. Table 8 shows that AFTER*TNRE
is positively associated with all types of discretionary accruals, while TNRE is negatively
associated with all types. These results mean that the impact of tenure that existed prior the
introduction of the rule was offset by the rule, which is consistent with the results shown in
Table 7.
[Insert Table 7 about here]