kim et al - 2004 - selective auditor rotation and earnings management - evidence from korea

1



Selective Auditor Rotation and Earnings Management:
Evidence from Korea




Jeong-Bon Kim,
*
Chung-Ki Min and Cheong H. Yi















Current Draft
June 2004

___________
The first and third authors are at the School of Accounting and Finance, Faculty of
Business, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong.
The second author is at the Department of Economics, The Hankuk University of Foreign
Studies, Seoul, Korea. The research is partially funded by the Hong Kong Polytechnic
University. We have received useful comments from Kwan Choi, Michael Firth, Dan
Simunic, and workshop participants of The Hong Kong Polytechnic University,
University of British Columbia, City University of Hong Kong, Korea University, the
2002 AAAA Annual Conference and the 2003 AAANZ Annual Conference. The usual
disclaimer applies.

*Please forward all correspondence to Jeong-Bon Kim (Phone: 852-2766-7046; Fax:
852-2330-9845; E-mail:
)
2


Selective Auditor Rotation and Earnings Management:
Evidence from Korea

Abstract

In Korea, the regulatory authority designates external auditors for firms that are
deemed to have high incentives and/or great potential for opportunistic earnings

management, and mandates these firms to replace their incumbent auditors by new
designated auditors and requires them to keep the designated auditor for a period of
typically one to three years. We call this regulatory regime selective auditor rotation. This
paper investigates whether the selective auditor rotation rule in Korea is effective in
deterring income-increasing earnings management. Consistent with our hypothesis, we
find that the level of discretionary accruals is significantly lower for firms with
designated auditors than firms that freely select their auditors. We also find that the level
of discretionary accruals is significantly lower during the designation period, compared to
the level during a year prior to the imposed rotation. The above findings are robust to a
battery of robustness checks. Overall, our results are consistent with the notion that the
selective auditor rotation enhances audit quality and thus the credibility of financial
reporting.

JEL classification: G38; L15; L84

Keywords: Selective Auditor Rotation; Audit Quality; Earnings management


3


Selective Auditor Rotation and Earnings Management:
Evidence from Korea


I. INTRODUCTION

Since 1991, the regulatory authority in Korea has designated external auditors for
a group of firms that are deemed to have high incentives or great potential for accounting
manipulation (hereafter problematic firms). These problematic firms must replace their

incumbent auditors with new auditors designated by the regulatory authority, and retain
them for a specific period, typically one to three years. During our sample period of 1991
to 2000, about 15 to 17 percent of all firms listed on the Korea Stock Exchange were
required to replace their incumbent auditors with designated auditors. We call this
regulatory regime ‘selective auditor rotation’ in the sense that auditor changes are
mandated only for select problematic firms and not for all firms.
1

The selective auditor rotation (SAR) rule in Korea has several interesting features.
Firstly, the SAR rule mandates auditor changes only for problematic firms, thereby
preserving a competitive, market-based audit engagement for firms that are not deemed
problematic. In this regard, the SAR contrasts with suggested or existing mandatory
rotation that typically requires auditor changes for all firms after a certain period of an
initial engagement. By mandating auditor for select problematic firms, the SAR rule
intent is to alleviate alleged problems arising from long-term, auditor-client relationships
or repeat audit engagements. Secondly, the SAR mandates retention of newly designated
auditors for a certain period, thus protecting designated auditors from early dismissal, and
thereby mitigating a major threat to auditor independence. Finally, by requiring that

1
Appendix explains detailed criteria used for selecting these problematic firms.
4


incoming auditors be chosen by the regulatory authority rather than by audit clients, the
SAR rule limits managerial influence over auditor selection and any potential for low-
balling, thereby enhancing auditor independence and audit quality.
To our knowledge, no country other than Korea has ever introduced or considered
a selective auditor rotation policy. This institutional feature unique to Korea provides us
with an ideal laboratory setting in which one can compare differences in audit quality

between two distinct groups, namely firms with mandated auditor changes (treatment
group) and those with a free selection of auditors (control group).
In this paper, we take advantage of this unique institutional feature to provide
systematic evidence on the effect of selective auditor rotation on audit quality. Audit
quality is not directly observable, and thus it is difficult to measure empirically.
Consistent with prior research (Becker et al. 1998; DeFond and Subramanyam 1998;
Myers et al. 2003), we use accounting accruals measures to draw inferences about audit
quality. High quality auditors are more likely to detect accounting irregularities, object to
the use of questionable accounting practices, and limit discretion over accrual choices for
client firms (Reynolds and Francis 2001; Frankel et al. 2001). To the extent that the SAR
rule is effective in improving auditor independence and thus audit quality, designated
auditors are likely to be less lenient towards managerial discretion over opportunistic
accrual choices, and thus they are more effective in deterring aggressive earnings
management by their clients, compared to non-designated auditors (auditors chosen by
clients). As a consequence, firms with designated auditors should report lower
discretionary accruals than those with non-designated auditors, other things being equal.
5


To test this proposition, we first compare the level of discretionary accruals
between firms whose auditors are designated and firms whose auditors have never been
designated. In addition, we also compare the level of discretionary accruals during the
pre-designation period with the level during the designation period, using firm-year
observations with designated auditors, to see if the level of discretionary accruals
decreases after the imposition of mandated auditor changes. We perform our analyses
using 752 firm-year observations with designated auditors and 2,784 firm-year
observations with non-designated auditors during our sample period, 1991-2000. To
address potential measurement error problems associated with the Jones model (Kothari
et al. 2002), we include a control for firm performance in estimating discretionary
accruals.

Briefly, our analyses reveal that firms with mandated auditor changes report
significantly lower discretionary accruals than firms with no mandated auditor changes,
after controlling for other factors that have been shown to be associated with
discretionary accruals. We also find that for firms with mandated auditor changes, the
level of discretionary accruals becomes significantly lower (more negative) during the
designation period, compared with the level during the last one year prior to auditor
designation. This evidence further corroborates our initial results. Taken together, our
results indicate that the selective auditor rotation rule enhances audit quality and is
effective in deterring opportunistic earnings management. Put differently, our results
suggest that the imposition of selective auditor rotation leads to the expected changes
desired by the regulatory authority with respect to the problematic firms at which the
SAR rule is targeted. Finally, we conduct a battery of sensitivity analyses to check the
6


robustness of our findings. Overall, the results of various sensitivity checks reveal that
our findings are robust to alternative test designs and/or explanations.
This paper adds to the existing literature on auditor independence and audit
quality (e.g., Frankel et al. 2002; Ashbaugh et al. 2003; Myers et al. 2003) in the
following ways. Recent incidents of audit failures in the US (e.g., Enron debacle) have
triggered a world-wide regulatory debate over how to mitigate potential threats to auditor
independence ranging from managerial influence over auditor choice to auditors’
economic dependence on clients. In response to the above concern, the US Congress
enacted the Sarbanes-Oxley Act of 2002 in which the lead or the coordinating audit
partner for a certain client must be rotated every five years. In addition, the Act called for
further research into the potential effects of requiring mandatory rotation of audit firms,
not just audit partners within the same firm.
2
Proponents of mandatory auditor rotation
argue that mandating auditor changes after some specified period would truncate

potential economic gains to auditors arising from repeat audit engagements with the same
client, and provide stronger incentives for auditors to act independently. The selective
auditor rotation (SAR) is similar to the mandatory auditor rotation in the sense that both
measures aim to enhance auditor independence by truncating the client-specific rents to
auditors. As mentioned earlier, however, the requirement for mandated auditor changes
under the SAR rule applies only to problematic firms, while the mandatory rotation rule
would apply to all firms. The selective auditor rotation may be viewed as a regulatory
measure that lies between the mandatory rotation system and the voluntary rotation

2
Sec. 207 of the Sarbanes-Oxley Act of 2002 states that the Comptroller General of the Unite States shall
conduct a study and review of the potential effects of requiring the mandatory rotation of audit firms. In
fact, the imposition of mandatory rotation has been discussed at various times over the last three decades in
7


system. Korean evidence reported in this paper suggests that a form of limited regulatory
intervention on auditor changes targeted at potentially problematic firms could be
considered as a possible alternative to complete mandatory rotation.
Several studies (e.g., Davies et al. 2003; Geiger and Raghunandan 2002; Myers et
al. 2003) have recently examined the effect of imposing limits on auditor tenure on
financial reporting quality. The evidence on whether extended auditor tenure improves or
deteriorates audit quality, however, is mixed. For example, Myers et al. (2003), using
several accruals measures to proxy for earnings quality, find evidence suggesting that
longer auditor tenure increases earnings quality. In contrast, Davies et al. (2003) find
results suggesting that discretionary accruals increase with auditor tenure. In addition,
since the above evidence is obtained in a voluntary auditor rotation setting, it is difficult
to extrapolate the effects of mandated auditor changes from the evidence. A notable
exception is Dopuch et al. (2001) who analyze the question in a laboratory setting, and
provide experimental evidence in favor of mandatory auditor changes. The results

reported in this paper provide useful insights into the expected effect on audit quality, if
the mandatory auditor rotation policy is imposed in the US and other countries, in
particular, for firms with high incentives for opportunistic earnings management.
The remainder of the paper is organized as follows. The next section presents an
overview of the Korean audit services market. Section III develops our hypothesis.
Section VI discusses research design, including sample selection and empirical model
specification. Section V presents empirical results. Section VI reports the results of
additional analyses. The final section concludes the paper.

the U.S. (e.g., the U.S. Senate’s Metcalf Subcommittee Report 1977; the AICPA’s Cohen Commission
Report 1978; SEC 2000), although it has never been adopted.
8



II. INSTITUTIONAL BACKGROUND

Overview of the Korean Audit Services Market

During late 1970s and early 1980s, the Korean economy experienced phenomenal
growth, along with a rapid development in capital markets, which in turn increased the
demand for credible financial reporting and external auditing. This change in the
economic environment prompted the regulatory authority in Korea to introduce the Act
on External Audit (AEA) that was enacted in 1980. The AEA fermented many changes in
the accounting and auditing professions in Korea. On the demand side, the AEA
increased significantly the number of firms that are subject to external audits by requiring
the financial statements of a firm with more than 6 billion Korean Won in total assets to
be audited by an independent auditor. About 7,000 firms, including unlisted firms, were
subject to external audits in 2000. On the supply side, the AEA loosened restrictive
licensing procedure for certified public accountants (CPAs), resulting in an increase in

the number of CPA. The Korean Institute of Certified Public Accountants (KICPA),
established in 1954 to improve CPA skills and monitor professional conducts of its
members, had 5,309 CPAs registered in 2000. 32 local audit firms were practicing as of
2000 and many of them have a member firm relationship with international accounting
firms such as Big Five audit firms.
While the introduction of the AEA led to the growth of the auditing profession,
there are several aspects of institutional and socio-economic environments that pose a
threat to auditor independence and audit quality in Korea. First, founding families of
Korean firms typically exercise significant control over operations and other aspects of a


9


firm’s activities (Joh 2003). Members of founding families can effectively maintain
control over firms through pyramidal ownership and cross shareholdings, even though
they have relatively low cash flow rights. This separation of ownership and control
creates agency conflicts between controlling shareholders and minority shareholders
(Johnson et al. 2000a; Claessens et al. 2002). Controlling shareholders also exert
significant influence over firms’ financial reporting and auditor selection (Fan and Wong
2002). The conflict of interest between controlling shareholders and minority
shareholders motivates controlling shareholders to engage in opportunistic earnings
management and hire auditors more acquiescent to their demands regarding accounting
choices. Controlling shareholders’ influence over auditors becomes even greater when
auditors earn client-specific rents in the form of the provision of non-audit services.
Second, the Korean society and economy have traditionally been operated, in
large part, on the basis of personal relationships. These personal relationships typically
take the form of family ties (heol yeon), school ties (hak yeon), or regional ties (chi yeon).
These relationships are deeply rooted in long-held Confucian cultural traditions, and
loyalty to organizations and obedience to seniors within the tie-based networks have been

considered as the most important factor leading to an individual success or successful
operations of organizations in Korea. In this relationship-based economy, managers are
likely to select auditors based on personal ties, which may cause inherent limitations on
auditors being independent of managers. In a related vein, conflicts among interested
parties have typically been resolved in Korea through direct communications among
them to preserve future business within the relationship-based network. Not surprisingly,
shareholder litigation against auditors has been less common in Korea, and the amount of
10


court-awarded damage has been relatively small.
3
In this relationship-based economy
with low litigation risk, it is a daunting task to maintain auditor independence.
Introduction of Auditor Designation

In response to concerns over audit quality and the credibility of financial reporting
voiced by financial statement users, the Financial Supervisory Commission (FSC),
equivalent to the Securities and Exchange Commission (SEC) in the U.S., established the
Committee for External Audit Improvement with a mandate to propose a remedial action
plan for improving auditor independence in 1989. Following the Committee’s
recommendation, the Amendment of the AEA became effective in December 1989.
Under the Amendment, the FSC is empowered to impose mandated auditor changes for a
group of problematic firms. As described in Appendix in details, Article 10 of the FSC
Regulation associated with the AEA stipulates the type of firms that could be subject to
mandated auditor changes. A close examination of Article 10 reveals that these
problematic firms may have poor corporate governance or high agency problems (e.g.,
insufficient separation of ownership and management, and excess loans to related
parties), may face financial troubles (e.g., excessive reliance on debts, industry
restructuring, and trading on administrative post), and may make questionable auditor

changes, or violate GAAP in preparing annual reports. These firms are believed to have
high incentives or great potential for accounting manipulation or opinion shopping. To
increase audit quality, the FSC is empowered to mandate firms falling under the
categories described in the Appendix (what we call problematic firms earlier) to replace
their incumbent auditors with auditors designated by the FSC, and to retain designated

3
Since the first case of auditor litigation happened in 1991, there have been a total of 22 litigations against
auditors during the period 1991-1999. The largest amount of court-awarded damage was far less than
11


auditors over a certain period, typically one to three years. Since the selective auditor
rotation rule was first introduced in 1991, the FSC has imposed mandated auditor
changes for about 15 to 17 percent of all firms listed on the KSE every year.

III. HYPOTHESIS DEVELOPMENT

Audit quality is often defined as the probability of both detecting and reporting a
breach in the financial statements (DeAngelo 1981; Watts and Zimmerman 1986). The
reporting of a detected breach partially requires auditor independence. In the relationship-
based economy with relatively low litigation risk where external auditors tend to be
chosen on the basis of their personal relations with incumbent managers or corporate
insiders, the selective auditor rotation (SAR) rule could be an effective regulatory
measure for enhancing auditor independence and audit quality. Under the SAR rule,
control over the hiring and firing of external auditors is taken away from management,
and designated auditors must be retained for the designation period. The selective auditor
rotation could minimize managerial influence over the firing and hiring of external
auditors, in particular, for firms with high incentives for opportunistic earnings
management. As a result, designated auditors would be in a stronger position to resist

clients’ pressure to bias financial reporting, and they are more likely to make objective
professional judgments.
On the other hand, the selective auditor rotation could decrease audit quality.
Incoming auditors designated by the regulatory authority may be less competent because
they are not industry experts or they lack knowledge of a client’s business and operations.
Since auditors’ understanding of their clients increases with auditor tenure, imposing

US$1.0 million.
12


limits on auditor tenure may increase the number of audit failures. Consistent with this
argument, the AICPA’s Quality Control Inquiry Committee of the SEC Practice Section
reported that allegations of audit failure occur much more frequently when auditors are in
their first or second year of a new audit engagement (AICPA 1992). Similarly, Geiger
and Raghunandan (2002) found that more audit reporting failures occur in the earlier
years of the auditor-client relationship. It should be noted, however, that the above US
evidence on the earlier-year audit failures has been obtained in an environment of no
mandatory auditor rotation and thus be interpreted cautiously (Imhoff 2003).
In this study, we empirically investigate whether the selective auditor rotation for
problematic firms improves audit quality, which, in turn, enhances the ability of an
auditor to detect and report accounting manipulation or substandard reporting. Higher
quality auditors are better able to withstand clients’ pressure to allow substandard
reporting and they are more likely to object to questionable accounting practices, and
constrain management’s ability to distort the true financial performance of the firm
(Levitt 1998; Public Oversight Board 2000). To the extent that the selective auditor
rotation for problematic firms is effective in constraining opportunistic earnings
management by such firms, the level of discretionary accruals should be lower for firms
with designated auditors than for firm with non-designated auditors. Hence, we test the
following hypothesis in an alternative form:

H
A
: Firms with designated auditors report lower discretionary accruals than
firms with non-designated auditors, other things being equal.

Our analysis focuses on the level of discretionary accruals or income-increasing
earnings management for several reasons: First, evidence shows that managers are more
likely to be involved in income overstatement than income understatement (DeFond and
13


Jiambalvo 1991, 1993; Kinney and Martin 1994; Becker et al. 1998). Second, when
examining the issue of earnings management in a general context, income-decreasing
earnings management is inherently less interesting than income-increasing earnings
management. Third, previous research (e.g., Dechow et al. 1996) suggests that poor
corporate governance facilitates earnings manipulation. During the 1990s (which includes
our sample period), the Korean corporate sector had experienced low profitability due
primarily to poor corporate governance. For example, Joh (2003) shows that low
profitability is associated with poor corporate governance in Korea, which allows
controlling shareholders to divert assets and profits from the firm for their private gains at
the expense of minority shareholders. Low profitability, along with poor corporate
governance, is likely to create incentives for controlling shareholders and incumbent
managers to boost reported earnings via income-increasing accruals to camouflage poor
earnings performance. Consistent with this conjecture, more than 80 % of firms receiving
qualified audit opinions for departures from GAAP allegedly overstated reported earnings
during our sample period (Financial Supervisory Commission 1999). Finally, unlike
accounting standards in Japan and Germany, Korea’s accounting standards are not
closely tied to tax accounting rules, thus lowering potential opportunity costs associated
with income-increasing earnings management. For the above reasons, we are primarily
interested in examining the effect of selective auditor rotation on the level of

discretionary accruals.

14



IV. RESEARCH DESIGN

Sample Description

The initial sample for this study consists of all firms listed on the Korea Stock
Exchange (KSE) that are included in the 2001 KIS-DATA files developed by the Korea
Information Service (KIS). The KIS is a credit rating agency in Korea and provides
corporate financial and ownership information of all listed firms on the KSE. For our
auditor-designated sample, we obtained a list of 362 firms (948 firm-year observations)
whose auditors are designated during 1991-2000 from the Accounting System and Audit
Review Department of the Korea Financial Supervisory Service, the operating arm of the
FSC.
4
We delete 6 firms (12 firm-years) in the financial services industry (commercial
banking, investment brokerage, insurance, etc) from the sample because the nature of
accruals for firms in this industry differs from that in other industries. 52 firms (184
firm-years) are also excluded from the sample due to missing data for measuring research
variables in this study.
Table 1 summarizes our sample selection procedure. As outlined in Panel A of
Table 1, we identify a total of 304 firms (752 firm-year observations) with designated
auditors over the period 1991-2000 after applying the above selection criteria. Panel B of
Table 1 reports the distribution of firms with designated auditors (our treatment sample),
along with firms with non-designated auditors (our control sample) by year of auditor
designation. Our control sample consists of 2,784 firm-year observations (between 250

and 293 firms each year) with sufficient data that never have their auditors designated by

4
We thank Professor Choi Kwan of SungKyunKwan University in Seoul, Korea for providing the list of
firms subject to auditor designation during the period 1991-1993.
15


the FSC during the sample period. Panel C classifies our designated sample according to
the designated period.
5
Over 62% of the firm-years are designated for three years or less.
[INSERT TABLE 1 ABOUT HERE]

Table 2 displays our treatment sample by reason for auditor designation. The table
indicates that two major reasons for auditor designation are excess reliance on external
debts (51%) and insufficient separation of ownership and management (28%). Before the
Korean financial crisis erupted in December 1997, 71% of our firm-year observations
have their auditors designated because of high leverage, while concentrated ownership
structure is the major reason for auditor designation after the crisis.
[INSERT TABLE 2 ABOUT HERE]
Measuring Discretionary Accruals

Like many other studies, this study uses the level of discretionary accruals to
proxy for the extent of opportunistic earnings management. Previous research indicates
that it is important to control for the effect of firm performance on accruals when
measuring discretionary accruals using the Jones (1991) model or its variants. Dechow et
al. (1995) and Kasznik (1999) show that estimated discretionary accruals are correlated
with earnings performance. Firms with low (high) earnings tend to have negative
(positive) discretionary accruals. Further, testing for earnings management could yield

biased results if measurement error in the estimated discretionary accruals is correlated
with the partitioning variable of interest (e.g., whether a firm is auditor-designated or not
in our study). To the extent that firms with poor earnings performance are more likely to

5
In Panel C of Table 1, the total number of frequency (345) is different from the total number of firms
subject to auditor designation (304) because several firms are auditor designated more than once at
different points in the sample period.
16


be auditor-designated, failure to control for this correlated omitted variable would induce
a bias in our study, which may lead to erroneous inferences.
Kasznik (1999) and Kothari et al. (2002) suggest two approaches to control for
correlated omitted variables: (1) to use a matched firm (or portfolio) technique; or (2) to
include a control for firm performance in the regression model used to estimate
nondiscretionary accruals. We take the latter approach in this study since there is a risk of
mismatching on the conditioning variables (firm performance and industry membership)
given the relatively small number of listed firms (704 firms as of year 2000) on the KSE.
6

Imperfect matching will introduce noise into discretionary accruals, which weakens the
power of our test for earnings management. However, the weakness of the latter approach
is that it imposes a linear relation linking accruals to earnings performance in the cross-
section.
To estimate discretionary accruals (DAC), we first compute total accruals as the
change in non-cash current assets minus the change in current liabilities excluding the
current portion of long-term debt minus depreciation and amortization expenses.
Formally,
jtjtjtjtjtjt DepSTDCLCashCATAC

−
∆
−
∆
−∆−∆= )()( (1)
where, for firm j and in year t:
TAC
jt
= total accruals;

CA∆
jt

= change in current assets;

Cash∆
jt

= change in cash and cash equivalents;

6
We tried to control for the impact of performance on estimated discretionary accruals using a
performance-matched firm’s discretionary accruals. We attempted to match performance (lagged ROA)
within the 80%-120% filter in the same industry and year. However, given the small number of firms listed
on the stock exchange, we were unable to identify a matched firm for more than 60% of our treatment
sample.
17

CL∆
jt

= change in current liabilities;

STD∆
jt

= change in long-term debt included in current liabilities; and

Dep
jt

= depreciation and amortization expenses.

As discussed above, we employ an augmented version of the cross-sectional
modified Jones (1991) model to decompose total accruals (
TAC) into discretionary and
nondiscretionary accruals. Specifically, we include last year’s return on assets (ROA) as a
control for firm performance in the model:
jtjt
jtjtjtjtjtjtjt
eROA
APPEAREVAATAC
++
+
∆
+
=

−
−−−−
][
]/[]/[]/1[/
14
1312111
α
α
α
α
(2)
where:
TAC
jt

= total accruals for firm j in year t;

A
jt-1

= total assets for firm j in year t-1;

∆
REV
jt

= change in net revenues for firm j in year t;

PPE
jt

= gross property, plant, and equipment for firm j in year t;

ROA
jt-1

= return on assets (earnings before extraordinary items divided by
lagged total assets) for firm j in year t-1; and

e
jt

= error term.

For each year and industry, we estimate regression parameters in Eq. (2) using
cross-sectional observations. Following the industry classification of the KSE, we
classify firms into 18 industries.
7
To obtain meaningful cross-sectional estimates of

7
The distribution of auditor-designated firms (304 firms) across 18 industries is as follows: Fishing (0);
Mining (0); Food & Beverage (26); Textile & Apparel (31); Paper & Wood (21); Chemicals (38); Medical
Supplies (14); Non-metallic Minerals (10); Iron-Metals (32); Machinery (15); Electrical & Electronic
Equipment (44); Medical & Precision Machines (0); Transport Equipment (25); Distribution (0); Electricity
& Gas (19); Construction (2); Transport & Storage (20); and Services (7).
18


regression parameters in Eq. (2), we require that at least 10 firms exist for each industry

in each sample year. The parameters from Eq. (2) are used to calculate nondiscretionary
accruals with a performance control (
NDAC):
][]/[
]//[]/1[
1
4
1
3
11
2
1
1
−
∧
−
∧
−−
∧
−
∧
++
∆−∆+=
jtjtjt
jtjtjtjtjtjt
ROAAPPE
ARECAREVANDAC
αα
αα
(3)

where
jt
REC∆ is the change in net receivables. DAC is equal to TAC minus NDAC. As in
other studies, DAC is considered to be the outcome of managers’ opportunistic accrual
choices.
Model Specification

We test our hypotheses by estimating the following OLS regression model linking
DAC with auditor designation and other control variables:
itit
itititititit
IndustryBIG
TACLLEVSIZEOCFDESIGDAC
εβ
β
β
β
β
β
β
+++
+
+
+
++=
) Dummies(5
1
6
543210
(4)

where, for firm i and year t:
DAC
it
= the level of discretionary accruals computed using Eq. (2);
DESIG
it
= a dummy variable that takes the value of 1 if a firm’s auditor
is designated and zero otherwise;

OCF
it

= operating cash flows (earnings before extraordinary items
less total accruals) scaled by total assets;

SIZE
it

= the natural logarithm of the book value of total assets;

LEV
it


= the ratio of total debts to total assets;

L1TAC
it

= last year’s total accruals scaled by total assets;

BIG5
it

= a dummy variable that takes the value of 1 if the auditor is a
Big Five and 0 otherwise;

Industry dummies
= dummy variables controlling for industry differences; and
19



ε
it


= error term.


Our variable of interest, DESIG, is an indicator variable that takes the value of 1
when a firm’s auditor is designated. Its coefficient, β
1,
captures the difference in the level
of discretionary accruals between the designated and non-designated samples after
controlling for all other variables included in Eq. (4). To the extent that auditor
designation is effective in constraining the ability of managers to boost reported earnings
through income-increasing accrual choices, one would observe a negative coefficient for
DESIG in Eq. (4).
We include several control variables in Eq. (4) that may affect the level of

discretionary accruals. Previous research documents a negative correlation between DAC
and cash flow performance (Dechow et al. 1995; Becker et al. 1998; Kim et al. 2003).
We thus include operating cash flows scaled by lagged total assets (OCF) in Eq. (4) to
control for the potential confounding effects of OCF on our results. We also include the
natural logarithm of the book value of total assets (SIZE) and the ratio of total debts to
total assets (LEV) as control variables because previous research suggests that they may
affect discretionary accrual choices in the current period (DeFond and Jiambalvo 1993;
Becker et al. 1998; DeFond and Park 1997; Kim et al. 2003).
As in Ashbaugh et al. (2003), we include the prior year’s total accruals (L1TAC)
to capture the reversal of accruals over time. Previous research provides evidence that
Big Five auditors are effective in constraining managers’ abilities to boost reported
earnings through income-increasing accrual choices (Becker et al. 1998; Francis et al.
1999; and Kim et al. 2003). Fan and Wong (2002) also document that in East Asian
20


economies, Big Five auditors are more effective in mitigating agency problems. To
control for the effect of audit quality differentiation on our results, we include a dummy
variable (BIG5) indicating auditor type in Eq. (4).
8
Finally, we include industry dummies
to control for potential confounding effects of industry differences on DAC.

V. EMPIRICAL RESULTS

Descriptive Statistics and Univariate Tests

Table 3 presents descriptive statistics for major financial variables, along with
univariate tests for differences between the two samples, namely one with designated
auditors and the other without designated auditors. Panels A and B of the table present

the mean and median for the sample of firm-years with designated auditor and the sample
of firm-years with non-designated auditors, respectively. Panel C presents the results of
parametric and non-parametric tests for differences in each variable between the two
samples. Overall, Table 3 reveals that the designated sample tends to be less profitable,
smaller, and more leveraged than the non-designated sample. The designated sample also
generates less cash flows than the non-designated sample. Compared with the designated
sample, a larger portion of the non-designated sample is audited by Big Five auditors.
The level of lagged total accruals (L1TAC) is negative for both the designated and non-
designated samples, although L1TAC is more negative for the designated sample.
Finally, the firm-years with designated auditors have a lower level of
discretionary accruals (DAC) than the firm-years with non-designated auditors. The mean
(median) DAC is -1.1% (-0.2%) of lagged total assets for the designated sample,

8
Big Five auditors have a member firm relationship with large local audit firms since Big Five auditors are
not allowed to run their own operations in Korea without partnering with local firms. The local firms are
provided services of technical expertise and quality control from Big Five firms.
21


compared to 0.1 % (0.0%) for the non-designated sample. As indicated in Panel C, the
difference is statistically significant with p < 0.01 (p < 0.10) for parametric test (non-
parametric test). The significant difference in DAC between the two samples is
consistent with our hypothesis (H
A
) that firms with designated auditors report relatively
lower discretionary accruals than firms with non-designated auditors. However,
significant differences in other firm characteristics observed between the two samples are
likely to confound the univariate comparison. In what follows, our analysis thus focuses
on various multivariate tests.

[INSERT TABLE 3 ABOUT HERE]

Multivariate Analysis

Panel A of Table 4 reports the regression results for Eq. (4),
9
using the full sample
(with both positive and negative DAC). The regressions in Table 4 include industry
dummies to control for potential confounding effects of industry-level variations in
accounting standards and regulations on our results, though their coefficients are not
reported for brevity. As shown in the Panel A of Table 4, the coefficient of the auditor
designation dummy (DESIG), namely β
1
, is highly significant with an expected negative
sign. The β
1
-coefficient of -0.026 is significant (p < 0.01) enough to support our
hypothesis (H
A
) that, ceteris paribus, firms with designated auditors report lower DAC
than those with non-designated auditors. Given that mean return on assets (ROA),
measured by earnings before extraordinary items divided by lagged total assets, is -1.9%
and 3.3% for firms with designated auditors and those of non-designated auditors,

9
To avoid cross-correlation of residuals in our analysis, we estimate the model in Eq. (4) separately for
each of the ten years in our sample. Our results (untabulated) show that the average coefficient of DESIG is
–0.024 with t=-2.11, confirming the result of the pooled regression. That is, firms with designated auditors
report lower DAC than firms with non-designated auditors.
22

respectively, as shown in Table 3, this incremental DAC difference (2.6% of lagged total
assets) is economically significant as well. The above results suggest that auditor
designation for Korean firms is effective in limiting management’s ability to boost
reported earnings through income-increasing accrual choices.
The coefficient on OCF is highly significant with a negative sign. This is
consistent with the findings of Dechow et al. (1995), Becker et al. (1998) and Kim et al.
(2003). DAC is positively associated with firm size, but not significant. The relation
between DAC and LEV is significantly negative. This negative relation is consistent with
the debt-monitoring hypothesis that outside debt suppliers, primarily commercial banks,
monitor managerial opportunism such as opportunistic earnings management (Jensen
1986, 1989). The association between DAC and L1TAC is highly significant with a
negative sign, consistent with the findings of Ashbaugh et al. (2003). The coefficient for
the Big Five auditor dummy (BIG5) is negative but not significant.
Next, we partition our sample into two groups based on the sign of firms’
discretionary accruals to examine whether there is any differential relation between the
auditor designation and our measures of discretionary accruals conditional upon the sign
of discretionary accruals (DAC). Panel B of Table 4 reports the results separately for
firms with positive DAC and for firms with negative DAC
10
. We find the association
between DAC and DESIG is significantly negative for firms with positive DAC. This
finding lends further support to our hypothesis that mandated auditor changes effectively
limit managers’ ability to boost reported earnings through income-increasing accrual


10
Previous studies (Myers et al. 2003; Ashbaugh et al. 2003) note that the OLS estimates are, in general,
biased toward zero when a sample is truncated. To address this problem, we also use a maximum likelihood

23


choices. We also find a negative association between DAC and DESIG for firms with
negative DAC. On the one hand, this is consistent with the notion that auditor designation
leads to lower DAC regardless of whether firms have positive or negative DAC. On the
other hand, designated auditors may be more likely to
exercise a heightened degree of
professional skepticism on managers’ income-increasing accrual choices and their impact
on current earnings, compared to non-designated auditors, because firms subject to
auditor designation (i.e., problematic firms) presumably have high incentives to overstate
current earnings and the period of designation is rather short. Thus, to the extent that
firms with negative DAC are more likely to have engaged in income-decreasing earnings
management, the above finding may be interpreted in such a way that designated auditors
are less effective in constraining income-decreasing accrual choices than non-designated
auditors.
[INSERT TABLE 4 ABOUT HERE]
The Effect of Extreme Performances on Regression Results
Table 3 suggests that our designated sample is in a relatively poor financial health
measured by ROA and OCF. Although we controlled for individual firm performance
when estimating discretionary accruals, a concern still remains that the results we present
in Table 4 may be attributable to differences in firm performance between the designated
and non-designated samples. To mitigate this concern, we re-estimate Eq. (4) after
deleting observations in the extreme (top and bottom) deciles of ROA (measured by year
and industry). Deleting firms that fall into the extreme deciles of ROA reduces the sample
size to 2,940 observations: 568 firm years of designated firms and 2,372 firm years of

truncated regression approach. The results are similar to those reported in the table and our inferences
based on OLS estimations are unchanged.
24

non-designated firms. The regression results are presented in Table 5. Overall, the
regression results reported in Table 5 are similar to those reported in Panel A of Table 4,
suggesting that the regression results reported in table 4 are not driven by extreme
performance. In particular, we find the coefficient on DESIG is significant with an
expected negative sign (p < 0.01), which further supports our hypothesis, H
A
.
[INSERT TABLE 5 ABOUT HERE]
Analyzing Auditor-designated Firms Only
In the above analyses, we have examined the difference in the levels of
discretionary accruals between the sample of firms with designated auditors (treatment
sample) and that with non-designated auditors (control sample) to draw inference about
the impact of selective auditor rotation. In this subsection, we further investigate the
behavior of discretionary accruals only for the sample of firms with designated auditors
before and after the imposition of mandated auditor changes. For this test, we require that
the sample of firms with auditor designation must have data required to compute
discretionary accruals and control variables in the last one year prior to auditor
designation and in each of the years during the designation period. This requirement
yields 1,072 firm year observations. The results are reported in Table 6. Panel A of Table
6 shows that the level of discretionary accruals is significantly lower during the auditor
designation period compared to the level in the last one year prior to auditor designation,
suggesting selective auditor rotation is effective in constraining managers’ income-
increasing accounting choices.
To further examine any differential effect of differences in the auditor designation
period on our analysis, we include several dummy variables to capture potential
25

differences in the designated period. Panel B shows that, relative to the benchmark year
i.e., the last one year prior to auditor designation, the level of discretionary accruals is
significantly more negative in the first and second years of the designation, and that after
the second year, becomes less negative and less significant, suggesting that selective
auditor rotation is more effective in constraining income-increasing earnings
management in earlier years of mandated auditor changes than in later years. Taken
together, our results suggest that the imposition of selective auditor rotation resulted in
the expected changes desired by the FSC, in particular, for firms that are deemed to
engage in aggressive earnings management.
[INSERT TABLE 6 ABOUT HERE]

VI. ADDITIONAL ANALYSES
The Impact of the 1997 Asian Economic Crisis

The Korean economy has undergone a structural change since the Asian
economic crisis erupted in December 1997. Firm value plummeted and many firms went
bankrupt or underwent restructuring as a result of the crisis. Since the crisis, several
reform measures have been introduced to improve auditor independence and corporate
governance in Korea. They include mandated outside directors, enhanced minority
shareholders’ rights, and strengthened penalties for auditors and management for
accounting irregularities. We also note that as shown in Table 2, Korean firms have
auditors designated for different reasons before and after the crisis. To examine the effect
of this structural change on our analysis, we partition our sample into two periods (before
the crisis and after the crisis) and re-estimate Eq. (4) for each sub-period. Table 7 reports