Summarize and PowerPoint

Make a summary and PowerPoint and answer case study (A,B) (Please use easy words and easy sentences)

Save Time On Research and Writing
Hire a Pro to Write You a 100% Plagiarism-Free Paper.
Get My Paper

A.prepare a 3 or 4 pages summaryof each paper (3 Article) (including the Libby box)

Focus on summarizing the most salient points of the article.

. After that make sure the summaries address the following questions (I will send them).

B.and make PowerPoint slides for each study.

Save Time On Research and Writing
Hire a Pro to Write You a 100% Plagiarism-Free Paper.
Get My Paper

Three Articles in the attachments

Article
Does the Timing of Auditor
Changes Affect Audit
Quality? Evidence From the
Initial Year of the Audit
Engagement
Journal of Accounting,
Auditing & Finance
2020, Vol. 35(2) 263–289
ÓThe Author(s) 2017
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0148558X17726241
journals.sagepub.com/home/JAF
Cory A. Cassell1, James C. Hansen2, Linda A. Myers3,
and Timothy A. Seidel4
Abstract
We focus on the first year of the auditor–client relationship and investigate whether audit
quality varies with the timing of the new auditor’s appointment. We find that audit quality is
not lower for companies that engage new auditors before the end of the third fiscal quarter
than for companies that do not change auditors. However, companies that engage new
auditors during or after the fourth fiscal quarter are more likely to misstate their audited
financial statements than companies that engage new auditors earlier in the year and companies that do not change auditors. In additional tests, we find that the decrease in audit
quality associated with late auditor changes is more pronounced for companies with complex operations (i.e., more operating segments). These results suggest that the extent to
which audit quality suffers in the first year of audit engagements is affected by both the
amount of time required to understand the client’s business, assess risks, and perform the
audit (all of which are driven by client complexity), as well as the amount of time available
for auditors to perform these tasks.
Keywords
audit quality, auditor changes, misstatements, auditor independence
Introduction
Deficiencies in public company audits, identified through ongoing Public Company
Accounting Oversight Board (PCAOB) inspections, have prompted standard setting and
rulemaking activities focused on auditor independence, objectivity, and professional
1
University of Arkansas, Fayetteville, USA
Weber State University, Ogden, UT, USA
3
The University of Tennessee, Knoxville, USA
4
Brigham Young University, Provo, UT, USA
2
Corresponding Author:
Linda A. Myers University of Tennessee, Knoxville Haslam College of Business Department of Accounting &
Information Management Stokely Management Center Knoxville, Tennessee 37996-0560, Knoxville, TN, USA.
Email: lmyers16@utk.edu
264
Journal of Accounting, Auditing & Finance
skepticism, with the goal of improving audit quality.1 Although there appears to be a
common belief among stakeholders that improvements to the current audit model are
needed (Franzel, 2012), one of the PCAOB’s proposals—the Concept Release on Auditor
Independence and Audit Firm Rotation—received especially strong opposition. In fact,
94% of approximately 600 comment letters received in the initial comment period opposed
audit firm rotation (Ernst & Young LLP, 2012).
The comment letter process and outreach efforts of the PCAOB have resulted in proposals for a number of alternatives to mandatory audit firm rotation. These include strengthening audit committees and increasing disclosure about their activities, requiring targeted
audit firm rotation (i.e., mandating audit firm rotation only when specific criteria are met),
and requiring periodic rebidding (tendering) of the audit (Franzel, 2012). Each of these
alternatives has been proposed with the intention of improving what regulators, stakeholders, and others suggest is the root cause of the problem—a lack of sufficient auditor
independence, objectivity, and professional skepticism. Indeed, some argue that
setting a limit on the continuous stream of audit fees that an auditor may receive from one client
would free the auditor, to a significant degree, from the effects of management pressure and
offer an opportunity for a fresh look at the company’s financial reporting. (PCAOB, 2011, p. 17)
Importantly, a key commonality among the proposed regulatory remedies is that they
would result in more frequent auditor changes.
A primary argument made by stakeholders who oppose regulations that could result in
more frequent auditor changes relates to concerns about the ‘‘audit learning curve’’ experienced on new audit engagements. For example, James Copeland, former CEO of Deloitte
& Touche, stated,
There is strong evidence that requiring the rotation of entire firms is a prescription for audit failure. It would result in the destruction of vast stores of institutional knowledge and guarantee that
auditors would be climbing a steep learning curve on a regular basis. (PCAOB, 2011, p. 13)
Arguments like this are supported by evidence from academic research which generally
finds that financial reporting quality is lower for companies with shorter auditor tenure.
Ideally, the audit learning curve should not affect the outcome of the audit because auditors should evaluate a prospective client and accept the engagement only if they can devote
the resources and have (or can obtain) the expertise necessary to perform an audit that
achieves reasonable assurance. However, this is difficult to achieve in practice for a couple
of reasons. First, all public companies, regardless of circumstances, are required to file
audited financial statements with the Securities and Exchange Commission (SEC) by the
required filing deadline (based on public float). Thus, although a given audit firm may reject
a prospective client due to insufficient resources and/or expertise, that same client will continue to pursue other audit firms until it is able to find an audit firm willing to perform the
audit.2 Second, certain audit tasks (e.g., risk assessments, complicated judgments, etc.) may
be difficult to address by simply devoting additional audit personnel to the task because they
require an audit team to spend time assessing the client and its associated audit risks.
Collectively, these arguments suggest that the quality of some engagements may be affected
by the timing of auditor changes due to constraints on the production of the audit.
Understanding whether an audit learning curve exists, and factors that affect the shape
of the curve (e.g., its length and steepness), would certainly be of interest to regulators and
Cassell et al.
265
to a number of other stakeholders (e.g., audit committees, investors, etc.). Unfortunately,
researchers cannot provide such evidence because we are only able to observe audit outcomes (e.g., misstatements, accruals quality, etc.). Data on audit inputs (e.g., the number of
audit personnel and hours, experience levels of the engagement team, etc.) are not publicly
available. Although we cannot directly measure the length or steepness of the audit learning
curve due to these data limitations, we can identify situations in which constraints on the
time available to the auditor may influence the level of assurance provided early in the
auditor–client relationship. Specifically, we examine whether the timing of the successor
auditor’s engagement constrains audit production and affects audit quality in the year of
the auditor change. We also investigate whether client complexity affects the impact of the
timing of an auditor change on audit quality.
We identify 76,430 company-year observations from 2000 through 2014 with sufficient
data to construct the variables in our models (7,715 of which involve an auditor change).
Given the timing of a typical audit and the related quarterly reviews, we classify auditor
changes that occur before the end of the third fiscal quarter as ‘‘early changes’’ (4,916
observations) and those that occur during or after the fourth fiscal quarter as ‘‘late
changes’’ (2,799 observations). Univariate tests suggest that companies with late auditor
changes differ from companies with early auditor changes and companies not changing
auditors across a broad array of company risk characteristics. As such, we use propensityscore matching (PSM) to create three matched samples (companies changing auditors late
matched with companies changing auditors early, companies changing auditors early
matched with nonauditor change companies, and companies changing auditors late matched
with nonauditor change companies).
We use financial statement misstatements (as revealed through subsequent restatements)
as our proxy for audit quality. Revealed misstatements provide direct evidence that the
auditor failed to detect and/or report a material misstatement in the financial statements,
and a recent survey of investors and auditors suggests that restatements are the leading indicator of low audit quality (Christensen, Glover, Omer, & Shelley, 2016). In multivariate
tests, we find no evidence that companies engaging a new auditor early in the year are
more likely to misstate than companies that do not change auditors. However, companies
that change auditors late in the year are significantly more likely than companies that
change auditors early in the year and companies that do not change auditors to misstate. In
subsequent tests, we find no evidence that the timing of the prior year’s auditor change
affects the likelihood of misstatement in the subsequent year.
Next, we investigate situations where we expect the effect of the timing of auditor
changes on audit quality to be especially pronounced. Specifically, we expect client complexity to affect audit production constraints because the planning and risk assessment process requires more time and effort as client complexity increases. Using operating
segments to proxy for client complexity, we find that the reduction in audit quality
observed among companies that change auditors late in the year is more pronounced for
companies with more operating segments. In contrast, we find no evidence of decreased
audit quality among complex and noncomplex companies that change auditors early in the
year. Collectively, our results suggest that the extent to which audit quality suffers in the
first year of audit engagements is affected by both (a) the amount of time required to
understand the client’s business, assess risks, and perform the audit (all of which are driven
by client complexity), and (b) the amount of time available for auditors to perform these
tasks.
266
Journal of Accounting, Auditing & Finance
To further strengthen the inferences from our primary tests, we use the demise of Arthur
Andersen as an exogenous shock to test our predictions. As the survival of Andersen
became uncertain, clients began to change auditors. Importantly, these auditor–client realignments differ from normal transitions in that they were necessary rather than voluntary.
As a result, client characteristics such as profitability, risk, and growth (among others)
were likely less important in initiating the change and determining the timing of the
change. Thus, these tests provide us with an alternative setting in which to isolate the
effect of the timing of auditor changes on audit quality. We reperform our tests after limiting the sample to auditor changes occurring during 2002 and 2003 as a result of the demise
of Arthur Andersen. We find that although engagements occurring late in the fiscal year
are not associated with misstatements on average, there is a higher likelihood of misstatement when the successor auditor is engaged by a client with complex operations late in the
year, consistent with our primary findings.
Finally, we investigate whether time constraints in first-year audit engagements result in
audit production inefficiencies, where audit production inefficiencies are reflected in higher
audit costs. Because audit costs (e.g., hours worked, personnel mix, etc.) are unobservable,
we use audit fees to proxy for audit costs. Consistent with our inference that our results are
likely attributable to audit production inefficiencies, we find that audit fees are higher for
clients that engage auditors late in the year relative to clients that engage auditors early in
the year. However, we urge caution in the interpretation of these results because of the difficulties associated with using audit fees to proxy for audit production costs in our setting.
Our study contributes to the literature that finds a relation between audit quality and
auditor tenure. Although this research provides evidence that audit quality improves as
auditor tenure lengthens, some of these studies omit the first year of auditor tenure (e.g.,
Johnson, Khurana, & Reynolds, 2002; Myers, Myers, & Omer, 2003) while others do not
(e.g., Carcello & Nagy, 2004; Chi, Myers, Omer, & Xie, 2017; Jenkins & Velury, 2008;
Stanley & DeZoort, 2007). Other studies examining the first year of an auditor–client relationship fail to find a significant decline in financial reporting quality (e.g., Carver,
Hollingsworth, & Stanley, 2011; DeFond & Subramanyam, 1998; Nagy, 2005) except
when companies switch to smaller audit firms. However, none of these studies consider the
timing of the new auditor’s engagement. In contrast, we show that audit quality is lower in
the first year of auditor tenure when the successor auditor is engaged late in the year, particularly when client operations are complex.
We expect our results to be of interest to regulators who have proposed various remedies
to improve auditor independence and skepticism. For example, the European Parliament
recently implemented rules that require public companies in the European Union (EU) to
rotate auditors every 10 to 24 years. This new regulation will affect not only European
companies but also the United States and other cross-listed multinationals.3 Our study also
answers calls from the PCAOB for evidence on the possible effects of audit firm rotation
on audit quality. For example, in their concept release on auditor independence and audit
firm rotation, the PCAOB specifically seeks commenters’ views on the following questions
(PCAOB, 2011):
Does audit effectiveness vary over an auditor’s tenure on a particular engagement? For example, are auditors either more or less effective at the beginning of a new client relationship? If
there is a ‘‘learning curve’’ before auditors can become effective, how long is it, and does it
vary significantly by client type? (p. 20)
Cassell et al.
267
Our results suggest that possible negative effects of regulatory changes which would
result in more frequent auditor changes could be mitigated if these regulatory changes
include restrictions related to the timing of auditor changes. In addition, our results should
be of interest to investors, audit committees, auditors, and other stakeholders interested in
the impact of auditor changes and the length of the auditor–client relationship on audit
quality.
The remainder of the article is organized as follows. We discuss prior research and
make predictions in ‘‘Prior Literature and Predictions’’ section. We provide a description
of our empirical methods in ‘‘Research Design and Methodology’’ section. We describe the
sample and present our results in ‘‘Sample Selection and Empirical Results’’ section. We
discuss additional tests in ‘‘Additional Tests’’ section. The final section summarizes the
key results and provides conclusions.
Prior Literature and Predictions
Prior research suggests a positive association between financial reporting quality and auditor tenure. For example, Johnson et al. (2002) find lower financial reporting quality (i.e.,
higher absolute value of unexpected accruals and less persistent accruals) when auditor
tenure is short and Carcello and Nagy (2004) find a greater likelihood of fraud when auditor tenure is short (both define short tenure as three years or less). Similarly, Myers et al.
(2003) find a negative association between auditor tenure and the magnitude of discretionary accruals, Stanley and DeZoort (2007) find a negative relation between auditor tenure
and financial restatements, and Jenkins and Velury (2008) find a positive relation between
auditor tenure and conservative financial reporting.
Although prior work suggests that audit quality is compromised during the early years
of the auditor–client relationship, prior research examining the effect of auditor changes on
audit quality only finds differences in the first year of an auditor engagement in specific
settings. DeFond and Subramanyam (1998) find that discretionary accruals are not significantly different from zero in the first year of tenure. Nagy (2005) finds that discretionary
accruals of smaller clients forced to switch auditors due to Arthur Andersen’s demise are
less extreme in the first year with the successor auditor but this relation does not exist for
larger clients. Carver et al. (2011) find no change in discretionary accruals of companies
changing to an auditor of similar size, but find that companies changing to a smaller auditor report a significant increase in signed discretionary accruals in the 2 years following the
change. Overall, the results of these studies do not suggest a deterioration in audit quality
in the first year of a new audit engagement, except when companies switch to a smaller
audit firm.
Although prior research suggests a positive association between financial reporting quality and auditor tenure, strong inferences about audit quality in the first year of an auditor–
client relationship cannot be drawn from these studies because some exclude the initial
year of the engagement. Furthermore, prior research that specifically examines financial
reporting quality in the first year of an auditor–client relationship does not consider the
timing of the new auditor’s engagement.
In contrast to prior work, we investigate whether audit quality varies with the timing of
the successor auditor’s appointment in the initial year of the engagement. We posit that the
timing of an auditor change is crucial because a high-quality audit requires knowledge
about the client and its business, an appropriate assessment of the risk of material
268
Journal of Accounting, Auditing & Finance
misstatement, and an appropriate response to this risk. Importantly, these tasks require sufficient time to complete and time may be too constrained in certain situations. Time constraints can increase the difficulty of coordination and communication within the audit
team and with client personnel, increasing the likelihood of errors or pressures to forgo
obtaining sufficient audit evidence. As discussed above, it may not be possible to overcome
time constraints because all public companies must file audited financial statements with
the SEC by the filing deadline and because of the complex nature of certain audit tasks
(e.g., for some audit tasks, additional resources in the form of more personnel cannot substitute for auditor learning accumulated over time).4 Thus, we argue that auditors engaged
late in the year are less likely to have a sufficient amount of time to appropriately understand the client’s business, plan the audit, assess client risks, and execute audit procedures.
This leads to our first prediction:
Prediction 1: Audit quality will be lower in the year of an initial audit engagement
when the engagement of a successor auditor occurs late in the year.
We predict no reduction in audit quality when companies engage new auditors earlier in
the year. Thus, our second prediction is as follows:
Prediction 2: Audit quality will not differ in the year of an initial audit engagement
when the engagement of a successor auditor occurs early in the year.
Finally, we argue that the amount of time necessary for auditors to obtain an understanding of the company and its risks should vary based on the complexity of the company’s
operations. Related to this, Ashton, Graul, and Newton (1989) find that client complexity is
positively associated with audit delay, suggesting that audits of complex companies require
more time and effort. As such, auditors engaged late in the year by companies with more
complex operations may be subject to even greater audit risk because incoming auditors
have less time to understand the company and identify and assess client risks. This leads to
our third prediction:
Prediction 3: Client complexity affects the relation between the timing of an initial
audit engagement and audit quality in the year of the engagement.
Research Design and Methodology
We perform tests using financial statement misstatements (as revealed through subsequent
restatements) as our proxy for audit quality. A growing body of literature uses misstatements to proxy for audit quality (e.g., Francis & Michas, 2013; McGuire, Omer, & Sharp,
2012; Schmidt, 2012; Stanley & DeZoort, 2007) and survey results from Christensen et al.
(2016) support its use. Specifically, Christensen et al. (2016) find that restatements are
viewed as the leading indicator of poor audit quality by both auditors and investors.
Some prior research examining material misstatements focuses primarily on particularly
egregious cases, such as SEC enforcement actions related to fraudulent financial reporting.
However, such cases are infrequent and suggest very low rates of audit failure. Although
SEC enforcement actions are evidence of audit failure, we conjecture that any error or misapplication of Generally Accepted Accounting Principles (GAAP) represents a failure by
Cassell et al.
269
the auditors to detect and report the misstatement, regardless of management’s intent.
Furthermore, using financial statement misstatements rather than solely egregious misstatements results in a larger sample of low quality audits.5
To test our first prediction, we estimate the following logistic regression model:
PrðMISSTATEit = 1Þ = b0 + b1 LATEit + b2 GCOit + b3 LOSSit + b4 LEVERAGEit
+ b5 ROAit + b6 LATEFILERit1 + b7 ICMWit + b8 M&Ait + b9 NEWFINANCEit
+ b10 SIZEit + b11 MTBit + b12 BIGNit + b13 TENUREit + b14 SPECIALISTit
ð1Þ
+ b15 COMPLEXITYit + b16 RESIGNit + b17 DOWNWARDit + b18 UPWARDit
+ bm Industry FE + bn Year FE,
where MISSTATE = an indicator variable set equal to one if the year t annual financial
statements were misstated (as revealed through a subsequent restatement), and zero otherwise; LATE = an indicator variable set equal to one if a new auditor was engaged during
or after the fourth fiscal quarter, and zero otherwise; and all other variables are as defined
in the appendix. The coefficient of interest in Equation 1 is b1, which indicates whether the
likelihood of misstatement differs for companies that engage a new auditor: (a) late in the
year, relative to early in the year, or (b) late in the year, relative to companies that do not
change auditors, depending on which matched sample is being used.6
We include a number of controls for financial health and performance (i.e., receipt of
a going-concern opinion, losses, leverage, and return on assets) because prior research
generally finds a positive association between misstatements and financial distress
(Blankley, Hurtt, & MacGregor, 2012; Cao, Myers, & Omer, 2012). In addition, regulators suggest that financial reporting issues can lead to future restatements (PCAOB,
2011), and Blankley et al. (2012) find that internal control weaknesses are positively
associated with future restatements. As such, we control for whether the prior-year financial statements were filed after the required deadline and whether a material weakness in
internal control over financial reporting was reported. We also control for transactions
that could influence the likelihood of a misstatement. Following prior research, we control for mergers and acquisitions since they can increase the likelihood of misstatement
due to the integration of systems and business units and because of the complexity of the
accounting required to record these transactions.7 Following Dechow, Ge, Larson, and
Sloan (2011), we control for new financing, and consistent with Cao et al. (2012), we
control for company size and market-to-book ratio. We also control for auditor characteristics that may influence the likelihood of a misstatement. Specifically, we include controls for auditor size, industry specialization, and tenure because prior research suggests
that these auditor characteristics can improve audit quality.8 We control for client complexity because prior research suggests that the likelihood of misstatement is increasing
in client complexity (Cao et al., 2012). In tests using the matched sample of late and
early auditor changes, we include three additional variables to control for variation in the
nature of the auditor change other than the timing of the change (i.e., resignation versus
dismissal and upward versus downward versus lateral change).9 Finally, we include
industry and year fixed effects to control for variation in the frequency of misstatements
across industries and over time, and we cluster standard errors by company to control for
serial dependence (Petersen, 2009).
270
Journal of Accounting, Auditing & Finance
To test our second prediction, we estimate the following logistic regression model:
PrðMISSTATEit = 1Þ = O0 + O1 EARLYit + O2 GCOit + O3 LOSSit + O4 LEVERAGEit
+ O5 ROAit + O6 LATEFILERit1 + O7 ICMWit + O8 M&Ait + O9 NEWFINANCEit
+ O10 SIZEit + O11 MTBit + O12 BIGNit + O13 TENUREit + O14 SPECIALISTit
ð2Þ
+ O15 COMPLEXITYit + Om Industry FE + On Year FE,
where EARLY = an indicator variable set equal to one if a new auditor was engaged prior
to the start of the fourth fiscal quarter (i.e., in fiscal quarters one through three), and zero
otherwise; and all other variables are as previously defined. Model variables are consistent
with Equation 1 except that we exclude controls for the type of auditor switch (RESIGN,
DOWNWARD, UPWARD) because the control group consists of clients that do not
change auditors. The coefficient of interest in Equation 2 is O1, which indicates whether
the likelihood of misstatement differs for companies that engage a new auditor early in the
year, relative to companies that do not change auditors.
To test our third prediction, we use the number of operating segments to proxy for client
complexity.10 In many cases, operating segments represent different product and/or service
lines, which can result in clients being subject to different revenue recognition requirements,
having different inventory management processes, more complex consolidation procedures,
and so on. In addition, companies with more than one material operating segment are subject
to increased disclosure requirements. Consistent with our third prediction, we expect that
these complexities also affect the amount of time required for the auditor to plan and execute
the audit. We use the following logistic regression model to test our prediction:
PrðMISSTATEit = 1Þ = a0 + a1 TIMINGit + a2 TIMINGit 3COMPLEXITYit + a3 GCOit
+ a4 LOSSit + a5 LEVERAGEit + a6 ROAit + a7 LATEFILERit1 + a8 ICMWit
+ a9 M&Ait + a10 NEWFINANCEit + a11 SIZEit + a12 MTBit + a13 BIGNit
+ a14 TENUREit + a15 SPECIALISTit + a16 COMPLEXITYit + a17 RESIGNit
+ a18 DOWNWARDit + a19 UPWARDit + am Industry FE + an Year FE,
ð3Þ
where TIMING = either EARLY or LATE; and all other variables are as previously
defined. The coefficient of interest in Equation 3 is a2, the interaction between
COMPLEXITY and TIMING (where TIMING is either EARLY or LATE, depending on
which matched sample is being used). This coefficient indicates whether client complexity
influences the effect (if any) of the timing of an auditor’s engagement on the likelihood of
a misstatement in the year of an auditor change. All control variables are consistent with
those in Equations 1 and 2 depending on which matched sample is used.
To construct our first matched sample based on the propensity to change auditors late in
the year, we first estimate the following model:
PrðLATEit = 1Þ = d0 + d1 GCOit + d2 LOSSit + d3 LEVERAGEit + d4 ROAit
+ d5 LATEFILERit1 + d6 ICMWit + d7 M&Ait + d8 NEWFINANCEit + d9 SIZEit
+ d10 MTBit + d11 BIGNit + d12 SPECIALISTit + d13 COMPLEXITYit + d14 RESIGNit
+ d15 DOWNWARDit + d16 UPWARDit ,
ð4Þ
where all variables are as previously defined.
Cassell et al.
271
To construct our matched samples where early or late auditor changes are matched with
companies that do not change auditors, we estimate the probability of an early or late auditor change as in Equation 4, after excluding variables for the type of auditor switch
(RESIGN, DOWNWARD, UPWARD) because these variables are not defined for companies that do not change auditors.
Equation 4 includes several auditor and client risk characteristics that could influence
the timing of an auditor change including financial distress as indicated by a going-concern
opinion, reported losses, leverage, return on assets, prior period financial reporting delays,
the presence of a material weakness in internal controls, merger and acquisition activity,
new debt or equity financing, company size, market-to-book, whether the company is
audited by a Big N or industry specialist auditor, and client complexity.
For each of the three PSM samples, we use the estimated propensity scores and match
(without replacement) within a maximum caliper distance of 3%.11 For the first PSM sample,
we estimate Equation 4 using all available observations where there was an auditor change
late in the year and all available observations where there was an auditor change early in the
year. The dependent variable is LATE. This results in 1,807 successful matches for a total of
3,614 observations. For the second PSM sample, we estimate Equation 4 using all available
observations where there was an auditor change early in the year and all available nonauditor
change observations. The dependent variable is EARLY. This results in 4,904 successful
matches for a total of 9,808 observations. Finally, for the third PSM sample, we estimate
Equation 4 using all available observations where there was an auditor change late in the
year and all available nonauditor change observations. The dependent variable is LATE. This
results in 2,746 successful matches for a total of 5,492 observations.
Sample Selection and Empirical Results
Sample Selection
As outlined in Table 1, we obtain data on restatements and auditor changes from Audit
Analytics for the years 2000 through 2014. We merge Audit Analytics data with financial
data from the Compustat Fundamentals Annual and Compustat Segments databases. After
excluding observations with missing data needed to construct our variables and auditor
changes that occurred as the result of an audit firm merger, we have 76,430 available company-year observations.
Included in this sample are 7,715 auditor change observations: 4,916 new auditor
engagements occur before the end of the third fiscal quarter and 2,799 occur during or after
the fourth fiscal quarter. Table 1 also presents auditor changes by year and by group (i.e.,
early versus late). We find a large number of auditor changes in 2002 (which coincides
with the collapse of Arthur Andersen) and a declining trend subsequent to the passage of
the Sarbanes-Oxley Act of 2002 and the adoption of important sections of the Act related
to internal controls. In general, there are more early auditor changes than late auditor
changes, but the percentage of auditor change observations in each category fluctuates in a
nonsystematic manner throughout the sample period.
Descriptive Statistics and Univariate Results
Table 2 presents descriptive statistics for the sample of available observations. We winsorize all continuous variables at the 1/99 percent level to mitigate the influence of outliers.
272
EARLY
LATE
TOTAL
120
107
227
2000
249
306
555
2001
997
322
1,319
2002
335
284
619
2003
438
329
767
2004
485
220
705
2005
401
211
612
2006
329
153
482
2007
243
132
375
2008
Auditor changes by year
291
132
423
2009
Observations from 2000 through 2014 with data from Audit Analytics and Compustat
Less observations with missing data needed to construct our variables
Less observations with auditor changes due to an audit firm merger
Full available sample
Sample observations with auditor changes occurring before end of third quarter (Early changes)
Sample observations with auditor changes occurring during or after the fourth quarter (Late changes)
Auditor change observations included in full available sample
Table 1. Sample Selection and Composition.
241
132
373
2010
189
129
318
2011
186
150
336
2012
256
140
396
2013
156
52
208
2014
83,104
6,457
217
76,430
4,916
2,799
7,715
Sample
4,916
2,799
7,715
Total
Cassell et al.
273
Table 2. Descriptive Statistics.
Variable
MISSTATE
EARLY
LATE
GCO
LOSS
LEVERAGE
ROA
LATEFILERt-1
ICMW
M&A
NEWFINANCE
SIZE
MTB
BIGN
TENURE
SPECIALIST
COMPLEXITY
RESIGN
DOWNWARD
UPWARD
N
M
SD
25th percentile
Median
75th percentile
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
76,430
0.106
0.064
0.037
0.129
0.407
0.366
–0.450
0.116
0.072
0.103
0.844
5.302
2.205
0.614
6.910
0.189
0.056
0.024
0.030
0.008
0.307
0.245
0.188
0.335
0.491
0.900
2.016
0.320
0.258
0.304
0.363
2.842
10.623
0.487
3.843
0.392
0.230
0.152
0.170
0.090
0.000
0.000
0.000
0.000
0.000
0.020
–0.119
0.000
0.000
0.000
1.000
3.550
0.856
0.000
4.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.168
0.010
0.000
0.000
0.000
1.000
5.631
1.633
1.000
6.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
1.000
0.377
0.055
0.000
0.000
0.000
1.000
7.261
3.049
1.000
10.000
0.000
0.000
0.000
0.000
0.000
Note. All variables are defined in the appendix.
The distributions of our control variables are consistent with those in prior work.
Approximately 11% of sample observations have misstatements (as revealed through subsequent restatements) and 5.6% of sample observations have multiple operating segments.
Table 3 presents univariate tests comparing the mean values of our model variables for
observations with new auditor engagements that occur late in the year, observations with
new auditor engagements that occur early in the year, and nonauditor change observations.
Here, we find systematic differences across the three groups in terms of company size,
financial health, profitability, growth, auditor characteristics, risk, and complexity, highlighting the importance of using matched samples to investigate whether the timing of a
new auditor’s engagement affects audit quality in the year of an auditor change.
Table 4 presents the results from estimating Equation 3, our first-stage PSM model.
Column 1 presents results for the first matched sample (early and late auditor change observations), column 2 presents results for the second matched sample (early and nonauditor
change observations), and column 3 presents results for the third matched sample (late and
nonauditor change observations).
In column 1, we find that, relative to companies changing auditors early in the year, the
likelihood of a late auditor change increases with financial distress (i.e., a going-concern
report modification), prior period financial reporting delays, and when the predecessor
auditor resigns. The likelihood of a late auditor change is lower for companies obtaining
new debt or equity financing, larger companies, companies engaging a Big N auditor, and
companies switching to a smaller auditor. In column 2, we find that, relative to companies
not changing auditors, the likelihood of an early auditor change increases with the reporting
of a loss, return on assets, prior period financial reporting delays, the presence of internal
274
0.124
0.373
0.631
0.693
21.601
0.344
0.180
0.049
0.751
2.935
0.357
0.235
0.077
0.021
0.308
0.267
0.074
0.109
0.201
0.514
0.435
20.732
0.207
0.126
0.068
0.824
4.276
1.727
0.391
0.121
0.041
0.194
0.309
0.084
0.015*
0.172***
0.117***
0.258***
20.870***
0.137***
0.055***
20.018***
20.073***
21.341***
21.370***
20.156***
20.045***
20.019***
0.114***
20.042***
20.009
0.109
0.201
0.514
0.435
20.732
0.207
0.126
0.068
0.824
4.276
1.727
0.391
0.121
0.041
EARLY
(n = 4,916)
M
0.105
0.114
0.390
0.348
20.383
0.100
0.064
0.107
0.849
5.472
2.315
0.646
0.199
0.058
Diff. in
mean
0.005***
0.087***
0.124***
0.088***
20.350***
0.108***
0.062***
20.040***
20.025***
21.196***
20.588***
20.254***
20.077***
20.018***
No Change
(n = 68,715)
M
Diff. in
mean
EARLY
(n = 4,916)
M
Note. All variables are defined in the appendix.
p values are two-tailed.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
MISSTATE
GCO
LOSS
LEVERAGE
ROA
LATEFILERt – 1
ICMW
M&A
NEWFINANCE
SIZE
MTB
BIGN
SPECIALIST
COMPLEXITY
RESIGN
DOWNWARD
UPWARD
Variable
LATE
(n = 2,799)
M
EARLY vs. No change
LATE vs. EARLY
Table 3. Univariate Tests—Full Available Samples.
0.124
0.373
0.631
0.693
21.601
0.344
0.180
0.049
0.751
2.935
0.357
0.235
0.077
0.021
LATE
(n = 2,799)
M
0.105
0.114
0.390
0.348
20.383
0.100
0.064
0.107
0.849
5.472
2.315
0.646
0.199
0.058
No Change
(n = 68,715)
M
LATE vs. No change
0.019***
0.259***
0.241***
0.345***
21.218***
0.244***
0.116***
20.058***
20.098***
22.537***
21.958***
20.411***
20.122***
20.037***
Diff. in
mean
275
.017
\.001
.756
.275
.651
\.001
.164
.541
\.001
\.001
.514
.003
.177
.280
\.001
.005
.537
20.211**
0.284***
20.018
20.027
20.005
0.297***
0.098
20.066
20.234***
20.087***
20.001
20.232***
0.131
20.167
0.413***
20.168***
0.058
7,715
2,799
4,916
0.655
Intercept
GCO
LOSS
LEVERAGE
ROA
LATEFILERt – 1
ICMW
M&A
NEWFINANCE
SIZE
MTB
BIGN
SPECIALIST
COMPLEXITY
RESIGN
DOWNWARD
UPWARD
N
N LATE
N EARLY
Area under the ROC curve
0.788
4,916
0.684
21.848***
0.217***
0.134***
20.050***
0.004
0.613***
0.587***
20.064
20.074
20.086***
20.001
21.306***
0.183**
0.000
\.001
.011
\.001
.049
.078
\.001
\.001
.160
\.001
.002
.200
\.001
.243
.053
71,514
2,799
Coefficient
p value
73,631
21.690***
20.131**
0.146***
20.036**
0.015*
0.353***
0.497***
20.085
0.166***
20.026***
20.002
20.977***
0.059
0.149*
Coefficient
DV = LATE
\.001
\.001
.007
.004
.654
\.001
\.001
.485
.125
\.001
.237
\.001
.030
.999
p value
LATE and matched no change
Column 3
Note. This table presents results from estimating Equation 4. All variables are defined in the appendix. Standard errors are clustered by company. PSM = propensity-score
matching; ROC = receiver operating characteristic curve.
p values are two-tailed.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively.
p value
DV = EARLY
DV = LATE
Coefficient
EARLY and matched no change
LATE and matched EARLY
Variable
Column 2
Column 1
Table 4. PSM Model—Determinants of Late/Early Changes.
276
Journal of Accounting, Auditing & Finance
control material weaknesses, new debt or equity financing, and multiple operating segments. The likelihood of an early auditor change is lower for companies in financial distress, and for companies that are highly leveraged, larger, and subsequently audited by a
Big N auditor. In column 3, we find that, relative to companies not changing auditors, the
likelihood of a late auditor change increases with financial distress, the reporting of a loss,
prior period financial reporting delays, the presence of internal control material weaknesses,
and the use of an industry specialist auditor. The likelihood of a late auditor change is
lower for companies that are highly leveraged, larger, and subsequently audited by a Big N
auditor.
Table 5 presents univariate tests comparing the mean values of our model variables for
each of the three PSM samples. For the first matched sample (early and late auditor change
observations), the only significant difference is that late auditor change observations have a
higher incidence of internal control material weaknesses. For the second matched sample
(early auditor change and nonauditor change observations), we find that early auditor
change companies are larger, more likely to engage in merger and acquisition activity,
have a higher return on assets, and have a higher incidence of internal control material
weaknesses. Early auditor change companies also have a lower incidence of going concern
opinions, report fewer losses, and have lower leverage. For the third matched sample (late
auditor change and nonauditor change observations), we find that late auditor change companies have a higher incidence of going concern opinions, prior period financial reporting
delays, and internal control material weaknesses, and have lower return on assets.
Collectively, the results in Table 5 suggest that our matching procedures are partially
successful in mitigating the systematic differences across the three groups observed in the
full sample. There are far fewer significant differences in Table 5 than in Table 3, and the
magnitude of the differences is generally much lower. Importantly, our second-stage model
includes controls for each of these variables to further mitigate concerns related to residual
differences in company characteristics across the three groups.
Multivariate Results
Table 6 presents the results from estimating Equations 1 and 2. Column 1 presents the
results using our first matched sample, while columns 2 and 3 present the results using our
second and third matched samples, respectively. In column 1, the coefficient on LATE is
positive and significant (p value = .001), revealing that, relative to early auditor changes,
there is a greater likelihood of misstatement when a new auditor is engaged during or after
the fourth fiscal quarter. In column 2, the coefficient on EARLY is not statistically significant (p value = .929), revealing that the likelihood of misstatement is not significantly different for companies that do not change auditors versus companies that change auditors early in
the year. In column 3, the coefficient on LATE is positive and significant (p value = .091),
revealing that relative to nonauditor changes, the likelihood of misstatement is greater when
an auditor is engaged late in the year. Collectively, the results in Table 6 are consistent with
Predictions 1 and 2 and suggest that the timing of a new auditor’s engagement has a significant effect on audit quality.
Table 7 presents the results from estimating Equation 3 to test Prediction 3. Consistent
with Table 6, column 1 presents the results using the first matched sample, while columns
2 and 3 present the results using our second and third matched samples, respectively. In
column 1, we find a positive and significant coefficient on LATE (p value = .004) and a
positive and significant coefficient on the interaction between LATE and COMPLEXITY
277
0.243
0.553
0.490
20.942
0.259
0.163
0.061
0.792
3.758
1.553
0.297
0.099
0.028
0.232
0.315
0.090
Diff. in mean
0.263
0.564
0.465
20.892
0.255
0.141
0.060
0.775
3.869
1.494
0.320
0.113
0.032
0.255
0.328
0.089
20.019
20.011
0.025
20.050
0.004
0.022*
0.001
0.017
20.111
0.059
20.023
20.014
20.004
20.022
20.012
0.002
0.200
0.513
0.436
20.734
0.206
0.125
0.068
0.823
4.279
1.713
0.392
0.122
0.040
0.227
0.531
0.514
20.847
0.214
0.106
0.053
0.809
4.071
1.488
0.398
0.137
0.036
20.027***
20.018*
20.078***
0.113**
20.008
0.018***
0.015***
0.014
0.208***
0.224
20.006
20.016
0.004
Diff. in mean
No change
(n = 4,904)
M
EARLY
(n = 4,904)
M
EARLY
(n = 1,807)
M
Note. All variables are defined in the appendix. PSM = propensity-score matching.
p values are two-tailed.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
GCO
LOSS
LEVERAGE
ROA
LATEFILERt – 1
ICMW
M&A
NEWFINANCE
SIZE
MTB
BIGN
SPECIALIST
COMPLEXITY
RESIGN
DOWNWARD
UPWARD
Variable
LATE
(n = 1,807)
M
EARLY vs. No change
LATE vs. EARLY
Table 5. Univariate Tests—PSM Sample.
0.364
0.625
0.684
21.522
0.334
0.174
0.050
0.752
3.008
0.747
0.240
0.078
0.022
LATE
(n = 2,746)
M
0.322
0.617
0.622
21.307
0.307
0.141
0.051
0.767
3.112
1.456
0.236
0.080
0.026
No change
(n = 2,746)
M
0.042***
0.008
0.061
20.215**
0.027**
0.033***
20.001
20.015
20.104
20.709
0.004
20.002
20.004
Diff. in mean
LATE vs. No change
278
Journal of Accounting, Auditing & Finance
(p value = .060), suggesting that the adverse effects of late auditor changes on audit quality
are more pronounced in companies with complex operations. The joint test (LATE +
LATE 3 COMPEXITY) is positive and significant (p = .034). In column 2, we find an
insignificant coefficient on EARLY (p value = .783) and an insignificant coefficient on the
interaction between EARLY and COMPLEXITY (p value = .210), suggesting that there is
no difference in the likelihood of misstatement between companies engaging an auditor
early in the year relative to companies not changing auditors, even when client operations
are complex. Moreover, the joint test (EARLY + EARLY 3 COMPEXITY) is insignificant. In column 3, we do not find a significant coefficient on LATE (p value = .191) but
we find a positive and significant coefficient on the interaction between LATE and
COMPLEXITY (p value = .022), suggesting that relative to companies not changing auditors, the adverse effects of late auditor changes on audit quality are more pronounced for
companies with complex operations. Moreover, the joint test (LATE + LATE 3
COMPEXITY) is positive and significant (p = .026).12
Collectively, the results in Tables 6 and 7 reveal (a) an increased likelihood of misstatement when the auditor is engaged late in the year, and (b) the effect of late auditor changes
on the likelihood of misstatements is more pronounced for companies with complex operations. A key inference from these results is that we should not observe a deterioration in
audit quality in the second year of an auditor’s tenure, regardless of the complexity of
client operations.13 To investigate whether this inference is supported by our data, we
examine whether the likelihood of misstatement differs in the second year of auditor tenure
for companies that engage a new auditor (a) late in the previous year, relative to early in
the previous year; (b) early in the previous year, relative to companies that do not change
auditors in the previous year; or (c) late in the previous year, relative to companies that do
not change auditors in the previous year. To do this, we use the subsequent year observations for companies in our PSM matched samples. Thus, these samples are limited to
matched pairs of second-year auditor tenure observations with available data.
The three resulting matched samples are comprised of 2,984, 8,296, and 4,686 observations, respectively. For the first and third matched samples, we modify Equation 1 by replacing LATE with YR2LATE (an indicator variable set equal to one if a new auditor was
engaged in the prior year during or after the fourth fiscal quarter, and zero otherwise). For
the second matched sample, we modify Equation 1 by replacing EARLY with YR2EARLY
(an indicator variable set equal to one if a new auditor was engaged in the prior year before
the end of the third fiscal quarter, and zero otherwise). For the first and third matched samples, the coefficients on YR2LATE indicate whether the likelihood of misstatement differs
for second-year audits where the auditor change occurred late in the previous year, relative
to the matched observations (i.e., companies changing auditors early in the year or companies not experiencing an auditor change). For the second matched sample, the coefficient
on YR2EARLY indicates whether the likelihood of misstatement differs for second-year
audits where the auditor change occurred early in the previous year, relative to the matched
observations.
In untabulated tests, we find that the coefficients on YR2LATE and YR2EARLY
are not significantly different from zero. Taken together, these results reveal that the likelihood of misstatement in the second year of auditor tenure is not influenced by the timing
of an auditor change, consistent with what should be expected given the results in Tables 6
and 7.
279
Predictions
23.460***
0.008
\.001
.001
.473
.079
.179
.609
.099
\.001
.011
\.001
.007
.352
.179
.468
.088
.268
\.001
.361
23.610***
0.340***
0.123
0.189*
0.060
0.015
0.180*
0.615***
0.468**
0.617***
0.092***
0.004
20.166
20.015
0.392*
0.144
20.898***
0.163
Included
Included
3,614
0.688
Included
Included
9,808
0.668
0.024
0.139**
20.036
20.002
0.352***
0.824***
0.258**
0.241***
0.068***
0.000
0.097
20.008
20.121
20.029
Coefficient
p value
.833
.042
.799
.923
\.001
\.001
.041
.009
.002
.928
.834
.292
.134
.563
\.001
.929
p value
DV = MISSTATE
DV = MISSTATE
Coefficient
EARLY and Matched No Change
LATE and Matched EARLY
Included
Included
5,492
0.663
0.129*
0.015
0.008
0.026
0.028
0.303***
0.696***
0.373**
0.406***
0.045*
20.003
0.258
0.002
20.038
0.484**
23.657***
Coefficient
DV = MISSTATE
.091
.905
.473
.274
.172
.001
\.001
.022
\.001
.083
.389
.964
.537
.414
.028
\.001
p value
LATE and Matched No Change
Column 3
Note. This table presents results from estimating Equations 1 and 2, depending on the matched sample. All variables are defined in the appendix. Standard errors are clustered
by company. p values are one-tailed if a sign is predicted, and two-tailed otherwise. ROC = receiver operating characteristic curve; ?: two tailed p-values.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
Intercept
?
EARLY
?
LATE
+
GCO
?
LOSS
+
LEVERAGE
+
ROA
?
LATEFILERt – 1
+
ICMW
+
M&A
+
NEWFINANCE
+
SIZE
?
MTB
?
BIGN
TENURE
SPECIALIST
COMPLEXITY
+
RESIGN
?
DOWNWARD
?
UPWARD
?
Industry FE
Year FE
N
Area under the ROC curve
Variable
Column 2
Column 1
Table 6. The Effect of Auditor Engagement Timing on Audit Quality.
280
?
?
?
+
+
Predictions
23.410***
0.024
–0.448
\.001
.004
.060
.034
23.595***
0.303***
0.897*
Included
Included
Included
3,614
0.690
4.516**
1.447
Included
Included
Included
9,808
0.669
Coefficient
p value
.229
\.001
.783
.210
p value
DV = MISSTATE
DV = MISSTATE
Coefficient
EARLY and
Matched no change
LATE and matched
EARLY
4.963**
0.086
1.024**
Included
Included
Included
5,492
0.664
23.643***
Coefficient
.026
.191
.022
\.001
p value
DV = MISSTATE
LATE and matched
no change
Column 3
Note. This table presents results from estimating Equation 3. All variables are defined in the appendix. Reported standard errors are clustered by company. p values are onetailed if a sign is predicted, and two-tailed otherwise. ROC = receiver operating characteristic curve; ?: two tailed p-values.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
Intercept
EARLY
EARLY 3 COMPLEXITY
LATE
LATE 3 COMPLEXITY
Controls
Industry FE
Year FE
N
Area under the ROC curve
x2 test:
EARLY + EARLY 3 COMPLEXITY = 0
LATE + LATE 3 COMPLEXITY = 0
Variable
Column 2
Column 1
Table 7. The Effect of Auditor Engagement Timing on Audit Quality—Client Complexity.
Cassell et al.
281
Additional Tests
Arthur Andersen’s Demise
To confirm the inferences from our primary tests, we use the demise of Arthur Andersen as
an exogenous shock to test our predictions. As noted in Table 1, there was an abnormally
high level of auditor turnover in 2002 related to uncertainty about whether Arthur
Andersen would continue as a viable audit firm. As the collapse of the firm became evident, Arthur Andersen clients began to engage other auditors and these auditor transitions
differ from normal voluntary transitions. That is, client characteristics such as profitability,
risk, and growth (among others) were likely less important in initiating the change and
determining the timing of the change. Thus, these tests provide a strong setting in which to
isolate the effect of the timing of auditor changes on audit quality.
We reperform our tests after limiting the sample to 1,048 auditor changes occurring in
2002 and 2003 where Arthur Andersen was the predecessor audit firm. Table 8 presents the
results from estimating Equations 1 and 2. In column 1, we do not find a higher likelihood
of misstatement for companies engaging a new auditor late in the year (p = .248).
However, in column 2, we find that the effect of late auditor changes is more pronounced
for complex clients, consistent with our primary results.
Audit Production Costs
As discussed in the introduction, absent data on auditor costs (e.g., hours worked,
personnel mix, etc.), we cannot directly measure the length or steepness of the audit
learning curve. Using audit fees (or some measure of unexpected audit fees) to capture
audit effort or costs is challenging in our setting because we examine first-year audit
engagements and, although some companies disclose annual audit fees paid to their predecessor and successor auditors, some do not. In addition, prior research finds evidence
of low-balling in the initial years of an audit engagement to secure new clients (e.g., Deis
& Giroux, 1996; Desir, Casterella, & Kokina, 2014; Simon & Francis, 1988), suggesting
that audit fees are not necessarily representative of auditor effort in these years.
Despite these limitations, we perform tests to provide additional support for our inference that the negative association between audit quality and late auditor changes that we
document in our main tests is related to audit production constraints. Specifically, we perform tests to investigate the effects of late versus early auditor changes on audit production
costs (using audit fees as a proxy). To mitigate the influence of the issues described above on
our results, we limit this analysis to our matched sample of early and late auditor changes.
This matched sample is also best suited for this test because the matching procedure includes
variables for the type of auditor switch (RESIGN, DOWNWARD, UPWARD), which likely
influences audit fees.
Of the 1,807 matched pairs in our first matched sample, 1,577 pairs have available
audit fee data. Thus, our sample for this test consists of 3,154 company-year
observations. We use the following ordinary least squares regression model to investigate fee
differences for late versus early auditor changes, controlling for a number of client
characteristics:
282
Journal of Accounting, Auditing & Finance
Table 8. Andersen’s Collapse.
The Effect of Auditor Engagement Timing on Audit Quality.
Variable
Intercept
LATE
LATE 3 COMPLEXITY
GCO
LOSS
LEVERAGE
ROA
LATEFILERt – 1
ICMW
M&A
NEWFINANCE
SIZE
MTB
BIGN
SPECIALIST
COMPLEXITY
DOWNWARD
Industry FE
Year FE
N
Area under the ROC curve
x 2 test:
LATE + LATE 3 COMPLEXITY = 0
Column 1
Column 2
DV = MISSTATE
DV = MISSTATE
Predictions
Coefficient
p value
Coefficient
p value
?
+
+
?
+
+
?
+
+
+
+
?
?
2
2
+
?
23.627***
0.196
\.001
.248
20.406
0.119
0.397*
0.492
0.687**
.312
.288
.099
.100
.030
23.765***
0.043
1.988***
20.354
0.111
0.392
0.483
0.723**
\.001
.443
.004
.379
.302
.105
.107
.024
0.026
1.092***
0.089*
0.002
0.222
20.147
0.185
20.038
Included
Included
1,048
0.700
.477
.002
.097
.938
.623
.238
.295
.953
0.046
1.108***
0.100*
0.001
0.292
20.129
20.299
0.061
Included
Included
1,048
0.700
.461
.002
.068
.966
.654
.267
.754
.929
7.757***
.005
Note. This table presents results from estimating Equations 1 and 3. All variables are defined in the appendix.
Standard errors are clustered by company. p values are one-tailed if a sign is predicted, and two-tailed otherwise.
ROC = receiver operating characteristic curve; ?: two tailed p-values.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
LAFEEit = g0 + g1 LATEit + g2 GCOit + g3 LOSSit + g4 LEVERAGEit + g5 ROAit
+ g6 LATEFILERit1 + g7 ICMWit + g8 M&Ait + g9 NEWFINANCEit + g10 SIZEit
+ g11 MTBit + g12 BIGNit + g13 SPECIALISTit + g14 COMPLEXITYit
+ g15 SECONDTIERit + g16 BUSYit + g17 LNONAFEEit + g18 RESIGNit
ð5Þ
+ g19 DOWNWARDit + g20 UPWARDit + gm Industry FE + gn Year FE,
where all variables are as previously defined or as defined in the appendix. The results are
presented in Table 9. We find that audit fees are higher for clients that engage auditors late
in the year relative to clients that engage auditors early in the year (p = .084). We speculate
that this result is attributable to heightened time constraints (which are priced) that arise
due to production inefficiencies associated with late auditor changes. Although this result is
consistent with audit production inefficiencies, we urge caution in the interpretation of
these tests due to the limitations described above. Furthermore, although our model
Cassell et al.
283
Table 9. The Effect of Auditor Engagement Timing on Audit Fees.
LATE and matched EARLY
DV = LAFEE
Variable
Intercept
LATE
GCO
LOSS
LEVERAGE
ROA
LATEFILERt – 1
ICMW
M&A
NEWFINANCE
SIZE
MTB
BIGN
SPECIALIST
COMPLEXITY
SECONDTIER
BUSY
LNONAFEE
RESIGN
DOWNWARD
UPWARD
Industry FE
Year FE
N
Adjusted R2
Predictions
Coefficient
p value
?
+
+
+
+
2
+
+
+
+
+
+
+
+
+
+
+
+
?
?
?
8.104***
0.051*
0.041
0.289***
0.053***
20.047***
0.183***
0.190***
0.126**
0.059*
0.387***
0.002*
0.886***
20.067
0.471***
0.351***
0.169***
0.043***
0.134***
0.323***
20.057
Included
Included
3,064
.674
\.001
.063
.213
\.001
.003
\.001
\.001
\.001
.038
.081
\.001
.073
\.001
.842
\.001
\.001
\.001
\.001
.001
\.001
.388
Note. This table presents results from estimating Equation 5. All variables are defined in the appendix. Standard
errors are clustered by company. p values are one-tailed if a sign is predicted, and two-tailed otherwise; ?: two
tailed p-values.
*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
includes a number of client risk characteristics, and the matched sample of late and early
auditor changes are well-balanced across these risk characteristics, we cannot rule out the
possibility that the result is attributable to a risk premium.
Summary and Conclusion
In this article, we investigate whether the timing of a successor auditor’s engagement influences audit quality in the year of an auditor change. We use misstatements (identified
through financial statement restatements) to proxy for audit quality. We find that companies
engaging an auditor late in the year (i.e., during or after the fourth fiscal quarter) are more
likely to misstate their financial statements than are companies that change auditors early in
the year and companies that do not change auditors. We also find that companies engaging
an auditor early in the year are not more likely to misstate than are companies that do not
change auditors. In additional analyses, we find that the effect of late auditor changes on the
284
Journal of Accounting, Auditing & Finance
likelihood of misstatement is more pronounced when clients have multiple operating segments. This suggests that audit quality suffers in the year of an auditor change when the successor auditor is engaged late in the year, especially when client operations are complex.
The results of this study should be informative to various stakeholders (e.g., investors,
audit committees, audit firms, etc.) interested in the effects of auditor changes on audit
quality. In addition, the results should be of interest to regulators as they continue to deliberate proposed regulatory actions (many of which would lead to more frequent auditor
changes) intended to improve auditor independence, objectivity, and audit quality.
Appendix Variable Definitions.
Variable
Definition
BIGN
An indicator variable set equal to one if the auditor is from the Big 4 (or Arthur
Andersen LLP), and zero otherwise.
An indicator variable set equal to one if the client’s fiscal year ends in December or
January, and zero otherwise.
An indicator variable set equal to one if more than one operating segment is
reported in Compustat, and zero otherwise.
An indicator variable set equal to one if the auditor change was a downward switch
(e.g., from a Big N auditor to a non-Big N auditor, or from a 2nd tier auditor to a
non-Big N, non-2nd tier auditor), and zero otherwise.
An indicator variable set equal to one if a new auditor was engaged prior to the
start of the fourth fiscal quarter, and zero otherwise.
An indicator variable set equal to one if the company received a going-concern
opinion, and zero otherwise.
An indicator variable set equal to one if the company reported a material weakness
in internal control over financial reporting, and zero otherwise.
Industry indicator variables (where, following Ashbaugh, LaFond, & Mayhew, 2003;
Cao, Myers, & Omer, 2012, we use Standard Industrial Classification (SIC) codes
to define industries as follows: agriculture (0100-0999), mining and construction
(1000-1999, excluding 1300-1399), food (2000-2111), textiles and printing/
publishing (2200-2799), chemicals (2800-2824; 2840-2899), pharmaceuticals (28302836), extractive (1300-1399; 2900-2999), durable manufacturers (3000-3999,
excluding 3570-3579 and 3670-3679), transportation (4000-4899), retail (50005999), services (7000-8999, excluding 7370-7379), computers (3570-3579; 36703679; 7370-7379), and utilities (4900-4999).
The natural log of audit fees.
An indicator variable set equal to one if a new auditor was engaged during or after
the fourth fiscal quarter, and zero otherwise.
An indicator variable set equal to one if the company filed its annual report (10-K)
after the required filing deadline, and zero otherwise.
Long-term debt plus the current portion of long-term debt, scaled by total assets.
The natural log of non-audit fees.
An indicator variable set equal to one if net income is negative, and zero otherwise.
An indicator variable set equal to one if the year t annual financial statements were
misstated (as revealed by a subsequent restatement), and zero otherwise.
The market-to-book ratio, calculated as the market value of equity divided by the
book value of equity.
An indicator variable set equal to one if there was a merger or acquisition in the
year, and zero otherwise. Consistent with Cassell, Dreher, and Myers (2013), we
identify mergers and acquisitions using the Compustat variable AQP.
BUSY
COMPLEXITY
DOWNWARD
EARLY
GCO
ICMW
Industry FE
LAFEE
LATE
LATEFILER
LEVERAGE
LNONAFEE
LOSS
MISSTATE
MTB
M&A
(continued)
Cassell et al.
285
Appendix (continued)
Variable
NEWFINANCE
RESIGN
ROA
SECONDTIER
SIZE
SPECIALIST
TENURE
UPWARD
Year FE
YR2EARLY
YR2LATE
Definition
An indicator variable set equal to one if the company issued equity or debt in the
current year, and zero otherwise. Following Dechow, Ge, Larson, and Sloan
(2011), we create this variable by identifying companies with long-term debt
issuances (Compustat variable DLTIS) or sales of common or preferred stock
(Compustat variable SSTK).
An indicator variable set equal to one if the previous auditor resigned from the
client, and zero otherwise.
Return on assets, measured as net income divided by total assets.
An indicator variable set equal to one if the new auditor is from the second tier of
audit firms (defined as BDO Seidman, Crowe, Chizek and Company, Grant
Thornton, and McGladrey & Pullen), and zero otherwise.
The natural log of total assets.
An indicator variable set equal to one if the auditor is an industry specialist, and
zero otherwise. Following Reichelt and Wang (2010), we define an auditor as an
industry specialist if the auditor’s audit fee market share in the two-digit SIC code
is at least 50% at the Metropolitan Statistical Area (MSA) level and exceeds 30% at
the national level.
The length (consecutive years to date) of the auditor–client relationship.
An indicator variable set equal to one if the auditor change was an upward switch
(e.g., from a non-Big N, non-2nd tier auditor to a 2nd tier auditor, or from a nonBig N auditor to a Big N auditor), and zero otherwise.
Indicator variables for each year in the sample period.
An indicator variable set equal to one if a new auditor was engaged in the prior year
before the end of the third fiscal quarter, and zero otherwise.
An indicator variable set equal to one if a new auditor was engaged in the prior year
during or after the fourth fiscal quarter, and zero otherwise.
Authors’ Note
The data used are publicly available from the sources cited in the text.
Acknowledgments
We thank Agnes Cheng (2016 Journal of Accounting, Auditing and Finance Conference Editor), Bharat
Sarath (Editor), Dan Simunic (discussant), Ken Bills, James Myers, Karen Pincus, Nate Stephens,
Devin Williams, Aaron Zimbelman, workshop participants at the University of Arkansas and the
University of North Carolina at Charlotte, and conference participants at the 2013 Brigham Young
University Accounting Research Symposium, the 2014 American Accounting Association Annual
Meeting, the 2014 American Accounting Association Auditing Midyear Meeting, and the 2016 Journal
of Accounting, Auditing and Finance Conference for helpful comments and suggestions. Linda Myers
gratefully acknowledges financial support from the Haslam Chair of Business at the University of
Tennessee, Knoxville, and from the Garrison/Wilson Chair while at the University of Arkansas.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/
or publication of this article.
286
Journal of Accounting, Auditing & Finance
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this
article.
Notes
1. Examples include the Public Company Accounting Oversight Board (PCAOB) Proposed Rule
Improving Transparency through Disclosure of Engagement Partner and Certain Other
Participants in Audits (Release No. 2011-007), the PCAOB Final Rule Communications with
Audit Committees (AS 16), and the PCAOB concept release Auditor Independence and Audit
Firm Rotation (Release No. 2011-006).
2. Consistent with this, we do not observe a large number of clients that do not secure an independent auditor.
3. Although this new regulation does not directly apply to U.S. companies, the rules apply to any
entity, including subsidiaries of the United States or other multinationals that meet the definition
of a European Union (EU) public interest entity.
4. For example, consider the difference between an audit engagement where five people work 40 hr
a week for 1 month and an audit engagement where 100 people work 10 hr for 1 day. Although
the number of audit hours are the same, coordination and communication within the audit team
and with client personnel will likely be much more inefficient in the latter scenario. We thank
Dan Simunic (discussant) for sharing this example.
5. We exclude misstatements made in 2004 and 2005 that were solely related to lease accounting
because these were based on interpretive guidance newly issued by the Securities and Exchange
Commission (SEC).
6. Greene (2004) suggests that problems can arise when estimating a nonlinear model with fixed
effects. To ensure that our results are robust, we reestimate each model as a linear probability
model rather than as a logistic regression model. All results and inferences are consistent with
those tabulated.
7. In addition, Kravet, Myers, Sanchez, and Scholz (2016) suggest that the association between misstatements and acquisitions could exist because managers are more likely to make acquisitions
when equity is overvalued because of misstatements.
8. See, for example, Johnson, Khurana, and Reynolds (2002); Myers, Myers, and Omer (2003);
Carcello and Nagy (2004), Stanley and DeZoort (2007); Jenkins and Velury (2008); and Reichelt
and Wang (2010). When using the sample of matched late and early auditor changes, we exclude
auditor tenure (TENURE) because this variable equals one for both groups.
9. When using the sample of companies with late auditor changes matched with companies that do
not change auditors, we exclude controls for the type of auditor switch (RESIGN,
DOWNWARD, UPWARD) because the control group consists of clients that do not change
auditors.
10. Based on a review of prior research, the most common variable used to proxy for company complexity is the number (or natural log or square root) of operating or business segments. See, for
example, Whisenant, Sankaraguruswamy, and Raghunandan (2003); Ge and McVay (2005);
Ashbaugh-Skaife, Collins, and Kinney (2007); Zhang, Zhou, and Zhou (2007); Hogan and
Wilkins (2008); Francis and Yu (2009); Hoitash, Hoitash, and Bedard (2009); Li, Sun, and
Ettredge (2010); Doogar, Sivadasan, and Solomon (2010); Seetharaman, Sun, and Wang (2011);
Bens, Heltzer, and Segal (2011); Chan, Chen, Janakiraman, and Radhakrishnan (2012); Hennes,
Leone, and Miller (2014); and Kanagaretnam, Lobo, Ma, and Zhou (2016). Other, less common,
measures of company complexity include the number (or natural log or square root) of geographic segments and the presence of foreign operations, foreign sales, or foreign currency
exchange gains/losses. In untabulated analysis, we reperform our analyses using indicator variables for multiple geographic segments and an indicator for the presence of foreign operations.
Cassell et al.
287
Although results using the presence of foreign operations were close to significance at conventional levels, results using multiple geographic segments do not support our main results.
11. This maximum difference follows Lawrence, Minutti-Meza, and Zhang (2011) and Bills,
Cunningham, and Myers (2016).
12. We perform an additional (untabulated) test to assess the robustness of the results presented in
Tables 6 and 7. Given the large number of auditor changes in 2002, we exclude observations
from 2002 and reestimate our models. Second, rather than identifying as late auditor changes
those changes occurring after the third fiscal quarter, we identify late auditor changes as those
occurring after the filing of the third quarter 10-Q. This alternative definition does not allow the
new auditor to benefit from performing any of the quarterly reviews. For both sets of analyses,
our inferences are consistent with those tabulated. As discussed in Ai and Norton (2003), interaction effects in nonlinear models cannot be evaluated based on the magnitude, sign, and significance of the associated marginal effects. Norton, Wang, and Ai (2004) provide a detailed
discussion of the issues associated with the use of the marginal effect that is reported in most statistical packages and discusses a STATA command (INTEFF) that computes the ‘‘correct marginal effect.’’ In untabulated analyses, we use the INTEFF procedure to assess the robustness of
our inferences. We find that, with respect to the sign and significance of the interaction terms,
inferences drawn from these tests are consistent with those derived from the tabulated results.
13. Note that this inference is not consistent with evidence from prior work or with opinions
expressed by regulators. For example, in its concept release on mandatory audit firm rotation,
the PCAOB reiterated concerns expressed by the Cohen Commission in 1978—that ‘‘[r]otation
would considerably increase the costs of audits because of the frequent duplication of the startup and learning time necessary to gain familiarity with a company and its operations that is necessary for an effective audit’’ (PCAOB, 2011, p. 11). In addition, most instances of substandard
auditor performance identified by the Cohen Commission were in the first or second year of an
audit engagement (American Institute of Certified Public Accountants [AICPA], 1978).
Moreover, after reviewing 406 alleged audit failures of companies that file with the SEC, a committee of the AICPA (1992) concluded that audit failures are 3 times more likely in the first 2
years of an audit engagement than in subsequent years. As discussed previously, prior research
finds that earnings quality is lower when auditor tenure is shorter (Johnson et al., 2002; Myers
et al., 2003).
References
Ai, C., & Norton, E. C. (2003). Interaction terms in logit and probit models. Economics Letters, 80,
123-129.
American Institute of Certified Public Accountants. (1978). The commission on auditors’ responsibilities: Report, conclusions, and recommendations. New York, NY: Author.
American Institute of Certified Public Accountants SEC Practice Section. (1992). Statement of position regarding mandatory rotation of audit firms of publicly held companies. New York, NY:
American Institute of Certified Public Accountants.
Ashbaugh, H., LaFond, R., & Mayhew, B. W. (2003). Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review, 78, 611-639.
Ashbaugh-Skaife, H., Collins, D. W., & Kinney, W. R. (2007). The discovery and reporting of internal control deficiencies prior to SOX-mandated audits. Journal of Accounting and Economics, 44,
166-192.
Ashton, R. H., Graul, P. R., & Newton, J. D. (1989). Audit delay and the timeliness of corporate
reporting. Contemporary Accounting Research, 5, 657-673.
Bens, D. A., Heltzer, W., & Segal, B. (2011). The information content of goodwill impairments and
SFAS 142. Journal of Accounting, Auditing & Finance, 26, 527-555.
288
Journal of Accounting, Auditing & Finance
Bills, K., Cunningham, L., & Myers, L. A. (2016). Small audit firm membership in associations, networks, and alliances: Implications for audit quality and audit fees. The Accounting Review, 91,
767-792.
Blankley, A. I., Hurtt, D. N., & MacGregor, J. E. (2012). Abnormal audit fees and restatements.
Auditing: A Journal of Theory & Practice, 31, 79-96.
Cao, Y., Myers, L. A., & Omer, T. C. (2012). Does company reputation matter for financial reporting
quality? Evidence from restatements. Contemporary Accounting Research, 29, 956-990.
Carcello, J., & Nagy, A. (2004). Audit firm tenure and fraudulent financial reporting. Auditing: A
Journal of Theory & Practice, 23(2), 55-69.
Carver, B. T., Hollingsworth, C. W., & Stanley, J. D. (2011). Recent auditor downgrade activity and
changes in client’s discretionary accruals. Auditing: A Journal of Practice & Theory, 30(3), 33-58.
Cassell, C. A., Dreher, L. M., & Myers, L. A. (2013). Reviewing the SEC’s review process: 10-K
comment letters and the cost of remediation. The Accounting Review, 88, 1875-1908.
Chan, L., Chen, T.-Y., Janakiraman, S., & Radhakrishnan, S. (2012). Reexamining the relationship
between audit and nonaudit fees: Dealing with weak instruments in two-stage least squares estimation. Journal of Accounting, Auditing & Finance, 27, 299-324.
Chi, W., Myers, L. A., Omer, T. C., & Xie, H. (2017). The effects of audit partner pre-client and
client-specific experience on audit quality and on perceptions of audit quality. Review of
Accounting Studies, 22, 361-391.
Christensen, B. E., Glover, S. M., Omer, T. C., & Shelley, M. K. (2016). Understanding audit quality:
Insights from audit professionals and investors. Contemporary Accounting Research, 33, 16481684.
Dechow, P. W., Ge, W., Larson, C., & Sloan, R. (2011). Predicting material accounting misstatements. Contemporary Accounting Research, 28, 17-82.
DeFond, M. L., & Subramanyam, K. R. (1998). Auditor changes and discretionary accruals. Journal
of Accounting and Economics, 25, 35-67.
Deis, D. R., & Giroux, G. (1996). The effect of auditor changes on audit fees, audit hours, and audit
quality. Journal of Accounting and Public Policy, 15, 55-76.
Desir, R., Casterella, J. R., & Kokina, J. (2014). A re-examination of audit fees for initial audit
engagements in the post-SOX period. Auditing: A Journal of Practice & Theory, 33(2), 59-78.
Doogar, R., Sivadasan, P., & Solomon, I. (2010). The regulation of public company auditing:
Evidence from the transition to AS5. Journal of Accounting Research, 48, 795-814.
Ernst & Young LLP. (2012). Technical line: Respondents to PCAOB overwhelmingly oppose mandatory audit firm rotation. New York, NY: Author.
Francis, J. R., & Michas, P. N. (2013). The contagion effect of low-quality audits. The Accounting
Review, 88, 521-552.
Francis, J. R., & Yu, M. D. (2009). Big 4 office size and audit quality. The Accounting Review, 84,
1521-1552.
Franzel, J. M. (2012, March 28). Opening remarks. PCAOB Investor Advisory Group Meeting,
Washington, DC. Retrieved from http://pcaobus.org/News/Speech/Pages/03282012_Franzel
Remarks.aspx
Ge, W., & McVay, S. (2005). The disclosure of material weaknesses in internal control after the
Sarbanes-Oxley Act of 2002. Accounting Horizons, 19, 137-158.
Greene, W. (2004). The behavior of the maximum likelihood estimator of limited dependent variable
models in the presence of fixed effects. The Econometrics Journal, 7, 98-119.
Hennes, K. M., Leone, A. J., & Miller, B. P. (2014). Determinants and market consequences of auditor dismissals after accounting restatements. The Accounting Review, 89, 1051-1082.
Hogan, C. E., & Wilkins, M. S. (2008). Evidence on the audit risk model: Do auditors increase audit
fees in the presence of internal control deficiencies? Contemporary Accounting Research, 25, 219242.
Hoitash, U., Hoitash, R., & Bedard, J. C. (2009). Corporate governance and internal control over
financial reporting: A comparison of regulatory regimes. The Accounting Review, 84, 839-867.
Cassell et al.
289
Jenkins, D. S., & Velury, U. (2008). Does auditor tenure influence the reporting of conservative earnings? Journal of Accounting and Public Policy, 27, 115-132.
Johnson, E., Khurana, I., & Reynolds, K. J. (2002). Audit-firm tenure and the quality of financial
reports. Contemporary Accounting Research, 19, 637-660.
Kanagaretnam, K., Lobo, G. J., Ma, C., & Zhou, J. (2016). National culture and internal control material weaknesses around the world. Journal of Accounting, Auditing & Finance, 31, 28-50.
Kravet, T. D., Myers, L. A., Sanchez, J. M., & Scholz, S. (2016). Do financial statement misstatements facilitate corporate acquisitions? (Working paper).
Lawrence, A., Minutti-Meza, M., & Zhang, P. (2011). Can Big 4 versus non-Big 4 differences in
audit-quality proxies be attributed to client characteristics? The Accounting Review, 86, 259-286.
Li, C., Sun, L., & Ettredge, M. (2010). Financial executive qualifications, financial executive turnover, and adverse SOX 404 opinions. Journal of Accounting and Economics, 50, 93-110.
McGuire, S. T., Omer, T. C., & Sharp, N. Y. (2012). The impact of religion on financial reporting
irregularities. The Accounting Review, 87, 645-673.
Myers, J. N., Myers, L. A., & Omer, T. C. (2003). Exploring the term of the auditor-client relationship and the quality of earnings: A case for mandatory auditor rotation? The Accounting Review,
78, 779-799.
Nagy, A. L. (2005). Mandatory audit firm turnover, financial reporting quality, and client bargaining
power: The case of Arthur Andersen. Accounting Horizons, 19(2), 51-68.
Norton, E. C., Wang, H., & Ai, C. (2004). Computing interaction effects and standard errors in logit
and probit models. The Stata Journal, 4, 154-167.
Petersen, M. A. (2009). Estimating standard errors in financial panel data sets: Comparing
approaches. Review of Financial Studies, 22, 435-480.
Public Company Accounting Oversight Board. (2011). Concept release on auditor independence and
audit firm rotation (PCAOB release No. 2011-006). Washington, DC: Author.
Reichelt, K. J., & Wang, D. (2010). National and office-specific measures of auditor industry expertise and effects on audit quality. Journal of Accounting Research, 48, 647-686.
Schmidt, J. J. (2012). Perceived auditor independence and audit litigation: The role of nonaudit services fees. The Accounting Review, 87, 1033-1065.
Seetharaman, A., Sun, Y., & Wang, W. (2011). Tax-related financial statement restatements and auditor-provided tax services. Journal of Accounting, Auditing & Finance, 26, 677-698.
Simon, D. T., & Francis, J. R. (1988). The effects of auditor change on audit fees: Tests of price cutting and price recovery. The Accounting Review, 63, 255-269.
Stanley, J., & DeZoort, F. (2007). Audit firm tenure and financial restatements: An analysis of industry specialization and fee effects. Journal of Accounting and Public Policy, 26, 131-159.
Whisenant, S., Sankaraguruswamy, S., & Raghunandan, K. (2003). Evidence on the joint determination of audit and non-audit fees. Journal of Accounting Research, 41, 721-744.
Zhang, Y., Zhou, J., & Zhou, N. (2007). Audit committee quality, auditor independence, and internal
control weaknesses. Journal of Accounting and Public Policy, 26, 300-327.
Copyright of Journal of Accounting, Auditing & Finance is the property of Sage Publications
Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder’s express written permission. However, users may print,
download, or email articles for individual use.
AUDITING: A JOURNAL OF PRACTICE & THEORY
Vol. 39, No. 3
August 2020
pp. 161–184
American Accounting Association
DOI: 10.2308/ajpt-18-152
Mandatory Audit Partner Rotations and Audit Quality in
the United States
Huan Kuang
University of Massachusetts Amherst
Huimin Li
University of New Hampshire
Matthew G. Sherwood
University of Massachusetts Amherst
Robert L. Whited
North Carolina State University
SUMMARY: This study uses a sample of mandatory partner rotation events hand collected from SEC filings to
investigate the relation between mandatory audit partner rotation and audit quality in the United States. Across a
variety of control groups and audit quality proxies, we do not find evidence consistent with rotation materially
improving audit quality (i.e., ‘‘fresh look’’). Although somewhat limited, the only statistically significant evidence we
document suggests that audited financial statements may be more likely to contain a material misstatement (i.e.,
subsequently be restated) following a mandatory audit partner rotation, particularly when the audit firm tenure is
short. We also provide evidence from client disclosures that mandatory rotation rules trigger auditor-client
realignment. Together, our results provide important evidence on the merits of mandatory partner rotation rules in the
United States.
JEL Classifications: M40; M41; M42.
Data Availability: Data are publicly available from sources identified in the article.
Keywords: U.S. mandatory audit partner rotations; material misstatements; audit quality.
I. INTRODUCTION
S
ection 203 of the Sarbanes-Oxley Act (SOX) requires audit engagements of publicly traded companies to change the
lead (and concurring) engagement partner(s) at least every five years. Thus, if a lead engagement partner has served for
five consecutive years in that capacity, the audit firm must assign a new lead engagement partner to that client.
Proponents of mandatory partner rotation contend that the new lead audit partner provides a fresh perspective to the audit
engagement while retaining client-specific knowledge from previous audits that would be lost under a mandatory audit firm
rotation regime. For this reason, many consider mandatory audit partner rotation a key element for maintaining the
independence that is necessary for high-quality audits. However, with the ‘‘fresh look’’ also comes a ‘‘learning curve’’ for the
new engagement partner, which may cause a temporary decline in audit quality or a deterioration in the auditor-client
relationship as partner-specific knowledge about the client is lost on the transition.
We thank Chan Li (editor) and two anonymous reviewers for their helpful comments, as well as participants at the 2019 AAA Annual Meeting.
Editor’s note: Accepted by Chan Li, under the Senior Editorship of Christopher P. Agoglia.
Submitted: November 2018
Accepted: February 2020
Published Online: February 2020
161
Kuang, Li, Sherwood, and Whited
162
Because partner identities for U.S. audit engagements have historically not been disclosed publicly, research on the effects
of mandatory partner rotation and partner tenure on audit quality in the United States is scant.1 PCAOB Auditing Standard AS
3101.10 requires that public accounting firms, not the individual audit partners, sign audit reports in the United States; thus,
without access to proprietary data, it has been extremely difficult to identify when audit partner rotation occurs. Extant research
on the effects of partner rotation either uses proprietary data (Gipper, Hail, and Leuz 2019; hereafter, GHL), uses publicly
available data from comment letters to infer rotation (Laurion, Lawrence, and Ryans 2017; hereafter, LLR), which cannot
distinguish between mandatory and voluntary rotation, or examines audits of non-U.S. firms (e.g., Stewart, Kent, and
Routledge 2016; Lennox, Wu, and Zhang 2014; Chi, Huang, Liao, and Xie 2009). In this study, we implement a novel
approach that uses publicly available data to identify audit partner rotations that occur due to mandated audit partner term
limits. To do so, we analyze over 1.3 million SEC filings and identify instances when companies disclose in their proxy
statement or 8-K filings that the lead external audit partner will rotate to comply with the five-year audit partner term limit as
specified by SOX Section 203. We also use an expanded mandatory rotation sample identified using a modification of the
comment letter approach from LLR to identify partner changes that occur after five years of partner tenure. Using these data, we
investigate the relation between mandatory partner rotations and audit quality using a variety of audit quality proxies. We find
no evidence suggesting that the new audit partner provides a ‘‘fresh look’’ that improves audit quality. In fact, we find limited
evidence that audited financial statements are more likely to be materially misstated (i.e., subsequently restated) in the initial
year(s) following mandatory audit partner rotation than in the terminal year(s) of partner tenure, particularly when audit firm
tenure is short. However, this evidence should be interpreted in light of the limited sample size and infrequency of material
restatements. Nonetheless, the weight of our evidence is not consistent with partner rotation yielding material improvements to
the audit process due to the ‘‘fresh look’’ of the new audit partner.
Our study makes several important contributions to the literature. We are the first study to our knowledge to use publicly
available data to directly explore the effects of mandatory lead audit partner rotation in the United States as required by SOX
Section 203. Second, we present a new way to identify mandatory partner rotations for future research interested in the effects
of mandatory audit partner rotation. The audit firms and PCAOB do not provide engagement-level data to the broad research
community regarding mandatory audit partner rotations, so our study identifies an innovative way in which researchers may
identify mandatory partner rotations even without access to proprietary data. We find that partner rotation disclosures are
increasing in recent years, providing a resource for researchers interested in the effects of mandatory audit partner rotation
before the Form AP filings have been disclosed for five years when researchers can infer mandatory rotations. Finally, we find
that some clients change auditors at the end of a partner’s tenure due to the audit firm’s inability to comply with partner rotation
rules, suggesting an additional cost of mandatory rotation rules.
While our results provide important insights, they are subject to several caveats. First, the companies in our sample are
larger and have longer audit firm tenure than the average U.S. public company. As such, the conclusions from this study may
not generalize to all companies. Second, while our sample of rotation events is similar to prior research (LLR), statistical
insignificance in some settings may be the result of a lack of statistical power. As such, we follow the advice in Cready et al.
(2019) and evaluate the evidence in light of the 95 percent confidence intervals.
II. BACKGROUND AND RESEARCH QUESTIONS
Background
Regulators and standard setters have long expressed concerns that economic and social bonds between the audit partner
and the client may increase over the length of a relationship, possibly diminishing the audit partner’s professional skepticism
and independence. For example, an audit partner might fear that disagreeing with an audit client’s financial reporting decisions
will result in losing the audit engagement and its fees, adversely affecting the audit partner’s portfolio of clients and possibly
the partner’s compensation and standing in the firm (Nelson 2009). As the partner’s tenure with a client increases, their
economic dependence may likewise increase, particularly if the client represents a large portion of the partner’s ‘‘book of
business’’ or a significant contribution to the audit firm or audit office’s bottom line.
In addition to career or fiscal motives, an audit partner might feel pressure not to disappoint a client when strong social ties
have developed between the partner and the client. Strong social ties could make the audit partner hesitant to push back on the
client’s accounting decisions (Bazerman, Moore, Tetlock, and Tanlu 2006), particularly when the treatment involves a
significant amount of judgment. Regulators and standard setters have implemented lead partner rotation rules to mitigate such
1
We note that the recent requirement to file a Form AP with the PCAOB results in the disclosure of lead engagement partner identities. However, Form
AP data do not indicate when audit partner rotations are mandatory. Thus, until Form AP filings have been around for at least five years, it will not be
clear when rotations are mandatory.
Auditing: A Journal of Practice & Theory
Volume 39, Number 3, 2020
Mandatory Audit Partner Rotations and Audit Quality in the United States
163
concerns about independence impairments. The first chief accountant of the SEC, Carman Blough, suggested that requiring
audit supervisors and lead partners to change on a regular basis might limit bonding between auditors and clients (Blough
1951). In response to the Metcalf Report’s recommendation of mandatory audit firm rotation, the American Institute of
Certified Public Accountants (AICPA 1978) introduced a requirement limiting the tenure of lead audit partners on SEC client
engagements to seven consecutive years. SOX Section 203 reduced the number of consecutive years an audit partner may serve
as the lead or concurring audit partner to five years for public companies and extended the cooling off period to five years.
While the term limit has not changed since the inception of SOX, the PCAOB has suggested that partner rotation may not
sufficiently address independence problems and has considered instituting mandatory audit firm rotation (Doty 2011).
Audit Quality
Proponents of mandatory rotation rules often contend that the new audit partner provides a ‘‘fresh look’’ on the engagement
that can identify overlooked aspects of the audit, thereby improving audit quality. In theory, the new lead audit partner provides
a new perspective on the existing audit program and might be able to identify insufficient or nonexistent procedures necessary
for a high-quality audit (Lennox and Wu 2018). Partner rotation (as opposed to audit firm rotation) does not sacrifice all of the
cumulative audit knowledge and experience gained by the external audit firm when the new partner rotates on the engagement.
For the most part, empirical studies on the effects of audit partner rotation examine data from countries such as Australia
(Stewart et al. 2016), China (Lennox et al. 2014), and Taiwan (Chi et al. 2009), as these countries require the disclosure of
audit partner identities (see Lennox and Wu [2018] for a more comprehensive review of this literature). However, the degree to
which conclusions drawn from studies of foreign audit firms generalize to U.S. audit firms is somewhat limited. First, the
litigation environment, coupled with the strong regulatory oversight in the United States, incentivize audit partners to maintain
independence, even absent impending rotation. Thus, the benefits of a mandatory rotation regime are likely lower in the United
States. Second, the large U.S. public accounting firms (with the majority of public clients) are substantially larger than the
largest public accounting firms in other countries; thus, the firms have significantly more potential partners with public
company auditing experience in the relevant industry that can rotate onto a vacant engagement. Therefore, the implications of
partner rotation likely differ between U.S.-based audit engagements and those conducted internationally.
Further highlighting the need for additional study into the ramifications of audit partner rotation is the fact that the findings
of prior research on the subject are inconsistent. For example, Lennox et al. (2014) examine the ‘‘fresh look’’ argument using
proprietary data on audit adjustments in the setting of mandatory rotation of Chinese audit partners. Their findings suggest
improvements in the quality of the audits performed in the year preceding and the initial year following a required rotation.
They conclude that mandatory rotation in China improves audit quality through both a fresh perspective of the new partner and
by disciplining the outgoing audit partner to provide high audit quality in the year preceding a rotation (i.e., review effect).
However, using data from Taiwan, Chen, C.-J. Lin, and Y.-C. Lin (2008) find that firm-level discretionary accrual levels
decrease as audit partner tenure increases, suggesting that longer-tenured audit partners provide higher-quality audits. Likewise,
Chi et al. (2009) find that Taiwanese audit quality is significantly lower in the year of mandatory partner rotation. The
combination of location-specific differences in findings (i.e., differences between studies using data from Taiwan and China) as
well as the noted differences between the U.S. regulatory and legal environment and those of foreign jurisdictions highlight the
need for research in the area of mandatory audit partner rotations in the United States.
Research on the relation between audit partner rotation (and by extension audit partner tenure) and audit quality within the
U.S. setting is rather limited. Using a hand-collected sample of public companies, Manry, Mock, and Turner (2008) find an
inverse relationship between discretionary accrual levels and audit partner tenure among small companies but find no
relationship among large companies. Under the assumption that audit firms only change audit partners when required by the
Section 203 mandate, Litt, Sharma, Simpson, and Tanyi (2014) identify the fifth consecutive fiscal year following an auditor
switch by a public company and assume that audit partner rotations occur following this year. Their findings suggest that a
reduction in financial reporting quality accompanies the assumed change in the audit partner. Likewise, Sharma, Tanyi, and Litt
(2017) adopt a similar approach and find that the initial year following assumed rotation events is associated with higher fees
and a longer audit report lag.
A recent study by LLR infers engagement partner identity by using the names of audit firm personnel copied on comment
letter correspondences between the SEC and U.S. public companies. The authors find mixed evidence on the benefits of
changing audit partners. They find no relation between partner rotation and the likelihood that the financial statements contain a
misstatement. However, they do find an increase in the frequency of restatement announcements and an increase in deferred tax
allowances following a rotation event. They conclude that their results provide ‘‘some evidence suggesting that U.S. partner
rotations support a fresh look at the audit engagement’’ (Laurion et al. 2017, 209).
We view our study as an extension of LLR due to our ability to address certain limitations the authors note in their study.
First, the authors must assume that the accounting firm personnel copied on the SEC comment letter responses reflect the
Auditing: A Journal of Practice & Theory
Volume 39, Number 3, 2020
Kuang, Li, Sherwood, and Whited
164
current lead audit partner on the engagement, as opposed to other audit firm employees (e.g., concurring partner, client
relationship partner, office managing partner, prior period audit partner). Our primary sample does not use comment letters and,
therefore, does not rely on such an assumption. Second, the authors note that ‘‘endogenous partner rotations are a concern in our
setting because we cannot observe whether a partner rotation is mandatory or voluntary’’ (Laurion et al. 2017, 211). If low
financial reporting/audit quality, which might be more likely for audit engagements receiving an SEC comment letter, triggers
voluntary audit partner rotation events (i.e., rotations before reaching the five-year engagement partner term limit), this would
bias evidence toward supporting the ‘‘fresh look’’ hypothesis. This issue is important in light of the findings of the current
working paper by GHL, which suggests that 38 percent of all audit partner rotations are nonmandatory and that poor audit
quality tends to precede voluntary partner rotations. Finally, because client firms often do not receive SEC comment letters in
consecutive years, there are instances where LLR cannot determine whether the rotation event occurred in year t or t1 and
must make assumptions about the partner transition year. Our primary approach for identifying mandatory rotation events
addresses some of these limitations by using audit committee disclosures to identify known mandatory audit partner rotations as
well as the exact year of the rotation.2 However, it is important to acknowledge that this approach is not without its limitations.
Unlike comment letters, the rotations identified in this study are a result of a firm disclosure choice. The endogenous nature of
the disclosure may affect the inferences of our findings, as well as the generalizability of our results. We discuss these issues in
depth in Section III. For this reason, we do not suggest that our approach is ‘‘better’’ than that of LLR. Rather, we adopt an
alternative approach that identifies a new set of more than 100 mandatory rotation events in the U.S., which has its own
strengths and weaknesses. We suggest that our paper represents a valuable complement to LLR, whereby our results can be
evaluated in conjunction with theirs.
A current working paper by GHL explores the effects of mandatory partner rotation in the United States using proprietary
engagement data provided by the PCAOB. While GHL primarily focus on the economic effects of audit partner tenure on firms
in terms of audit fees and audit hours, Gipper et al. (2019, Table 3) also perform limited tests on the effects of partner tenure on
audit quality. GHL do not directly compare year-before to year-after partner rotations when assessing rotation effects on audit
quality. Rather, they include a partner tenure count variable and do not find evidence of an average linear relation between
partner tenure and audit quality (they do find some evidence of a negative relation between restatement announcements and
partner tenure). GHL do document that audit hours decrease, and audit fees increase over the tenure cycle, suggesting economic
benefits to audit firms of prolonged partner tenure. The proprietary data afford GHL a large sample in which they can identify
mandatory rotation events with a high degree of confidence. Our study differs in two important respects. First, our paper
focuses primarily on audit quality effects and considers the effects in the initial year following rotation, while GHL investigate
the linear relation b…

Still stressed from student homework?
Get quality assistance from academic writers!

Order your essay today and save 25% with the discount code LAVENDER