cameran et al - 2013 - are there adverse consequences of mandatory auditor rotation - evidence from the italian experience

Online Early — Preprint of Accepted Manuscript
This is a PDF file of a manuscript that has been accepted for publication in an American
Accounting Association journal. It is the final version that was uploaded and approved by the
author(s). While the paper has been through the usual rigorous peer review process for AAA
journals, it has not been copyedited, nor have the graphics and tables been modified for final
publication. Also note that the paper may refer to online Appendices and/or Supplements that are
not yet available. The manuscript will undergo copyediting, typesetting and review of page proofs
before it is published in its final form, therefore the published version will look different from this
version and may also have some differences in content.
We have posted this preliminary version of the manuscript as a service to our members and
subscribers in the interest of making the information available for distribution and citation as
quickly as possible following acceptance.
The DOI for this manuscript and the correct format for citing the paper are given at the top
of the online (html) abstract.
Once the final published version of this paper is posted online, it will replace this
preliminary version at the specified DOI.
The Accounting Review • Issues in Accounting Education • Accounting Horizons
Accounting and the Public Interest • Auditing: A Journal of Practice & Theory
Behavioral Research in Accounting • Current Issues in Auditing
Journal of Emerging Technologies in Accounting • Journal of Information Systems
Journal of International Accounting Research
Journal of Management Accounting Research • The ATA Journal of Legal Tax Research
The Journal of the American Taxation Association
preprint
accepted
manuscript


Are There Adverse Consequences of Mandatory Auditor Rotation?
Evidence from the Italian Experience

by



Mara Cameran*, Jere R. Francis**, Antonio Marra*, and Angela Pettinicchio*

*Bocconi University, Italy
**University of Missouri, USA



Draft Date: October 10, 2013

Contact Author:
Jere Francis
Email:


SUMMARY: Mandatory auditor rotation was recently proposed for the European Union and is
also under consideration in the United States. There has been little research into either the
benefits or costs of rotation in a true mandatory setting that could inform intelligent policy
making. Our paper helps fill this gap by examining Italy where mandatory rotation of auditors
has been required since 1975. We find that outgoing auditors do not shirk on effort (or quality),
but final year fees are 7 percent higher than normal which may indicate opportunistic pricing.
The fees of incoming auditors are discounted by 16 percent even though they have abnormally
higher engagement hours in the first year (17 percent), which is suggestive of low balling.
However, subsequent fees are abnormally higher and exceed the initial fee discount. Thus the
costs of mandatory rotation are nontrivial. Higher costs could be acceptable if rotation improves

audit quality, but we find evidence of the opposite. Namely, the quality of audited earnings is
lower in the first three years following rotation, relative to later years of auditor tenure. Since
rotation is costly and earnings quality improves with longer auditor tenure, the evidence from
Italy does not support the case for mandatory rotation.


Keywords: auditor rotation, audit fees, earnings quality, audit market regulation.


preprint
accepted
manuscript


Are There Adverse Consequences of Mandatory Auditor Rotation?
Evidence from the Italian Experience



SUMMARY: Mandatory auditor rotation was recently proposed for the European Union and is
also under consideration in the United States. There has been little research into either the
benefits or costs of rotation in a true mandatory setting that could inform intelligent policy
making. Our paper helps fill this gap by examining Italy where mandatory rotation of auditors
has been required since 1975. We find that outgoing auditors do not shirk on effort (or quality),
but final year fees are 7 percent higher than normal which may indicate opportunistic pricing.
The fees of incoming auditors are discounted by 16 percent even though they have abnormally
higher engagement hours in the first year (17 percent), which is suggestive of low balling.
However, subsequent fees are abnormally higher and exceed the initial fee discount. Thus the
costs of mandatory rotation are nontrivial. Higher costs could be acceptable if rotation improves
audit quality, but we find evidence of the opposite. Namely, the quality of audited earnings is

lower in the first three years following rotation, relative to later years of auditor tenure. Since
rotation is costly and earnings quality improves with longer auditor tenure, the evidence from
Italy does not support the case for mandatory rotation.


Keywords: auditor rotation, audit fees, earnings quality, audit market regulation.
preprint
accepted
manuscript
1

Are There Adverse Consequences of Mandatory Auditor Rotation?
Evidence from the Italian Experience

INTRODUCTION
We investigate if there are potential negative consequences of mandatory audit firm
rotation in Italy where listed companies have been required by law to rotate their auditors since
1975. The evidence is timely as the European Commission (hereafter EC) has proposed
mandatory rotation for all European listed companies (European Commission 2011). The EC
proposes a 6-year term limit for the audit appointment, with a 4-year “cooling off” period before
an auditor could be re-appointed. The Commission has the legal authority to issue a directive
requiring auditor rotation, which would then be implemented by each member country.
However, due to the broad reach of the proposal, including European-wide licensing and
supervision of auditors, the Commission has chosen to implement these reforms through a
statutory regulation which must be approved by the European Parliament. This approach would
impose stronger obligations on member countries to comply with auditor rotation as regulations
are immediately effective in each member state without the need to be approved by the separate
national parliaments of each European Union country. On April 24, 2013, the Legal Affairs
Committee approved a proposal for a 14-year rotation rule. However, the full European
Parliament has not yet acted on the recommendation.

The United States is also considering mandatory audit firm rotation. In June 2011, James
Doty, Chairman of the PCAOB suggested that rotation might be a desirable way to strengthen
auditor independence and objectivity (Doty 2011). The PCAOB followed this up with a
background paper on August 16, 2011 (PCAOB 2011). This document invited public comment
through December 14, 2011, with public hearings on March 21-22, June 28, and October 18,
2012. No further action has been taken. However, on July 8, 2013, the U.S. House of
preprint
accepted
manuscript
2

Representatives approved a bill that would ban mandatory rotation. The bill must yet be passed
by the U.S. Senate, and signed by the President, to become law.
The case for mandatory rotation rests on two assertions (European Commission 2011, p.
3; PCAOB 2011, pp. 11-12):
1. The Problem: auditors with long tenure can develop close relationships with clients that
impair their ability to be independent and which lowers audit quality.
1


2. The Solution: a periodic change of auditors will bring a fresh perspective to the audit, leading
to more skepticism, greater auditor objectivity, and improved audit quality.

While the idea of rotation has intuitive appeal, the European Commission does not cite empirical
evidence or make rigorous arguments, nor did the Commission’s earlier “Green Paper” (2010).
Instead, the case for rotation is supported by anecdotal assertions like the following: “Situations
where a company has appointed the same audit firm for decades seems incompatible with
desirable standards of independence” (European Commission 2010, p. 11). The PCAOB (2011)
presents a more balanced discussion of benefits and potential negative consequences of rotation,
but it too relies entirely on anecdotes to make the case for mandatory rotation. For example, the

PCAOB report (p. 12) cites John Biggs, then CEO of the pension fund TIAA-CREF, who asserts
“… auditor rotation is a ‘powerful antidote’ to auditor conflict of interests, which reduces
dramatically the financial incentives for the audit firms to placate management.” On p. 15, the
report also cites concerns of the PCAOB’s Investor Advisory Group: “Key to concern of
independence was the level of ‘coziness’ the firm had with management of the company being
audited . . . . Many of the auditors of large companies . . . had long running audit relationships
with those companies.”

1
The alleged independence threat arising from close relations with clients is termed self-serving bias by Bazerman
et al. (1997). Self-serving bias means that auditors are unable to conduct fully impartial and objective audits because
their self-interest is closely tied to clients. The bias operates at both a conscious and unconscious level, which leads
Bazerman et al. (1997) to conclude that it is impossible for auditors to be independent. This is a strong claim and has
been refuted by Nelson (2006) and King (2002).
preprint
accepted
manuscript
3

As is clear from the above quotes, at the heart of the case for mandatory rotation is the
belief that “bad things” can happen when auditors have long tenure. While there is only limited
research in true mandatory settings, there is a large body of research on audit firm tenure in a
voluntary setting. What have we learned from this research? Lennox (2013) provides a
comprehensive review, and concludes that with the exception of one study (Davis et al. 2009),
the general finding is that longer audit firm tenure (in voluntary settings) does not appear to harm
audit quality, and there is some counter-evidence that long tenure may improve audit quality
(e.g., Meyers et al. 2003). Interestingly, the PCAOB report (p. 16) also notes that PCAOB
inspection data show no association between long-tenure audits and negative comments in
inspection reports.
On the other hand, it may be the case that “bad things” do happen in the short-tenure

setting due to a learning curve effect. Short tenure occurs when there is a change in auditor, and
there is evidence that earnings quality is lower during the first few engagement years, which is
consistent with a learning curve on new audits (Johnson et al. 2002). Auditors are also more
likely to encounter material irregularities in the initial engagement year, which puts these audits
at greater risk (Loebbecke et al. 1989; Carcello and Nagy 2004; PCAOB 2011, p. 16).
Given the existing body of evidence, it appears the European Commission is advocating a
major change to the audit market that is not supported by extant research. Even, worse, the
Commission could cause lower quality audits since there would be more frequent auditor
changes under a mandatory rotation rule, and therefore more frequent audits with short tenure
and potentially lower quality.
While acknowledging academic research on tenure and audit quality, the PCAOB (2011,
p. 17) explicitly cautions that we should be careful in drawing the conclusion that audit quality
preprint
accepted
manuscript
4

suffers following an auditor change. Their reasoning is that voluntary auditor changes often
occur in a broader context of auditor-client disagreements, or other client difficulties such as
financial distress and declining performance, and that these circumstances may “overstate” the
negative effects on audit quality following auditor changes that has been documented in
academic research. Lennox (2013) also notes the difficulty in determining causality in a
voluntary rotation setting.
Given these concerns, the purpose of our study is to investigate the potential negative
consequences of auditor rotation in a true mandatory rotation setting. As Lennox (2013)
concludes, there is limited research in such settings. Our objective is to provide evidence that
could guide intelligent policy making and help regulators to better assess the consequence of
fixed-term limits on auditor appointments. To do so we draw on the Italian experience with the
mandatory rotation of auditors of listed companies that has been required since 1975.
2

While the EC proposal is silent on negative consequences, the PCAOB (2011) identifies
three potential negative consequences of mandatory rotation that guides our investigation:
1. The PCAOB report (p. 38) notes that at the end of the mandatory fixed-term appointment, the
outgoing auditor may not have incentives to perform high-quality audits since they cannot be
reappointed. This may create a moral hazard problem in which the outgoing auditor shirks on
effort resulting in lower-quality audits.
3



2
While our focus is on Italy, there has been some research on mandatory audit firm rotation in Spain and Korea.
Ruiz-Barbadillo et al. (2009) examine the effect of the proposed mandatory rotation rule in Spain on the likelihood
the auditor issues a going concern audit report. While they find no statistical association, the authors point out that
the rotation rule was withdrawn before it officially went into effect, so that their study only measures the impact of
the "announcement" of a mandatory audit firm rotation rule. In Korea, an auditor change can be imposed on Korean
companies judged by the Financial Supervisory Commission as having high potential to manipulate accounting
results. In this setting, Kim and Yi (2009) find that there is less earnings management following a regulator-imposed
auditor change. However, Kim and Yi (2009, p. 207) recognize the uniqueness of the Korean auditor replacement
rule and note that their conclusions cannot be generalized to a mandatory rotation setting.
3
It is also possible outgoing auditors will be especially careful and put in greater effort because the new auditor will
review their work. (PCAOB 2011, p. 17). We find no evidence of greater effort, but we do find evidence of
abnormally higher fees in the final year, which suggests there may be opportunistic pricing by the outgoing auditor.
preprint
accepted
manuscript
5

2. The PCAOB report (p. 11) recognizes that rotation could impose significant switching costs
on clients, including larger audit fees if audit start-up costs are large and these costs are
passed on to the client.

3. The incoming auditor may not be able to initially perform high-quality audits due to learning
curve effects (PCAOB 2011, p. 13 and pp. 16-17). In other words, even though new auditors
may bring a “fresh perspective” to the audit, they may not be as competent or experienced as
the prior auditor, at least in the short-term. Alternatively, if the EC and PCAOB are correct in
their assertions, then audit quality could improve following a mandatory rotation.

Our main findings are as follows. First, for the outgoing auditor, there is no evidence of
lower-quality audits due to shirking in the final-year engagement. However, there is some
evidence of abnormally higher fees, as the final-year fees are 7 percent higher than normal. This
suggests there may some opportunistic pricing since we find no evidence of abnormally higher
audit effort in the final-year engagement. The PCAOB report does not identify this as a negative
consequence, but our evidence suggests it adds to the cost of mandatory rotation. Second, for the
incoming auditor, audit effort (hours) is abnormally higher by 17 percent in the initial
engagement, but initial fees are discounted by 16 percent relative to ongoing engagements.
However, we also find that future audit fees following the first-year audit are abnormally higher
by approximately 76 percent of first-year fees. Interestingly, the PCAOB (2011, p. 17) speculates
that mandatory rotation might eliminate low balling and subsequent fee recovery which they
view as a threat to auditor independence, but our evidence suggests otherwise. Third, we
document that earnings quality is lower during the first three engagement years relative to later
years of auditor tenure (larger abnormal accruals and less timely loss recognition). On average,
abnormal accruals are 36% larger in the first three years relative to later years of tenure. The
results in Italy are comparable to Johnson et al. (2002) in a voluntary auditor change setting who
find that earnings quality is lower in the first three-years relative to engagements with longer
tenure. Thus the consequences of mandatory rotation appear to be (1) higher audit fees, and (2)
preprint
accepted

manuscript
6

lower quality audited earnings following rotation, which is consistent with the evidence in non-
mandatory settings.
The next section discusses the mandatory auditor rotation requirement in Italy. This is
followed by the sample, descriptive statistics, and empirical results. In the concluding section we
discuss how a mandatory rotation rule might affect larger scale audit markets in countries such as
the United States.
AUDITOR ROTATION IN ITALY
Mandatory rotation was adopted in Italy in 1975 by Presidential Decree D.P.R. 136/1975
which required audit firm rotation for all listed companies in Italy, and was later extended to
unlisted companies in some regulated industries such as insurance.
4
Prior to the “Legge Draghi”
reforms in 1998, the Italian market was considered to be very thin with auditors competing for a
relatively small number of statutory audits (Gietzmann and Sen, 2002). Further reforms in 1998
and 2003 substantially increased the number of companies subject to external auditing including
all non-listed companies which are controlled by listed companies, non-listed companies that
control listed companies, and some private (limited liabilities) companies. Even though the audit
market expanded with the 1998 and 2003 changes, the market is still relatively small compared
to countries like the United States and United Kingdom.
The Italian institutional setting also has some distinctive features that make it an attractive
research setting with respect to auditor rotation. Mandatory rotation is potentially of greater
value in a country like Italy with a thin audit market, but we know little about the costs or

4
The law originally stipulated an appointment term of three years, renewable two times, for a maximum of nine
years. The cooling off period was set at five years, but was later reduced to three years by a Legislative Decree
issued in 1998, the so called “Legge Draghi.” The current requirement is a maximum auditor appointment of nine

years, with a three year “cooling off period,” as well as the required rotation of the engagement partner every six
years (and a three-year cooling off period before an engagement partner can be re-appointed).
preprint
accepted
manuscript
7

consequences of the rotation rule.
5
In addition, Italian audit regulation prohibits the auditor from
providing many types of non-audit services. As a result the Big 4 earn approximately 90 percent
of their revenues from audit-only services (Cameran et al. 2010), which means empirical results
are less likely to be confounded by non-audit services/fees. Lastly, there has been relatively little
research on auditor rotation in Italy, and the results to date are mixed as described below.
Concurrent Italian Research
Cameran and Pettinicchio (2011) document a steady increase over time in the market share
in Italy held by large international accounting firms, with an even stronger effect in the Italian
market segments subject to mandatory rotation. This is ironic because a rationale for rotation was
to increase market competition but it appears to have had the opposite effect, which illustrates
how unintended negative consequences can occur in audit market regulation. They also find
more partner suspensions for poor quality work related to the first year of an audit engagement,
compared to all other years, and this finding provides some evidence that audit quality may
initially suffer as a result of auditor rotation, consistent with a learning curve effect on new
audits (PCAOB 2011, p. 16). In other related studies, SDA Bocconi School of Management
(2001) surveyed internal auditors, managers, and Big 5 auditors of Italian listed companies and
reports that the first-year audit engagement requires more time by both the auditor and the client.
While mandatory rotation could potentially affect audit quality, there is no evidence the stock
market reacts to news of auditor changes (SDA Bocconi School of Management 2001; 2004).

5

Gietzmann and Sen (2002) argue that the mandatory rotation rule would have the most beneficial effects in thin
markets like Italy because a few large clients dominate such audit markets. As a result, they argue that the risk of
collusion between the auditor and a large client is more likely in absence of a mandatory rotation rule or strong legal
liability laws. Italy is a civil law country and is characterized as having weak legal enforcement and weak investor
protection (Choi and Wong 2007). Italy also has low litigation risk based on the index in Wingate (1997)
. The
litigation risk index score is 6.22 for Italy, while Anglo-Saxon countries report scores above 10, with a maximum
score of 15 for the US. The score assigned to Italy is equal to other non-Anglo-Saxon European countries like
France, Germany, Netherlands, Norway, and Switzerland.

preprint
accepted
manuscript
8

Our focus on rotation costs differs from prior studies in Italy that do not examine how
mandatory rotation affects fees and engagement hours. However, there are two concurrent
studies which also examine the effect of mandatory rotation on earnings quality Cameran et al.
(2012) use pre-IFRS data for a sample of 1,184 firm-year observations from 1985-2004 in which
the appointment term was on a three year basis, renewable two times (the so called “3+3+3”
rule). Their focus is on the auditor’s incentives to be reappointed at the end of the first and the
second 3-year appointment periods by allowing clients greater discretion to manage earnings in
these periods. In contrast they predict earnings quality will be higher in the third (last) 3-year
appointment term compared to the previous two, because the auditor cannot be reappointed and
therefore has no incentives to allow clients to manage earnings. Their results are consistent with
this prediction. In contrast, our study examines the effect on earnings quality immediately after
rotation (the first three years), and the more general relation between auditor tenure and earnings
quality. Livne and Pettinicchio (2012) report evidence suggesting that mandatory rotation does
not improve earnings quality. Their focus and research design differs from our study as follows.
They investigate a setting where firms are expected to have a demand for high-quality audits (as

proxied by complexity, leverage, and level of intangibles), and test if mandatory rotation
improves earnings quality for these firms. Their focus is not on auditor tenure, per se, and they
find that auditor rotation does not improve reporting quality.
6

RESEARCH APPROACH
We draw on the PCAOB (2011) to motivate our investigation of how mandatory rotation
may affect the behaviour of the outgoing and incoming auditors with respect of audit fees,
engagement hours, and audit quality. The following questions are examined:

6
Another concurrent study by Corbella et al. (2012) reports that earnings management declines in the year
immediately following mandatory audit firm rotation. We do not find this result in our data, but given the many
research design differences, it is difficult to directly compare the two studies.
preprint
accepted
manuscript
9

1. Does rotation cause outgoing auditors to shirk on effort and quality, in the absence of re-
appointment incentives, or, is the behaviour of outgoing auditors unaffected by
mandatory rotation?

2. Does rotation cause incoming auditors to have higher engagement hours and charge
larger audit fees due to start-up costs?
7


3. Does rotation cause incoming auditors to have lower quality in the short-term following
rotation due to learning curve effects, or are new auditors able to maintain or even

improve upon the quality of the prior auditor? The pro-rotation argument assumes the
new auditor will have a fresh perspective and can be more objective, which could lead to
higher quality audits post-rotation.

While we could make formal directional predictions for our analyses, there is no theoretical or a
priori basis for doing so. Instead, our study is descriptive in the sense that we are using the Italian
experience to examine the potential negative consequences of mandatory rotation.
A unique feature of our study is the use of private data on audit fees and engagement
hours provided to us by the Big 4 accounting firms in Italy. This allows us to accurately measure
the effect of mandatory rotation on audit engagement hours and audit fees. In order to assess the
cost of mandatory auditor rotation, we examine if audit fees and audit effort (engagement hours)
in the last year of the outgoing auditor, and the first year of the new auditor, are significantly
different than fees/hours of other engagement years. An additional consequence of rotation
would occur if audit quality changes around the rotation event, and to assess this possibility we
examine the quality of audited earnings (abnormal accruals and timely loss recognition).


7
Alternatively, the incoming auditor may engage in low balling. DeAngelo (1981) argues that low balling (fee
discounting) occurs in the expectation an auditor can set higher fees in the future and earn economic rents due to the
transaction cost advantages of incumbency. A fixed-term limit could actually increase the level of low balling if the
rotation rule makes the auditor’s tenure over the appointment term relatively more certain than would otherwise be
the case in a voluntary unregulated setting. For example, a nine-year appointment with certainty would have a larger
expected value relative to a situation in which there is only a 40 percent ex ante likelihood of having 15-year tenure
(i.e., an expected tenure of only 6 years). Earlier studies report evidence consistent with low balling (e.g., Simon and
Francis 1988; Hay et al. 2006), but a more recent study by Ghosh and Lustgarten (2006) finds low balling has
dissipated or disappeared altogether in the United States.

preprint
accepted

manuscript
10

SAMPLE AND DATA
The sample is comprised of 204 publicly-listed companies in Italy audited by the Big 4
accounting firms over the period 2006-2009, resulting in 667 firm-year observations in the
sample, although the specific sample size varies from test to test depending on data availability.
During the sample period, there were 232 non-financial companies listed on the Milan Stock
Exchange, so the sample represents approximately 88 percent of these companies.
8
The industry
composition is as follows: 37 percent of companies are in “manufacturing activities,” 13 percent
are in the “professional, technical and scientific services” sector, 10 percent in “information and
communications” and the rest of the sample is evenly distributed across 12 other industry
sectors.
9
The sample has 52 auditor changes, 36 of which are mandatory rotations (17.6 percent
of firms), plus another 16 auditor changes which are voluntary (7.8 percent of firms). For
mandatory rotations, the largest switches are from PricewaterhouseCoopers to Deloitte (22
percent), and from Deloitte to Ernst & Young (14 percent).
As mentioned above, a unique feature of the sample is that the Big 4 accounting firms in
Italy provided us with proprietary data on actual audit engagement hours and audit fees for the
consolidated group-entity accounts. We use this data in conjunction with publicly-available data
from consolidated annual reports, to measure the effect of auditor rotation on audit fees and audit
effort, as well as the quality of audited earnings around the rotation event. Using the audit fee
data provided directly by auditors allows us to be more accurate in our analysis. Publicly-

8
We do not examine listed companies in the financial sector as audit fees and accruals are fundamentally different
in this industry sector, nor did the Big 4 accounting firms provide us with data for this sector.

9
We use the ATECO industry classifications. This is the Italian version of the European nomenclature, Nace Rev.
2, published in the Official Journal of 20 December 2006 (Regulation (EC) no 1893/2006 of the European
Parliament and of the Council of 20 December 2006). This classification results in 16 industry sectors. The other 12
industry sectors in addition to four larger sectors mentioned in the text are: agriculture, forestry and fishing; minerals
extraction; electric energy and gas supply; water supply and garbage disposal activities; construction activities;
wholesale and retail trading; transport and storing activities; lodging and catering services; real estate; hiring
services and travel agencies; entertainment and sport activities; other services.
preprint
accepted
manuscript
11

disclosed audit fee data has been required since 2007 but we find some evidence that reported
fees in Italy can be inaccurate.
10
While we believe the fee data provided to us by the Big 4 firms
is more accurate, as a robustness test and sensitivity analysis, we also examine the differences in
publicly-disclosed fees and the proprietary fee data provided by the Big 4 accounting firms, and
re-estimate the models using publicly-disclosed fees.
11

Variable definitions are presented in Table 1, and descriptive statistics are reported in
Table 2. Median audit fees are 219,476 Euros, with an inter-quartile range of 129,185 to 395,933
Euros. Median engagement hours are 2,810, with an interquartile range of 1,694 to 5,105
12
. The
median absolute value of abnormal working capital accruals (AWCA) is 3.9 percent of sales. The
median firm in the sample has total assets of 457.8 million Euros, total debt equals to 65.3
percent of assets, ROA of 2.6 percent, inventories that are 6.7 percent of assets, and receivables

that are 23.1 percent of assets. The median firm has 14 subsidiaries, 3 operating segments, and
23.4 percent of the firm-years in the sample had an operating loss in year t-1.
[Insert Tables 1 and 2 here]
With respect to the auditor rotation variables, 5.4 percent of firm-years are coded one for
the first engagement year following a mandatory rotation (MANROT), and 7.3 percent of firm-
years are coded one for the last year of the outgoing auditor prior to rotation (PRE_MANROT).
For completeness we also examine voluntary auditor changes, where 2.4 percent of firm-years

10
The problem is that amount of fees paid to the auditor and disclosed in the annual report may not necessarily
relate to the current fiscal year being audited. In addition, although companies are required to make separate
disclosures of fees for audit and non-audit services, in some cases the amount of audit fees declared in the annual
report includes fees paid for work that is unrelated to the audit opinion and vice-versa (fees classified as non-audit
which are in fact audit related). For example, one company in our sample reports in its notes to the consolidated
financial statements a fee for auditing services of approximately two million Euros, which includes an unspecified
amount of fees related to a new share issue filing and therefore is not technically audit-related fees. Another
company in our sample reported as a non-audit fee, work related to the audit of a subsidiary company that was part
of the overall audit of the company’s consolidated financial statements, so these fees were incorrectly excluded.
11
While the results are qualitatively similar, the magnitude of low balling and subsequent fee recovery is much
larger using the publicly-reported data.
12
Descriptive statistics in Table 2 report natural logs of audit fees (LN_AF), audit hours (LN_H), and firm size
(SIZE), as these are the variables used in regression analyses.
preprint
accepted
manuscript
12

are coded one for the first engagement year following a voluntary rotation (VOLROT) and 1.6

percent of firm-years are coded one for the last year of the outgoing auditor prior to a voluntary
auditor change (PRE_VOLROT).The default comparison group is all other engagement years.
RESULTS
This section reports the empirical findings. Table 3 reports model estimations in which
the dependent variables are audit fees, total engagement hours, and abnormal working capital
accruals. All models are estimated with year fixed effects to control for systematic temporal
variation. The models are also estimated with firm fixed effects which is a strong control for
potential omitted variables bias (Lennox 2013).
13
The reported p-values are based on
heteroskedasticity-robust standard errors that are clustered by each unique company. P-values are
reported as two-tail probabilities since no directional predictions are made. There is no evidence
of multicollinarity threats as VIF’s are all under 4, well below the threshold of 10 suggested by
Kennedy (2008).
Audit Fees and Audit Hours
We estimate the audit fee model in equation (1) to test the effect of rotation on audit fees
around the rotation event, using private data supplied to us by the Big 4 accounting firms:
LN_AF
i,t

=
β
0
+ β
1
MANROT
i,t
+ β
2
VOLROT

i,t
+ β
3
PRE_MANROT
i,t
+ β
4
PRE_VOLROT
i,t
+
β
5
LN_H
i,t +
β
6
IND_SPEC
i,t +
β
7
SIZE
i,t
+ β
8
LEVERAGE
i,t
+ β
9
LAG_LOSS
i,t

+ β
10
ROA
i,t
+β
11
INV
i,t
+ β
12
REC
i,t
+ β
13
NSUB
i,t
+ β
14
FOREIGNREV
i,t
+ β
15
NSEG
i,t
+ β
16
CFO
i,t
+ Firm
Fixed Effects + Year Fixed Effects + ε

i,t
(1)
where LN_AF is the natural log of audit fees for firm i in year t.
14
The test variables of interest
are MANROT, which is an indicator variable denoting the first year of the new auditor

13
As explained later in the section “Robustness Tests” we re-estimate the models using industry fixed effects in lieu
of firm fixed effects with comparable results on the test variables.
14
Audit fees are the total fees (in euros) for a given engagement provided to us by the audit firms. Following prior
studies, we use the natural log of fees for the analysis.
preprint
accepted
manuscript
13

following mandatory rotation, and PRE_MANROT, which is an indicator variable denoting the
final year of the outgoing auditor. Following the audit fee literature (Hay et al. 2006), we also
control for industry specialization, firm size, leverage, prior-year loss, profitability, inventories,
receivables, number of subsidiaries, foreign operations, number of operating segments, and
operating cash flows. Because we have proprietary data on engagement hours (LN_H) it is also
included as a control variable and should be positively related to the level of audit fees.
15
We
also analyze voluntary auditor changes (VOLROT and PRE_VOLROT), and these tests are
reported later in the paper.
We begin with a baseline model of audit fees in column (1) in which firm-year
observations with auditor changes are deleted.

16
The model is significant at p-value <0.01 and
has high explanatory power with an R-square of 0.901. Audit fees are an increasing function of
audit hours and firm size, but other control variables are not significant. However, this is a
consequence of estimating the model with firm fixed effects, plus the control variable log of
audit hours. When we drop firm fixed effects and log of audit hours, then other variables become
significant and are consistent with prior studies: specifically, LEVERAGE, REC,
FOREIGNREV and NSEG are positively associated with audit fees, along with SIZE. However,
the model R-square is higher when firm fixed effects and audit hours are included.
[Insert Table 3 Here]
The test of audit fees in equation (1) is reported in Table 3, Column (2). The test variable
MANROT has a negative coefficient of -0.179 and is significant at p-value = 0.000 (two-tail).

15
We address the possibility that audit hours and audit fees are endogenously determined by estimating a “2SLS
regression model” where lagged audit hours are used as instruments for audit hours in the audit fee model
(Caramanis and Lennox, 2008). This approach yields results comparable to those reported in Table 3 and low balling
is confirmed (coef.: - 0.265; p =0.000). We also use the same approach for the fee models in Table 4, and again the
tabled results are confirmed.

16
This explains the different number of observations reported in column (1) of Table 3, compared to columns (2),
(3) and (4).
preprint
accepted
manuscript
14

The test variable PRE_MANROT has a positive coefficient of 0.071 and is significant at p-value
= 0.011 (two-tail). Using the procedures in Craswell et al. (1995) and Simon and Francis (1988),

the coefficients indicate a fee premium of 7.4 percent by the outgoing auditor, and a fee discount
of 16.4 percent by the incoming auditor.
17

Next we test the effect of mandatory rotation on audit engagement hours:
LN_H
i,t
=
β
0
+ β
1
MANROT
i,t
+ β
2
VOLROT
i,t
+ β
3
PRE_MANROT
i,t
+ β
4
PRE_VOLROT
i,t
+
β
5
IND_SPEC

i,t +
β
6
SIZE
i,t
+ β
7
LEVERAGE
i,t
+ β
8
LAG_LOSS
i,t
+ β
9
ROA
i,t
+β
10
INV
i,t
+ β
11
REC
i,t

+ β
12
NSUB
i,t

+ β
13
FOREIGNREV
i,t
+ β
14
NSEG
i,t
+ β
15
CFO
i,t
+ Firm Fixed Effects + Year Fixed
Effects + ε
i,t
(2)
where LN_H is the log of engagement hours.
18
Control variables are the same as in the fee model
in equation (1) because the same underlying factors that affect fees should also affect effort.
19

The effect of rotation on engagement hours is reported in column (3) of Table 3. The
model is significant and the R-square is 0.530. The coefficient on MANROT is +0.158 and is
significant at p-value = 0.014 (two-tail), which suggests an average increase in total audit hours
of approximately 17 percent, using the procedure described in footnote 17. The coefficient on
PRE_MANROT is insignificant (p-value = 0.796), indicating that the outgoing auditor does not
shirk on effort (hours) relative to ongoing engagements.
Putting the above two results together, for the outgoing auditor there is no evidence of
shirking because engagement hours are no different from ongoing engagements, but there is

17
Following Craswell et al. (1995), the fee premium is calculated as e
z
-1, where the z exponent value is the
coefficient value of 0.071 for the variable PRE_MANROT. The fee discount is 16.4 percent in the model in column
(2), calculated by taking the antilog of the coefficients as described in Simon and Francis (1988). Specifically, the
procedure calculates the percentage effect of the intercept shift on the logged dependent variable and is defined as 1-
(1/e
z
), where the z exponent value is the coefficient value of 0.179 for MANROT (i.e. the absolute value of the
coefficient).
18
This is the total number of hours, provided by the audit firms, charged to the clients for the audit engagement.
19
As a robustness test, we also include corporate governance variables (such as CEO duality, board size, and board
independence, as defined in model 3) which might possibly be related to the audit fees and audit hours. All results
hold with these additional controls. Moreover this allows us to compare the results on audit fees and audit hours for
a consistent model specification.
preprint
accepted
manuscript
15

some evidence of opportunistic pricing in the final year as fees are abnormally higher by 7.4
percent. While opportunistic pricing by the outgoing auditor is not mentioned by the PCAOB
(2011), it appears to be a potential cost of mandatory rotation. For the incoming auditor, the first
year audit results in 17 percent more effort (more engagement hours), but the overall fee is
discounted by an average of 16 percent. Thus the tests suggest low balling of approximately 33
percent: the fee discount of 16 percent, plus the uncharged additional hours which are 17 percent

higher in the first engagement year. Contrary to the PCAOB’s (2011) speculation, mandatory
rotation does not appear to eliminate low balling. We report evidence later in the paper that audit
fees increase with each year of auditor tenure following the initial engagement year, which is
consistent with the initial-year low balling amount being “recovered” through higher future fees.
Thus while fees are initially lower, they are subsequently higher which represents another cost of
mandatory rotation.
Earnings Quality
The analysis of auditor rotation on earnings quality uses abnormal working capital
accruals based and is estimated as follows:
AWCA
,t
=
β
0
+ β
1
MANROT
i,t
+ β
2
VOLROT
i,t
+ β
3
PRE_MANROT
i,t
+ β
4
PRE_VOLROT
i,t

+
β
5
IND_SPEC
i,t +
β
6
SIZE
i,t
+ β
7
LEVERAGE
i,t
+ β
8
LAG_LOSS
i,t
+ β
9
ROA
i,t
+ β
10
NSUB
i,t
+
β
11
FOREIGNREV
i,t

+ β
12
NSEG
i,t
+ β
13
CFO
i,t
+ β
14
IND
i,t
+ β
15
CEODUAL
i,t
+ β
16
BDSIZE
i,t

+Firm Fixed Effects + Year Fixed Effects + ε
i,t
(3)
where the dependent variable is absolute value of abnormal working capital accruals.
20

Following DeFond and Park (2001), abnormal accruals are calculated as the difference between

20

We do not include audit hours in the AWCA regression as audit hours are not a typical control variable in the
earnings quality literature. However, all results hold if we include audit hours as an additional control. In addition,
we include a control variable using an instrument for audit hours as described in footnote 15. The variable is
negative and significant indicating that higher audit hours are associated with smaller accruals, but the rotation test
preprint
accepted
manuscript
16

actual accruals and expected accruals based on the prior-year relation between a firm’s sales and
its working capital accruals. Later, as a robustness test, we report results using timely loss
recognition (Basu, 1997). The control variables are the same as those in the tests of fees and
audit hours except for INV and REC, which are dropped as they are more specific to the fee
model literature. In addition, following prior earnings quality research (e.g., Anderson et al.,
2004; Xie et al., 2003; and Dechow et al., 1996), we include three additional controls for the
corporate governance effects in constraining low-quality of earnings: the percentage of
independent directors on the board (IND), the separation of CEO and Chair positions
(CEODUAL) following Dechow et al. (1996), and board size (BDSIZE) based on (Klein, 2002).
The test of earnings quality is reported in Column (4) of Table 3. The two test variables
of interest, MANROT and PRE_MANROT, are not significant at the 0.10 level. We also test
separately those observations with positive abnormal accruals and those with negative abnormal
accruals. The rotation variables continue to be insignificant. There is no evidence that rotation
causes a decline in first-year earnings quality (relative to other engagement years). However, the
next section shows that earnings quality improves with tenure and that the first three years of
earnings following a mandatory rotation are significantly lower in quality, on average, compared
to those engagements with longer auditor tenure.
Alternative Analysis Using Auditor Tenure
An alternative approach to assessing mandatory auditor rotation is to examine the more
general effect of auditor tenure on fees, hours, and earnings quality. To do so, we create an audit
firm tenure variable (FTEN) which is coded 0 for the first year of auditor tenure, 1 if the second

year, 2 if the third year, and so on. By coding the first-year audit as zero, this is the default

variables of interest are unchanged when audit hours are added as an additional control. Finally, as an additional
robustness test, we include all of the control variables from the fee/hours models, and the results are qualitatively
unchanged.
preprint
accepted
manuscript
17

comparison group and the variable FTEN captures the time trend effect after the first-year
engagement, rather than the one-period snap shot immediately after rotation as reported in Table
3. We use this alternative metric and re-estimate the models in Table 3 as follows:
LN_AF
i,t
=
β
0
+β
1
FTEN
i,t
+ β
2
LN_H
i,t
+ β
3
INDSPEC
i,t

+β
4
SIZE
i,t
+ β
5
LEVERAGE
i,t
+
β
6
LAG_LOSS
i,t
+ β
7
ROA
i,t
+β
8
INV
i,t
+ β
9
REC
i,t
+ β
10
NSUB
i,t
+ β

11
FOREIGNREV
i,t
+ β
12
NSEG
i,t

+ β
13
CFO
i,t
+ Firm Fixed Effects + Year Fixed Effects + ε
i,t
(4)
LN_H
i,t
=
β
0
+β
1
FTEN
i,t
+ β
2
INDSPEC
i,t
+β
3

SIZE
i,t
+ β
4
LEVERAGE
i,t
+ β
5
LAG_LOSS
i,t
+
β
6
ROA
i,t
+β
7
INV
i,t
+ β
8
REC
i,t
+ β
9
NSUB
i,t
+ β
10
FOREIGNREV

i,t
+ β
11
NSEG
i,t
+ β
12
CFO
i,t
+ Firm
Fixed Effects + Year Fixed Effects + ε
i,t
(5)
AWCA
i,t
=
β
0
+β
1
FTEN
i,t
+ β
2
INDSPEC
i,t
+β
3
SIZE
i,t

+ β
4
LEVERAGE
i,t
+ β
5
LAG_LOSS
i,t
+
β
6
ROA
i,t
+ β
7
NSUB
i,t
+ β
8
FOREIGNREV
i,t
+ β
9
NSEG
i,t
+ β
10
CFO
i,t
+ β

11
IND
i,t
+ β
12
CEODUAL

i,t
+ β
13
BDSIZE
i,t
+Firm Fixed Effects + Year Fixed Effects + ε
i,t
(6)
The results of estimating these models with FTEN as the test variable are reported in
Table 4, Panel A. All models are significant at p-value < 0.01, and there is no evidence of
multicollinarity threats as VIF’s are all under 4, well below the threshold of 10 suggested by
Kennedy (2008). The effect of tenure on audit fees is reported in Table 4, Panel A, Column (1).
The test variable FTEN is positive and significantly related to audit fees (coef.: 0.021; p-value
0.000, two-tail), which means that audit fees increase by 2.12 percent with each additional year
of auditor tenure following the initial engagement year (using the procedure in footnote 17). For
example, the year 2 fee (two years after the first year) is higher by two times 2.12 percent (4.24),
the year 3 fee is higher by three times 2.12 (6.36), and so on. To estimate the cumulative effect
over the eight-years of audit tenure after the first year (assuming the maximum nine-year limit),
we sum the yearly effects, i.e., the sum of years 1 through 8, or 36, and multiply 36 by the yearly
increase of 2.12 percent. This gives a cumulative increase in abnormal audit fees of 76 percent
relative to the first-year engagement fee, i.e., those observations where tenure is coded zero.
preprint
accepted

manuscript
18

Note that since audit engagement hours are included in the model as a control variable, the
subsequent fee increases would appear to occur through higher hourly rates.
This result is consistent with low balling theory which predicts fee discounting in the
initial year in the expectation of higher future fees. Recall that the estimated amount of initial
engagement low balling is 33 percent (a combination of fee cutting and uncharged higher
engagement hours). Our result is in contrast to other countries where a continuous auditor tenure
metric has generally been insignificant (Hay et al. 2006), and is at least suggestive that low
balling may be greater in the Italian mandatory rotation setting where there is less uncertainty
about the length of the auditor’s tenure compared to settings where auditor changes are voluntary
and could occur at any point in time. The point is that a fixed-term appointment under a
mandatory regime, may, surprisingly, lead to higher levels of low balling because the auditor is
pretty much guaranteed a nine-year appointment. The low balling amount is more than recouped
later through higher audit fees, which is another consequence of mandatory rotation. We believe
that these abnormal post-rotation audit fees represent an implicit cost of mandatory rotation in
Italy because, arguably, these abnormal fees would not have occurred in the absence of the low
balling induced by mandatory auditor rotation.
[Insert Table 4 Here]
The effect of tenure on engagement hours is reported in Table 4, Panel A, Column (2),
and the tenure variable FTEN shows no statistical association. Given that an auditor typically
stays for the maximum nine-year term, there may be little incentive for the auditor to pass on
learning curve savings to clients, particularly given the above evidence that abnormal audit fees
are increasing with tenure. In other words, it would be difficult to justify higher fees in the
presence of lower engagement hours.
preprint
accepted
manuscript
19

We also examine the association of tenure with abnormal working capital accruals. The
test variable FTEN is negatively associated with abnormal working capital accruals (AWCA).
The coefficient is -0.004 and is significant at p-value = 0.045 (two-tail). This analysis suggest
that earnings quality improves with each year of tenure after the first engagement year, and is
consistent with previous tenure literature in settings where auditor changes are voluntary (Myers
et. al., 2003; Lennox 2013).
In Table 4, Panel B, we repeat the analyses using an alternative indicator variable for
tenure, PERIOD_1 which compares the first three years of a new auditor following rotation with
all engagements of longer auditor tenure. The intuition is that low balling may occur over several
periods, and earnings quality could be impaired for several years during the initial learning phase
by the new auditor. Results in Table 4, Panel B, Column (1) shows that in the first three years
following auditor rotation a low balling effect is present (coef.:- 0.117; p-value = 0.000, two-
tail). Column (3) of the same table shows that abnormal accruals are also larger (coef.: 0.031; p-
value = 0.048, two-tail), indicating lower quality earnings in the first three years relative to later
periods of auditor tenure. Based on the coefficient value of 0.031 for the test variable Period_1,
abnormal working capital accruals (scaled by assets) increase by amount equivalent to 3.1% of
lagged assets. This represents a 36% increase in abnormal accruals, measured as 3.1% divided by
the sample mean value for AWCA of 8.5% in Table 2. Thus, compared to Table 3 where
earnings quality is not significantly different in the first engagement year (relative to other
years), we find that the first three years are of lower quality, on average, compared to earnings in
later years of auditor tenure. We believe the three-period test has greater statistical power than
the one-period test in Table 3, given that earnings appear to be of lower quality for multiple
periods following rotation.
preprint
accepted
manuscript
20

We re-estimate the model in column (3) of Table 4 separately for observations with

positive abnormal accruals and observations with negative abnormal accruals. For observations
with positive accruals, audit firm tenure is negatively associated with positive accruals: FTEN
has a coefficient of -0.006 (p = 0.030, two-tail). For observations with positive accruals, the
relation between accruals and PERIOD_1 is positive: the coefficient of PERIOD_1 is 0.037 (p =
0.092, two-tail). These results are consistent with the full sample results in Table 4. For
observations with negative abnormal accruals, the test variables FTEN and PERIOD_1 are not
significant at the 0.10 level. Thus the results in Table 4 are driven by positive (income-
increasing) abnormal accruals, which arguably have a more serious effect on earnings quality.
21

Voluntary Auditor Changes
The sample includes 16 voluntary auditor changes, and the models in Tables 3 include
two test variables for these observations: VOLROT, which is coded one for the first year of the
new auditor, and PRE_VOLROT, which is coded one for the last year of the outgoing auditor. In
Table 3, there is no evidence that audit fees are affected by a voluntary auditor change. Neither
the last-year fee of the outgoing auditor nor the first-year fee of the new auditor is significantly
different from fees of ongoing audit engagements. The same is true of audit engagement hours as
both VOLROT and PRE_VOLROT are statistically insignificant at the 0.10 level. However, for
abnormal working capital accruals, Column (4) of Table 3 shows that abnormal accruals are
significantly smaller (p-value = 0.051, two-tail) in the last year of the outgoing auditor, although
accruals in the first year of the new auditor are not significantly different from ongoing audits.
This result is broadly consistent with DeFond and Subramanyam (1998) in a study of voluntary

21
We also re-examine the year preceding a voluntary rotation and find that the variable PRE_VOLROT is
negatively associated with positive accruals (coef.: -0.070; p. 0.001), and positively associated with negative
accruals (coef.: 0.041; p. 0.099). In other words, in the last year of the outgoing auditor, positive accruals (income-
increasing) are smaller, and negative accruals are larger. Both would have the effect of reducing the level of reported
earnings relative to other engagement years.

preprint
accepted
manuscript
21

auditor changes in the United States. Specifically, they find that abnormal accruals are smaller
and reduce the level of reported earnings in the period leading up to an auditor change. They
interpret this as evidence of excessive auditor conservatism, and suggest that this conservatism
motivated the change to another (potentially less conservative) auditor. We observe a similar
pattern in the sample of voluntary auditor changes in Italy.
22

Alternative Analysis Using Publicly-Disclosed Audit Fees

Audit fees have been disclosed in annual reports of Italian listed companies since 2007,
and we previously discussed why these disclosures may be inaccurate. At the same time, we
cannot be certain of the accuracy of the private data provided to us by the Big 4 accounting
firms. Therefore, as a robustness test, we re-estimate the audit fee model in equation (1) using
publicly reported audit fees for a reduced sample over the period 2007-2009 in which audit fees
are publicly available.
23
Untabulated results indicate that first-year fees are significantly lower at
p-value <0.01. The coefficient is -0.259, which represents a fee discount of 22.8 percent relative
to ongoing audit fees, compared to a discount of 16.4 percent in Table 3. Using the same
approach as before, we estimate that future fees are abnormally higher by approximately 130
percent of the first-year fee, versus 76 percent using the data provided by the Big 4 firms. In
addition, with the public fee data we find evidence of larger opportunistic pricing in the final
year of the outgoing auditor of approximately 29.4 percent. Overall, while the results are
qualitatively the same, the magnitudes are somewhat larger using publicly-reported data.

22
We also estimate the models in Table 4 Panel B, column 1, 2 and 3, separately for mandatory rotations and
voluntary rotations. All of our main results are confirmed for the mandatory rotations, while voluntary rotations are
not significant.
23
In our sample for the period 2007-2009, publicly-disclosed fees are 23 percent larger, on average, than the fees
reported to us by the Big 4 accounting firms. The Pearson correlation is 0.80 between the fee data provided to us by
the Big 4 firms and the publicly-disclosed fees. However, the correlation between fees and engagement hours is only
74.3 percent using publicly-disclosed fees, compared to 97.2 percent when using the Big 4 fee data.
preprint
accepted
manuscript
22

Additional Robustness Tests and Sensitivity Analyses
We conduct a set of additional analyses to test the robustness of our results. We repeat
the analyses using industry fixed effects in lieu of firm fixed effects because prior research has
used industry fixed effects.
24
The results are consistent using industry fixed effects, with one
exception. Audit hours are not significantly higher in the first year of a new engagement. We
also repeat all the analyses by excluding observations with voluntary rotations, and the results for
mandatory rotations are unchanged. In our investigation on how mandatory rotation affects audit
fees, we re-estimate the model in Table 3, column (2) using a different specification for the
dependent variable. The variable LN_AF is replaced by the log of audit fees divided by audit
hours. Results are qualitatively the same: for the test variables of interest, MANROT is negative
and significant (-0.215, p-value=0.000), and PRE_ MANROT is positive and significant (0.073,
p-value=0.019).
We also attempt to rule out an alternative interpretation of the results. The results indicate

that earnings quality is similar in year 9 (prior auditor) and year 1 (new auditor). At face value
this suggests that rotation does not worsen audit quality in the short-term. To explore this further,
we delete firm-years in which auditor tenure is years 4, 5 and 6. We then create a dichotomous
tenure variable with a value 1 if tenure equals year 1, 2, or 3, and zero if tenure equals years 7, 8,
or 9. The purpose is to test a longer window around the mandatory rotation event. This analysis
shows that abnormal accruals are significantly larger (coef.: 0.054; p-value = 0.082, two-tail) in
the first three years, relative to the last three years preceding mandatory rotation (years 7, 8 and
9). This analysis is consistent with the results in Table 4, and further suggests that the one-year
snap shot around the rotation year may have low power and does not give an accurate picture of
the effect of rotation on earnings quality. We repeat the analysis in Table 4, Panel B, Column 3,

24
Year fixed effects remain and we continue to cluster standard errors at the firm level.