ruizâ-barbadillo et al - 2009 - does mandatory audit firm rotation enhance auditor independence - evidence from spain [mafr]

We test two com- peting hypotheses concerning the impact of mandatory rotation on the likelihood of auditors’ issuing going-concern modified audit opinions.. Our results suggest that aud

AUDITING: A JOURNAL OF PRACTICE & THEORY American Accounting Association

of audit firms on auditor independence using Spanish archival data Rotation of audit firms every nine years was mandatory in Spain from 1988–1995 Although the rule was never enforced, the Spanish context provides a unique setting to examine the effects that mandatory audit firm rotation has on auditor behavior We examine audit reports for a sample of financially stressed companies from 1991–2000 to compare audit re- porting behavior in a regime with rotation (mandatory rotation period: 1991–1994) and one without rotation (post-mandatory rotation period: 1995–2000) We test two com- peting hypotheses concerning the impact of mandatory rotation on the likelihood of auditors’ issuing going-concern modified audit opinions We find no evidence to sug- gest that a mandatory rotation requirement is associated with a higher likelihood of issuing going-concern opinions Our results suggest that auditors’ incentives to protect their reputation have a positive impact on the likelihood of issuing going-concern opin- ions, while auditors’ incentives to retain existing clients did not impact on their deci- sions in both the mandatory rotation and post-mandatory rotation periods Overall, our results provide empirical support for the arguments put forward by opponents of man- datory rotation.

INTRODUCTION

The objective of this paper is to examine the impact of mandatory audit firm rotation

on auditor independence.1 Motivation for this study comes from: (1) the interest ofregulators and legislators in many countries regarding mandatory auditor rotation as

1 Hereafter, we use the terms ‘‘rotation’’ and ‘‘auditor rotation’’ to mean the ‘‘rotation of audit firms,’’ unless stated otherwise.

Emiliano Ruiz-Barbadillo is a Professor and Nieves Go´mez-Aguilar is an Associate fessor, both at the University of Ca´diz, and Nieves Carrera is an Associate Professor at Instituto de Empresa.

Pro-We acknowledge helpful suggestions from Marleen Willekens (associate editor), two anonymous reviewers, Salvador Carmona, Isabel Gutierrez, Christopher Humphrey, and Tashfeen Sohail on earlier drafts of this paper.

We gratefully acknowledge the financial support provided by the Spanish Ministry of Education and Science (SEJ2006-14021 and SEJ2004-08176).

Editor’s note: Accepted by Marleen Willekens, under Dan Simunic’s editorship.

Submitted: June 2005 Accepted: October 2008 Published Online: May 2009

a mechanism for strengthening auditor independence (e.g., U.S General Accounting Office[GAO] 2003, 2004; U.K Coordinating Group on Accounting and Auditing Issues [CGAA]2003), and (2) the need for empirical research on the effects of such a policy (Dopuch et

al 2001; DeFond and Francis 2005).2

The mandatory rotation of auditors has been a subject of debate by practitioners(AICPA 1978), academicians (Geiger and Raghunandan 2002), and regulatory bodies (GAO2003) Subsequent to the collapses of Enron and WorldCom, legislators and regulatorsacross the world were compelled to discuss the adoption of a policy of mandatory rotation(GAO 2003; CGAA 2003) In fact, in the U.S., there were calls for the immediate imple-mentation of mandatory rotation of audit firms, although those provisions were rejected.Nevertheless, the Sarbanes-Oxley Act of 2002 mandated the U.S Comptroller General toconduct a study of the potential effects of mandatory rotation The U.S GAO undertookthis study (GAO 2003, 2004) and concluded that audit firm rotation ‘‘may not be the mostefficient way to strengthen auditor independence and improve audit quality’’ (GAO 2003,2) However, the GAO left open the possibility of revisiting mandatory audit firm require-ments, if deemed necessary (Nagy 2005).3Countries such as Brazil, India, Italy, Singapore,and South Korea have adopted rotation policies on the grounds that they may reduce threats

to auditor independence by avoiding long-term relationships between auditors and theirclients (e.g., GAO 2003)

The issue of mandatory audit firm rotation has been an area of interest in the auditingliterature for quite some time.4On the one hand, it is suggested that the auditor’s economicdependence on existing clients and managers’ influence over the reappointment of the in-cumbent auditor may affect auditor reporting behavior (DeFond et al 2002) Thus, a limit

on the time horizon for an audit engagement could improve audit quality by reducing theauditor’s incentives to issue biased reports as well as decreasing management’s ability toinfluence the auditor’s decisions In contrast, it is argued that rotation is unnecessary be-cause market-based incentives, such as threat of loss of reputation and litigation costs,dominate the expected benefits from compromising auditor independence (AICPA 1992;DeFond et al 2002; Geiger and Raghunandan 2002)

While there is a sizable amount of literature both supporting and rejecting the mentation of mandatory rotation, direct tests of this policy have been limited, as it is

imple-‘‘difficult to obtain empirical evidence on the costs and benefits of a proposed regulationprior to its implementation’’ (Dopuch et al 2001, 94).5 An evaluation of the rotation rule

is nevertheless possible in countries that have adopted a mandatory audit firm rotationpolicy One such country is Spain, where the mandatory rotation rule was in force from

2 As noted by DeFond and Francis (2005, 6), ‘‘Because there is a realistic concern that mandatory audit firm rotation may yet be proposed by the SEC, we encourage more research in this area.’’

3 Similarly, in the U.K., the CGAA analyzed the case for the mandatory rotation of audit firms The final report did not recommend its implementation (CGAA 2003) Rotation has been discussed and rejected in other coun- tries, including Canada and Ireland In Canada, the Standing Senate Committee on Banking Trade and Commerce (2003, 21) concluded: ‘‘We do not support a requirement for rotation of the audit firm, since in our view valuable company-specific experience would be lost.’’ In Ireland, the Review Group on Auditing concluded that

‘‘the introduction of mandatory auditor rotation could undermine the effectiveness of audits’’ (Department of Enterprise, Trade and Employment, Ireland 2000, 188) In Greece, the rotation of audit firms used to be man- datory, but the requirement was removed in 1994 (Arrun˜ada and Paz-Ares 1997).

4 For reviews, see Gietzmann and Sen (2002) and Cameran et al (2005).

5 We are not aware of any empirical study using archival data to examine the impact of mandatory rotation on auditor reporting behavior However, various studies have analyzed audit quality in settings with similar features

to those operating under mandatory rotation Nagy (2005) and Kealey et al (2007) analyzed the compulsory forced change of audit firms caused by the disappearance of Arthur Andersen Other authors have focused on the effects of audit partner rotation on audit quality (see, for example, Carey and Simnett 2006).

Does Mandatory Audit Firm Rotation Enhance Auditor Independence? Evidence from Spain 115

American Accounting Association

1988 until 1995 The 1988 Spanish Audit Law established a system where an auditor’sappointment could last for no less than three and no longer than nine years (Audit Law,art 8.4).6 At the end of the appointment period, the audit firm had to be replaced, andreappointment could not be sought for another three years The first audit contracts subject

to the mandatory rotation requirement were signed in 1988 The rotation rule was removed

in March 1995 by means of a provision included in the Limited Liability Companies Act

of 1995 (Ley de Sociedades de Responsabilidad Limitada 1995) After its removal, auditees

could renew an audit contract on a yearly basis once the initial contract expired (for moredetails, see Carrera et al 2007)

The change in the regulatory environment—from a regime with mandatory rotation toone without—provides a unique setting to analyze the regulatory effectiveness of audit firmrotation For seven years (1988–1994 inclusive), auditors’ economic incentives and report-ing decisions were conditioned by, and subject to, the existence of mandatory rotation Thatmandatory rotation was removed before any audit firm was actually required to leave aclient does not limit the contribution of our research On the contrary, a rule intended toenhance auditor independence is expected to influence auditor reporting behavior duringthe engagement period

We contribute to the literature by providing empirical evidence on the impact of datory rotation on auditor independence, using auditors’ propensity to issue going-concernaudit opinions as a proxy for auditor independence We examine the impact of rotation

man-on audit reporting behavior by comparing the mandatory rotatiman-on period (1991–1994) tothe post-mandatory rotation period (1995–2000) Specifically, we test whether the rotationrule affected auditor decision making by moderating auditor incentives (1) to maintain

quasi-rents from existing clients (economic dependence) and (2) to safeguard firm reputation (reputation protection).

Using established research methodology, we analyze the relationship between auditorreporting behavior and auditor independence for a sample of financially distressed com-panies (e.g., DeFond et al 2002; Geiger and Rama 2003).7 The underlying assumption of

our analysis is that, ceteris paribus, the auditor’s propensity to issue a going-concern

opin-ion to distressed companies is positively correlated with the auditor’s level of independence(e.g., DeFond et al 2002, 1249; Knechel and Vanstraelen 2007)

The auditing literature has analyzed the economic trade-off faced by an auditor whenassessing the going-concern assumption (Krishnan and Krishnan 1996; Louwers 1998;Reynolds and Francis 2001) On the one hand, the auditor faces the risk of losing a client

when issuing a qualification (economic dependence); on the other, failing to qualify the company’s accounts exposes the auditor to the risk of reputation loss (reputation protection).

It is noteworthy that prior research has consistently provided evidence that auditor tives are contingent on the legal regimes in which they operate (DeFond and Francis 2005).Assuming that rotation influences auditor reporting behavior by modifying auditor incen-tives and bolstering the possibility of independent action (Johnson et al 2002; Carcello andNagy 2004), we would expect to see different results in Spain on comparing the mandatory

incen-6 The Spanish Audit Law was enacted in 1988 in response to the Company Law Directives of the European Economic Union For the first time, firms were legally required to audit their financial statements.

7 The auditing literature has devoted considerable attention to auditor decisions in the evaluation of the concern status of companies because it is one of the most difficult and complex decisions faced by an auditor (Louwers 1998) Various academic studies have examined the relationship between auditor decisions and audi- tor independence in dealing with financially distressed companies (Craswell et al 2002; DeFond et al 2002; Geiger and Rama 2003).

going-and post-mgoing-andatory rotation periods Our competing hypotheses test whether mgoing-andatoryrotation mitigates the effect of economic dependence, thereby enhancing auditor indepen-dence, or, alternatively, whether mandatory rotation adversely affects auditor reputationconcerns and thereby undermines reputational capital as a mechanism to maintain auditorindependence.

Using a sample of 1,326 financially distressed Spanish companies during the period1991–2000, we find no evidence to suggest that a mandatory rotation requirement is as-sociated with a higher propensity for auditors to issue a qualified audit opinion Our analysisfinds no significant association between the auditor’s level of economic dependence andthe likelihood of issuing a going-concern report in both the mandatory rotation and post-mandatory rotation periods We do, however, find a positive association between auditors’incentives to protect their reputation and the likelihood of issuing going-concern opinions,regardless of the existence of a mandatory rotation regime Such positive association isstronger in the post-mandatory rotation period Overall, our results are consistent with theconcerns voiced by those opposing the implementation of mandatory audit firm rotation.Sensitivity analysis confirms that our results are robust across different subsamples andalternative model specifications, thereby providing further support to our findings

The remainder of the paper is organized as follows The next section discusses thepossible effects of mandatory auditor firm rotation on auditor independence and developsour hypotheses Then, we describe the sample data and methodology Following this, wepresent the results and robustness tests The final section discusses the study’s implicationsand limitations

HYPOTHESES Arguments Supporting Mandatory Audit Firm Rotation

Auditor independence may be adversely affected by long-term client relationships andthe desire to retain a client (GAO 2003) Previous research demonstrates that a company

is more likely to retain its auditor when the auditor gives the company a clean opinion,compared with situations where there is disagreement between the auditor and its client(Antle and Nalebuff 1991) Under such a scenario, auditors face the threat of dismissal ifthey seek to provide a qualified opinion Thus, if the auditor perceives the client as thesource of a perpetual annuity (Carcello and Nagy 2004), the economic interest of incumbent

auditors on their clients (economic dependence) may serve to increase management’s ability

to influence auditor reporting decisions As suggested by Gietzmann and Sen (2002), ditors that are capable of being indefinitely reappointed can be very concerned about main-

au-taining their existing client base, making them more susceptible, ceteris paribus, to collude

with management

Mandatory rotation has been advocated as one measure to overcome this collusionproblem Prior studies note that when audit firm rotation is fixed, auditor independence isless likely to be threatened because management cannot reengage the incumbent auditorindefinitely (Copley and Doucet 1993) In an experimental study, Dopuch et al (2001)found that auditors in a regime with rotation requirements were less willing to issue biasedreports than those in a regime without mandated rotation As the value of the individualquasi-rents associated with each client is lower, auditors have fewer incentives to give biasedopinions in response to management pressure, thereby mitigating the potential negative

impact of economic dependence on auditor decision making Accordingly, we expect that

an auditor will be more likely to issue going-concern modified audit opinions in a regimewith mandatory rotation than in a regime without rotation This leads to our first hypothesis

Does Mandatory Audit Firm Rotation Enhance Auditor Independence? Evidence from Spain 117

American Accounting Association

H ROTATION PROPONENTS : An auditor’s propensity to issue a going-concern modified

audit opinion to stressed companies is higher in the datory rotation period than in the post-mandatory rotation

man-period due to the economic dependence effect.

Arguments against Mandatory Audit Firm Rotation

Opponents of mandatory rotation argue that the audit market provides strong economicand institutional incentives for auditor independence, making mandated rotation unneces-sary.8 Prior research has shown that auditors’ incentives to protect firm reputation have amajor impact on auditor independence and, therefore, on audit quality (DeAngelo 1981;Reynolds and Francis 2001; DeFond et al 2002) An auditor’s reputation for performinghigh-quality audits is positively associated with the ability to earn higher fees and attractclients (Craswell et al 1995) Accordingly, the loss of reputation caused by the publicdisclosure of an audit failure can impose significant costs on auditors as it may significantlyreduce the present value of future revenue streams from both audit and nonaudit services(Krishnan and Krishnan 1996) As such, reputation concerns motivate auditors to maintainindependence because, in the event of an audit failure following a perceived lack of inde-pendence, they risk losing future quasi-rents (AICPA 1992) Thus, it could be argued that

in relatively well-developed audit markets, the reputation effect associated with the potentialloss of future business is sufficiently strong to prevent the risk of collusion between auditorand client, making the provision of rotation rules unnecessary.9 These arguments suggestthat reputation concerns may help maintain auditor independence

An auditor’s motivation to build firm reputation may differ from a regime with rotationrequirements to a regime without rotation Arrun˜ada and Paz-Ares (1997) suggest thatauditors’ incentives to build reputation are lower under forced rotation because rotation

‘‘drastically limits the possibility of economically realizing a substantial part of the tational capital of the audit firm’’ (Arrun˜ada and Paz-Ares 1997, 55–56) Accordingly,

repu-because of the reputation protection effect, the auditor’s propensity to issue qualified

going-concern opinions can be expected to be lower in the mandatory rotation period comparedwith the post-mandatory rotation period This leads to our second hypothesis

H ROTATION OPPONENTS : An auditor’s propensity to issuing a going-concern

modi-fied audit opinion to stressed companies is lower in themandatory rotation period than in the post-mandatory ro-

tation period due to the reputation protection effect.

DATA AND RESEARCH DESIGN Sample

Our analysis was conducted using a sample of companies extracted from the database

of the Spanish Securities and Exchange Commission (Comisio´n Nacional del Mercado de

8 The audit profession has strongly opposed the introduction of mandatory rotation Using cost-benefit analysis, the profession argues that there are doubts about whether the potential benefits derived from mandatory rotation overcome the associated transaction costs (AICPA 1978, 1992) The rotation rule is not costless, because it leads

to startup costs’ being incurred periodically As a result, under mandatory rotation, audit costs tend to increase, and audit fees are higher (Arrun˜ada and Paz-Ares 1997).

9 In addition, rotation may decrease audit quality as a result of the disruption of the ongoing relationship between the auditor and the auditee Previous research has shown that audit quality is lower in the early years of the engagement because the auditor is unfamiliar with the client’s business and industry (Carcello and Nagy 2004) Recent studies have provided empirical evidence on the higher frequency of audit failures when auditors perform their first and second audits (Geiger and Raghunandan 2002; Carcello and Nagy 2004).

Valores, CNMV) during the period 1991–2000 The database contains audited financial

information and the audit reports of all listed companies on the Madrid Stock Exchange.The database includes 4,817 audited company years for the period 1991–2000 Followingprior research (e.g., Geiger et al 2005), financial services and insurance companies wereexcluded because their financial ratios differ significantly from those in the nonfinancialsector, which could potentially generate misleading results Companies already in liquida-tion were excluded because there is no doubt concerning their going-concern problems and

so the audit reports provide little additional information to potential users Following thisprocedure, our sample was reduced to 3,119 observations

Consistent with prior studies, we analyzed a sample of financially distressed firms (e.g.,Hopwood et al 1994; DeFond et al 2002).10Specifically, we included only those companiesfor which auditors may reasonably be expected to issue a going-concern modified reportgiven their level of financial distress.11 Following Hopwood et al (1994) and Geiger andRaghunandan (2002), we classified a company as financially stressed if it exhibited at leastone of the following financial stress signals: (1) negative working capital, (2) negativeretained earnings, and (3) a bottom-line loss These variables are considered to be contraryfactors in Spanish accounting guidelines (ICAC 1991) In addition, we monitored the ex-istence of other factors that could potentially mitigate problems with financial distress.When there are mitigating factors, the auditor may feel justified in not qualifying the auditreport, even if there are going-concern problems Using a methodology similar to that ofReynolds and Francis (2001), we examined the subsequent fiscal year’s financial statements

to identify important sales of assets or the issuance of new debt or equity These mitigatingfactors may affect auditors’ reporting decisions Accordingly, companies that exhibited ei-ther or both of these factors were excluded This resulted in a total sample of 1,326 finan-cially stressed-company years, including 90 going-concern opinions, of which 33 were first-time going-concern modified audit opinions

Research Design

To assess the impact of mandatory rotation on auditor reporting behavior, we examinethe propensity of going-concern audit opinions in the mandatory rotation and the post-

mandatory rotation periods We initially use the following logistic regression model (main

effects model) to estimate the likelihood of an auditor issuing a going-concern opinion:

⫹ ␤4PROBFAIL⫹ ␤5LOSS⫹ ␤6LEVERAGE ⫹ ␤7SIZE

The variables are defined as follows:

GCO A dummy variable with a value of 1 if a going-concern modified audit report is

issued, 0 otherwise To identify going-concern uncertainties, we paid close attention to theexplanatory paragraphs in the audit reports In the case of mentioning going concern

10 As noted by DeFond et al (2002, 1255), ‘‘this is because the going-concern opinion is most salient among distressed firms.’’

11 Wilkerson (1987) suggested that, when investigating qualification decisions, it is important to collect mental and control samples of companies in which the overall degrees of economic uncertainty associated with such companies are similar If this is not done, then any differences may be related to differences in economic uncertainties, rather than to auditors’ decision making (Hopwood et al 1994).

experi-Does Mandatory Audit Firm Rotation Enhance Auditor Independence? Evidence from Spain 119

American Accounting Association

and / or expressed doubts regarding the ability of the firm to continue in existence, and / or

a going-concern disclaimer, GCO was coded as 1.

ROTATION A time dummy variable with a value of 1 for the years 1991–1994

inclu-sive, 0 otherwise If the rule is effective in enhancing auditor independence, we would

expect a positive coefficient for ROTATION Conversely, following the arguments put

for-ward by those against rotation, we would expect either no association or a negative

asso-ciation between ROTATION and the likelihood of issuing a going-concern opinion.

INFLUENCE A client’s sales (in natural log form) divided by the sum of the sales

(in natural log form) of all the auditor’s clients As in prior research (Chung and Kallapur2003), we use this ratio as a proxy for economic dependence.12 When fees from a clientrepresent a significant proportion of the auditor’s revenues, then the auditor’s power isweakened because of the fear of losing the engagement (DeAngelo 1981; Reynolds andFrancis 2001) This is particularly true when there are going-concern uncertainties (e.g.,

Louwers 1998) We expect a negative coefficient for INFLUENCE.

REPUTATION A dummy variable with a value of 1 if the auditor is a Big 6 firm, 0

otherwise.13 Prior studies indicate that large audit firms have a reputation advantage andmay provide a higher quality audit than small and medium-sized audit firms (DeAngelo1981; Francis and Wilson 1988; DeFond 1992) Reputation concerns are critical for largeaudit firms First, large firm auditors have more clients, and therefore, more quasi-rents tolose if their reputation is diminished Second, large audit firms face greater costs if alle-

gations of audit failure arise Thus, we expect a positive coefficient for REPUTATION.14

PROBFAIL Probability of bankruptcy using Zmijewski’s (1984) financial condition

score.15 Auditors issue going-concern opinions more often when the financial statementsindicate severe financial distress (Reynolds and Francis 2001) We expect companies with

a higher value of PROBFAIL to have a higher likelihood of receiving a going-concern

opinion

LOSS A dummy variable with a value of 1 if the company reported a loss in the

current or previous year, 0 otherwise Auditors are more likely to issue a going-concernmodified opinion when companies report accounting losses (Reynolds and Francis 2001;

DeFond et al 2002) Thus, we expect a positive coefficient for LOSS.

12 Audit fees would be a better proxy for economic dependence Unfortunately, data on fees are not available for the period under investigation Although we acknowledge that sales are a crude proxy for audit fees, we rely

on prior research, which has suggested that client size is an important determinant of audit fees (Simunic 1980).

13 During our research period, the Spanish audit market was dominated by the Big 6 firms: Arthur Andersen, Coopers & Lybrand, Deloitte, Ernst & Young, KPMG, and Price Waterhouse.

14 This variable can be problematic, as it does not capture the differences among big firms Neither does it differentiate among the reputations of small and medium-sized auditing firms To test if our results are affected

by the use of this proxy, we replaced this variable with an alternative measure Assuming that auditor size is highly correlated with audit quality, we used the ratio of sales of the clients of an auditor divided by the sales

of all companies in the audit market as a proxy for reputation (Francis and Wilson 1988) None of the results described later in this paper were sensitive to this alternative specification.

15 To obtain the variable PROBFAIL, we calculated the fitted probability of failure, p, as follows: ˆp ⫽ 1

⫺ F(⫺x ), where F is the standard normal cumulative distribution function, and the fitted values of the index ␤ ˆ were obtained using data on Spanish firms for the vector of variables of Zmijewski (1984), assuming that ˆ

x ␤

the coefficients given in Zmijewski (1984) are valid Although the use of Zmijewski’s model has some limitations (e.g., it was developed under a different time period), we relied on the coefficients provided by this model because: (1) a generally accepted model has not been established for Spanish companies, and (2) it has been widely used in research related to going-concern opinions (e.g., Carcello and Nagy 2004) To test whether our

results were sensitive to the use of this proxy for financial distress, we estimated a model wherein PROBFAIL

was replaced by individual ratios (return on assets, financial leverage, and liquidity) None of the results were sensitive to the alternative specification of the variable.

LEVERAGE The ratio of total debt to total assets A firm’s default status has been

found to be a powerful predictor of a going-concern opinion Chen and Churn (1992) foundthat covenant violations are positively associated with receiving a going-concern opinion

We expect a positive coefficient

SIZE The natural log of the auditee’s total assets (in thousands of Pts) Prior studies

have found that auditors are more likely to issue going-concern opinions to smaller clients(e.g., DeFond et al 2002; Knechel and Vanstraelen 2007) However, Craswell et al (2002)argued that the costs associated with litigation when a large client fails provide an incentivefor auditors to be more conservative in their opinion As with other European countries(see Gaeremynck and Willekens 2003; Knechel and Vanstraelen 2007), the Spanish envi-ronment is characterized by a low risk of litigation (Ruiz-Barbadillo et al 2000) Thus, we

expect a negative coefficient for SIZE.

SPECIALIST A dummy variable with a value of 1 if the auditor is an industry

spe-cialist, 0 otherwise As in prior studies (DeFond 1992; Craswell et al 1995), a threshold

of 10 percent of market share was required in an industry for a specific auditing firm to beclassified an industry specialist.16 Each industry was defined according to the two-digitSpanish SIC codes For each audit firm, we calculated the market share as its clients’ sales

in the industry divided by total sales in the industry Specialization enables auditors to gaincritical knowledge about the business and the industry, and this knowledge is used toevaluate companies’ financial positions (Casterella et al 2004) Biggs et al (1993) showedthat industry specialization is a determinant of an auditor’s ability to detect going-concern

problems Ceteris paribus, a specialized auditor will have a higher likelihood of issuing a

going-concern opinion than will a non-specialized auditor

AGE The company’s age measured in years As younger firms fail more often than

older firms, the stage of organizational development may significantly affect the probability

of receiving a going-concern opinion (DeFond et al 2002; Knechel and Vanstraelen, 2007)

We expect a negative coefficient for AGE.

To formally test our hypotheses, we added two interaction variables to the main effects

model: INFLUENCE*ROTATION and REPUTATION*ROTATION The logistic

regres-sion model with interaction terms is as follows:

⫹ ␤4PROBFAIL⫹ ␤5LOSS⫹ ␤6LEVERAGE ⫹ ␤7SIZE

⫹ ␤8SPECIALIST⫹ ␤9AGE⫹ ␤10INFLUENCE*ROTATION

HROTATION PROPONENTSpredicts that mandatory rotation mitigates the impact of economicdependence on an auditor’s decision to issue a going-concern opinion We test the hypoth-esis by including an interaction term for rotation and economic dependence in the main

effects model (INFLUENCE*ROTATION variable) A positive association between the teraction variable INFLUENCE*ROTATION and the likelihood of issuing a going-concern

in-modified opinion is consistent with HROTATION PROPONENTS, and implies that the rotation policyhelps to alleviate the expected negative impact of auditors’ economic dependence on au-ditors’ decision to issue a qualified report

16 Given the arbitrary nature of the 10 percent rule we used various cutoffs for auditor specialization (more than

15 percent and more than 20 percent) We obtained similar results.

Does Mandatory Audit Firm Rotation Enhance Auditor Independence? Evidence from Spain 121

American Accounting Association

TABLE 1 Variable Definitions and Expected Signs

Expected Sign Main Effects Model

GCO (dependent variable) ⫽ 1 if audit opinion is going-concern

modified, and 0 otherwise.

Binary

ROTATION ⫽ 1 if year is 1991, 1992, 1993, 1994,

and 0 otherwise.

INFLUENCE ⫽ total sales of a client of auditor X

divided by total sales of all clients

of auditor X.

PROBFAIL ⫽ probability of failure calculated

using Zmijewski’s (1984) coefficients.

two years, and 0 otherwise.

(thousands of Pts).

SPECIALIST ⫽ 1 if the market share of the auditor

in the specified industry is ⬎ 10 percent, and 0 otherwise.

Interactions Effects Model

INFLUENCE*ROTATION ⫽ total sales of a client of auditor X

divided by total sales of all clients

of auditor X if the year is 1991,

1992, 1993, 1994, and 0 otherwise.

REPUTATION*ROTATION ⫽ 1 if Big 6 auditor and the year is

1991, 1992, 1993, or 1994, and 0 otherwise.

According to HROTATION OPPONENTS, mandatory rotation moderates the impact of tion concerns on an auditor’s decision to issue a going-concern opinion To test the hy-pothesis, we included an interaction term for rotation and reputation in the main effects

reputa-model The interaction variable REPUTATION*ROTATION captures whether the effect of

reputation concerns on an auditor’s decision to issue a going-concern report is affected

by the mandatory rotation policy A negative association between the interaction variable

REPUTATION*ROTATION and the likelihood of issuing a going-concern modified opinion

is consistent with HROTATION OPPONENTS It implies that the existence of mandatory rotationundermines the potential positive impact of reputation concerns on auditors’ decision toissue a qualified audit opinion

Table 1 describes the variables and their predicted relationship with the likelihood ofissuing a going-concern opinion

TABLE 2 Descriptive Statistics

INFLUENCE⫽ total sales of a client of auditor X divided by total sales of all clients of auditor X;

REPUTATION⫽ 1 if Big 6 auditor, and 0 otherwise;

PROBFAIL⫽ probability of failure calculated using Zmijewski’s (1984) coefficients;

LOSS⫽ 1 if reported loss in either of past two years, and 0 otherwise;

LEVERAGE⫽ total debt divided by total assets;

SIZE⫽ natural log of total assets (in thousands of Pts);

SPECIALIST⫽ 1 if the market share of an auditor in the specified industry is ⬎ 10 percent, and 0 otherwise;

AGE⫽ number of years passed from the startup of the firm.

RESULTS Descriptive Statistics and Univariate Tests

Table 2 presents descriptive statistics about the sample

Pearson correlation coefficients are reported in Table 3 Only six correlations are nificant (p ⬍ 0.05), with 0.231 being the largest correlation (variables SIZE and LOSS).

sig-The correlation matrix suggests that there are no multicollinearity problems in the data.Table 4 reports the breakdown of the sample by the type of audit opinion (going-concernaudit reports versus other audit opinions) Table 4 shows that 90 companies (companies/year) received a going-concern modified opinion and 1,236 companies (companies / year)

did not receive a going-concern opinion The tests indicate that the variables PROBFAIL and LOSS are significantly different for both groups In particular, auditors issued more

going-concern modified opinions to companies that had a higher probability of failure and

had experienced losses in the last two years REPUTATION is also significantly different

for both groups, suggesting that companies audited by a large audit firm were more likely

to receive a qualified audit report

Table 5 shows descriptive data for the sample of firms partitioned into two time periods:the period with mandatory rotation (1991–1994) and the post-mandatory rotation period

(1995–2000) The means for PROBFAIL, LOSS, and LEVERAGE are higher for the rotation

period than for the post-mandatory rotation period, suggesting that the subsample for therotation period includes more financially stressed companies than the subsample for

the post-mandatory rotation period The mean of REPUTATION is also significantly

dif-ferent for the two time periods, indicating that the number of companies audited by largeaudit firms was higher in the post-rotation period compared with the rotation period.Table 6 shows the results of univariate tests comparing the relative frequency of going-concern modified opinions for the mandatory rotation period and the post-mandatory ro-tation period (Panel A, full sample; Panel B, reduced sample including only first-timemodified opinions) The results reveal that auditors issued more qualified audit opinions in

INFLUENCE⫽ total sales of a client of auditor X divided by total sales of all clients of auditor X;

REPUTATION⫽ 1 if Big 6 auditor, and 0 otherwise;

PROBFAIL⫽ probability of failure calculated using Zmijewski’s (1984) coefficients;

LOSS⫽ 1 if reported loss in any of past two years, and 0 otherwise;

LEVERAGE⫽ total debt divided by total assets;

SIZE⫽ natural log of total assets (in thousands of Pts);

SPECIALIST⫽ 1 if the market share of an auditor in the specified industry is ⬎ 10 percent, and 0 otherwise;

AGE⫽ number of years passed from the startup of the firm.

TABLE 4 Univariate Tests for Qualified Audit Reports

Variables

Qualified Going Concern

(nⴝ90) Mean

Not Qualified (nⴝ1,236) Mean

INFLUENCE⫽ total sales of a client of auditor X divided by total sales of all clients of auditor X;

REPUTATION⫽ 1 if Big 6 auditor, and 0 otherwise;

PROBFAIL⫽ probability of failure calculated using Zmijewski’s (1984) coefficients;

LOSS⫽ 1 if reported loss in any of past two years, and 0 otherwise;

LEVERAGE⫽ total debt divided by total assets;

SIZE⫽ natural log of total assets (in thousands of Pts);

SPECIALIST⫽ 1 if the market share of an auditor in the specified industry is ⬎ 10 percent, and 0 otherwise;

AGE⫽ number of years passed from the startup of the firm.

the post-rotation period compared with the rotation period (6.26 percent versus 7.56 percentfor the full sample; 2.39 percent versus 2.90 percent for the reduced sample), even thoughthe companies in the subsample of the rotation period showed higher indicators of financialdistress (see Table 5)

Multivariate Results

Results from the multivariate logistic regressions are presented in Table 7 Panel Aprovides the results for the main effects model (Equation (1)) and for the model withinteractions (Equation (2)) The main effects model is significant (Chi-square⫽ 165.329,

p ⬍ 000) The Hosmer-Lemeshow goodness-of-fit statistic is 7.395 (p ⫽ 0.328), whichprovides evidence of the model’s goodness-of-fit.17 The model with interaction effects isalso significant (Chi-square ⫽ 168.683, p ⬍ 0.000), and the Hosmer-Lemeshow statistic(6.231, p ⫽ 0.375) indicates a good fit

The coefficient of the ROTATION variable in the main effects model (Panel A, Table

7) is negative and significant (␤1 ⫽ ⫺1.125, p ⫽ 0.031), indicating that, ceteris paribus,

17 We use the Hosmer-Lemeshow goodness-of-fit statistic to assess model fit The test computes the estimated probability of receiving a going-concern opinion for each company on the basis of the model, ranks them into ten equal groups of ascending going-concern opinions, and then statistically evaluates the observed and estimated expected frequencies in each tenth under a Chi-square distribution A significant Hosmer-Lemeshow test suggests differences between the observed and expected frequencies in the groups and a lack of model fit (Hosmer and Lemeshow 2000) Conversely, an insignificant statistic (p ⬎ 0.05) indicates no significant differences between

the observed and predicted going-concern opinions suggesting a good model fit.