butterworth and houghton - 1995 - auditor switching - the pricing of audit services

T h e alleged benefit of this auditor change is that the new auditor cuts the price of the initial audit.' That is, after controlling for other factors that affect audit fees, a new audi

Journal Business Finance tAccounting, 22(3) April 1995, 0306 686X

AUDITOR SWITCHING: T H E PRICING OF AUDIT

SERVICES

s BUTTERWORTH AND K.A HOUGHTON

I N T R O D U C T I O N The cost of obtaining audit reports has, in recent years, come under increasing scrutiny A sub-committee of the United States Senate investigated this issue and found evidence of a lack of competition in the provision of audits to major

US corporations (Chung and Lindsay, 1988) Studies in the United Kingdom, Canada and the United States reveal that companies in these countries doubt that they are getting value for the money spent on audits (Firth, 1985) O n e strategy employed by auditees to mitigate these costs is to engage in audit tendering, and the common resultant phenomenon, to switch auditors T h e alleged benefit of this auditor change is that the new auditor cuts the price of the initial audit.' That is, after controlling for other factors that affect audit fees, a new auditor is said to charge less for an audit than an incumbent auditor The presence, or otherwise, of price-cutting has important implications for the independence of auditors and the level and type of competition in the auditing industry However, in Australia (and elsewhere), evidence as to the existence

of price-cutting through auditor switching is largely anecdotal,* and evidence

in the research literature is mixed This paper seeks to more systematically investigate the relationship between auditor switching and audit fees, using the population of companies under the monitoring control of the Western Australian Division of the Australian Securities Commission

T h e next section of the paper reviews the relevant theoretical and empirical literature and develops the hypothesis to be tested The third section contains

a description of the methodology used in the study The fourth section presents the empirical findings and the fifth section summarises the findings and describes the limitations to the study and those findings T h e sixth section describes some

T h e authors are from the University of Melbourne, Australia They gratefully acknowledge the assistance of the Australian Securities Commission, Western Australian Division, the financial suppon of the Coopers and Lybrand Accounting Education Research Fund and the helpful comments

of Jere Francis (University of Iowa), Bill Kinney (University of Texas), Don Stokes (University

of Southern Queensland), Stephen Taylor (University of Sydney), Philip Brown and Gary Monroc (University of W.A.) and Christine Jubb (The University of Melbourne) and participants at workshops at Monash University (Clayton), City University (London), the University of Texas (Austin), the University of Western Australia and the Annual Conference of the Accounting Association of Australia and New Zealand Any remaining errors are the sole responsibility of the authors (Paper received April 1993, revised and accepted July 1993)

Address for correspondence: K.A Houghton, Fitzgerald Professor of Accounting, Faculty of Economics and Commerce, the University of Melbourne, Parkville, Victoria 3052, Australia

@ Basil Blackwell Ltd 1995, 108 Cowley Road, Oxford O X 4 IJF UK

324 BUTTERWORTH AND HOUGHTON

of the policy implications of the results The final section contains conclusions and proposals for future research

THEORY AND LITERATURE REVIEW AND HYPOTHESIS FORMULATION This section consists of three parts, the first of which reviews the theoretical reasons as to why an auditor might cut initial audit fees The second part reviews the empirical evidence as to the existence or otherwise of price-cutting, and the third part formulates the hypothesis to be tested

Theoretical Formulation

In a competitive market for audit services one can theoretically justify the presence of fee-cutting for initial audit^.^ Despite this, the professional audit literature condemns price-cutting as being irrational and harmful to the profession For example, practising accountants have decried the possibility that the practice may lead to sub-standard audits which harm financial report users.' It can also be argued (see the penultimate section) that price-cutting reduces auditor independence, and thereby violates the ethical requirement that audits should be priced so as to maintain auditor impartial it^.^ Given this apparent professional predisposition, one could be excused for believing that price-cutting would be a rare phenomenon However, the anecdotal evidence

in Australia indicates that the practice is becoming increasingly common Contracting theory may explain this behaviour Coase (1937) suggests that

a firm auditee may be viewed as a nexus of contracts The substance of those contracts is that factor-providers (claim-holders) give control of their factors

of production to an entrepreneur, in return for a pay-off from the firm in a future period Those pay-off functions are specified in terms of accounting numbers which are calculated according to the rules set down in the claim- holders' ex-ante contracts with the firm T h e main role of the auditor is to examine the accounting numbers produced by the auditee to determine the degree of correspondence between those accounting numbers, and the numbers that would be produced by applying the contracted-for accounting rules If there is a high degree of correspondence, an auditor will signal this to the claim- holders by issuing a clean audit report, or if the degree of correspondence is low (or if a serious flaw exists), the signal will be in the form of a qualified report

A commonly accepted set of rules for determining accounting numbers may evolve over time (Anderson and Stokes, 1986) However, each auditee would modify those rules to reflect the particular nexus of contracts that constitute that auditee This leads to a demand by different auditees for different audit systems and procedures (Anderson and Stokes, 1986) The incumbent auditor will develop unique systems and procedures that enable it to deliver the required

0 Bad

AUDITOR SWITCHING: THE PRICING OF AUDIT SERVICES 325 service in the most efficient manner possible (Anderson and Stokes, 1986) Therefore, during the initial stage of the audit, an incoming auditor’s systems and procedures may not be perfectly adapted to the task of monitoring the claimholders’ ex-ante contracts Consequently, an auditee is likely to bear auditor-switching costs in familiarising a new auditor with the accounting and information systems of the auditee As a result, an auditee may not costlessly change auditors This means that the possibility exists for an incumbent auditor

to set a fee above the avoidable cost (including opportunity costs) of doing the audit, without the auditee finding it profitable to change to a new auditor That

is, an auditor is able to earn quasi-rents (de Angelo, 1981b).6 As has been argued elsewhere, quasi-rents may lead to the practice by potential auditors

of ‘low-balling’, or the setting of initial audit fees below total audit costs (de Angelo, 1981a), to attract new auditees and the fees (including quasi-rents) that may flow from them

T h e next part of the paper examines evidence of the existence of fee cutting

in the market for audit services

Empirical L iteralure

O n e could argue that the seminal work in this area is Simunic (1980) Using

a regression approach he found no evidence of significant price-cutting in the audits of the 397 US companies in his sample Subsequent related studies (including Chung and Lindsay, 1988; Francis, 1984;7 and Palmrose, 1986a) arrived at similar conclusions

T h e first such study to find price-cutting was Francis and Simon (1987), but it was based on a small sample size (12 changes) That paper motivated

a much more substantial study involving 214 auditees switching auditors, and

a similar number that did not (Simon and Francis, 1988) This later study did find a significant association between auditor switching and price-cutting in the early years of an engagement This study is an important contribution but suffers from certain difficulties arising out of the fact that in the U S audit fee information is non-public and therefore fee data is drawn from responses to questionnaires T h e response rate was 24% with no reported information on any possible non-response bias Arguably, those who responded were more likely to be those who were reporting ‘good news’ (a decline in audit fees) rather than other facts Such a bias, if it exists, would provide a n over-estimate of the general presence of price-cutting O n e would not be surprised therefore

if such a study found the existence of fee-cutting

In more recent studies (not using the methodology adopted by the above cited researchers), Turpen (1990) found that fees for new engagements were,

interalia, lower than fees for predecessor auditors, and Ettredge and Greenburg

(1990) found a mean audit fee reduction of 25% after auditor switching A laboratory simulation by Schatzberg (1990) found evidence of ‘low balling’,

a finding consistent with the presence of price-cutting on initial engagements

0

326 BUTTERWORTH AND HOUGHTON

Table 1

Results of Empirical Studies Investigating Price-Cutting

5 Francis & Simon (1987)

6 Simon & Francis (1988)

7 Schatzberg (1990)

8 Turpen (1990)

9 Ettredge & Greenburg (1990)

Baber, Brooks & Ricks (1987)

Public Sector

10

1 1 Rubin (1988)

USA Can

Aust USA USA USA USA USA USA

USA USA

1980-4

1982

No" No"

No

No Yesb Yes' Yes" Yes Yes

Mixed

Nod

Notes:

a Indirect text of price-cutting only

Potential small sample problem

The fee data from 'change' auditees was collected by questionnaire with only a 24% response rate

" However, indirect evidence of price-cutting was found

In two studies of audit fees in the public sector, Baber, Brooks and Ricks

(1987) and Rubin (1988) found some evidence of audit price-cutting, but due

to sample size or other methodological difficulties, were unable to report the 'change' variable in the regression functions as being significant

In 1988, Simon and Francis described the 'research evidence as inconclusive',

and little has changed since A summary of the findings is provided in Table 1

FOGW ofthe Present Study

The previous sections have suggested that price-cutting is theoretically justifiable, and that its presence can be empirically tested by comparing the fees charged by new and incumbent auditors T o determine whether or not this phenomenon is present in the Australian market, the following research question is tested:

In a given year, are the fees charged by auditors who are in their first year of an engagement, all things being equal, significantly lower than those charged by incumbent auditors?

@

AUDITOR SWITCHING: T H E PRICING OF AUDIT SERVICES 327

If this question is answered in the positive, then evidence of price-cutting will have been found

For the purpose of our study, auditor change (or switch) is defined as a change

in the audit firm specified in the auditee’s annual report between the two years

of our inquiry (1987 and 1988) As with the audit fees themselves this information is publicly available, and there exist non-trivial regulatory penalties for disclosure of false or misleading information ‘Changes’ in auditor due to the merging of audit firms, or audit firm name changes are regarded as ‘no change’ An auditor is regarded as incumbent where no change exists between

1987 and 1988 Note that the reasons behind auditor switching may not be fee-based For example, auditor switching may occur because of auditee takeover and because of changes in company directorates.’ Also an auditor change may occur because of qualitative reasons not associated with price O n e important non-price factor may be the provision and extent of availability of non-audit business services by the audit firm

Recent evidence from the Trade Practices Commission Enquiry into the accounting profession showed that the market for audit services was very competitive and that auditor choice was determined by a number of factors including price and expertise (which may incorporate both audit and non-audit services)

T h e present study differs from the preceding research in one or more of the following three ways First, the study is based on a population of companies, and not a sample (either derived randomly or o t h e r ~ i s e ) , ~ second, the study

is based on information on the public record, and third, the model which has been developed in the present study allows for control of potentially confounding variables beyond those which have so far been considered in the literature These are considered in some detail in the section which follows

METHODOLOGY

T o test for price-cutting it is necessary to control for those factors (in addition

to price-cutting) which affect audit fees The first sub-section below identifies those factors T h e second determines a research design which controls for those factors and which can be used to test the hypothesis of interest and defines the variables to be used in the study The third and fourth sub-sections outline the variable measurements used and describe the company database respectively

Audit Fee Factors

In order to execute the audit task, the auditor must perform both compliance and substantive testing lo Client-specific characteristics will affect the carrying out of these tests and therefore, the audit fee T h e literature reveals that four

0 Basil Blackwell Ltd

328 B U T T E R W O R T H AND H O U G H T O N

client-related factors are important I ' These are:

Client Size: Consistent with the findings of numerous other audit fee

studies," it is expected that as client size increases, the audit fee increases This is because as the auditee increases in size, then all other things being equal, the auditor will have to perform more work to ensure adequate compliance and substantive testing However, the audit fee should increase at a decreasing rate because the larger the auditee, the easier it is for the auditor to achieve economies of scale in its production function (Firth, 1985) Provided that the audit market is competitive, such savings will be passed on to the auditee Alternatively, auditing is a sampling-based procedure As the auditee grows

in size, the sample size required to give a certain degree of control will increase

at a decreasing rate (Simunic, 1980), thereby allowing the auditor to d o proportionately less work (and therefore charge less to the auditee) for each unit increase in auditee size

Client Complexify: Given the results of most other studies,13 it is expected that

as client complexity increases, so too does the audit fee T h e reason for this

is that the more complex the client, the more time the audit manager must spend co-ordinating the audit function, and the greater the difficulties incurred

in performing audit duties This suggests that the more complex the auditee, the more resources must be consumed in order to carry out a n audit of a given quality and therefore, the higher the audit fee This function may also increase

at a decreasing rate For example, Palmrose (1986a) found that the audit fee was significantly and positively related to the natural log of a measure of complexity

Client Risk:'* Another factor that may affect the audit fee is the risk that the auditor might be successfully sued by claim-holders of a n auditee who have suffered a negative pay-out, caused (at least in part) by the auditor negligently concluding that the auditee's accounts were drawn u p in accordance with the contracted-for accounting rules when this was not the case This could impose significant costs upon the auditor To reduce this risk, the auditor may devote more resources to testing, and pass this cost on to the auditee Alternatively, the auditor may demand a premium to bear the risk of auditing the client Given either of these propositions, it is expected that auditors will demand higher fees from clients who expose them to a greater risk of being successfully sued by claimholders (Firth, 1985)

Client Industry: Some prior research has suggested that a n audit fee will be influenced by the auditee's industry.I5 Possibly this is due to differences in audit risk across industry groupings (Simunic, 1980; and Palmrose, 1986a) Alternatively, different industries may have different audit requirements that may require differing amounts of work in the audit and, consequently, attract different audit fees In the context of the present study, a critical industry differentiation exists between companies in the mining and extractive industry and non-mining companies All other things being equal, the less diverse nature

of the asset base and generally qualitatively simpler nature of those companies

0

AUDITOR SWITCHING: T H E PRICING OF AUDIT SERVICES 329 (at least in Western Australia) leads to the expectation that mininglextractive industry companies will have lower fees than other companies

T o test for price-cutting in the first year of an audit, four client-specific factors are controlled for: size, complexity, risk and industry It is also necessary to control for factors relating to the auditor

Auditor Size: Empirical studies suggest that auditor size is a n important

determinant of the cost of a n audit.I6 A number of reasons have been suggested for this Firstly, it is aruged that large auditors can achieve economies

of scale in their production function, and provided that the audit market is sufficiently competitive, those savings will be passed on to the auditee (Firth, 1985) Secondly, it is argued that large auditors will deliver a higher quality audit (de Angelo, 1981b) Consequently, one would expect larger firms to charge higher fees, even if the market for audits is competitive (Rosen, 1974) Alternatively, it is possible that large auditors may form a cartel so as to charge

a monopoly price to those firms who wish to obtain higher quality audits (Firth, 1985) Despite some contradictory results in previous Australian ~ t u d i e s , ’ ~ the expectation is that Big 8 auditors will have higher fees than Non-Big 8 auditors

Non-Audit Services Provided by the Auditor: Simunic (1984), Palmrose (1986b)

and Turpen (1990) found an association between audit fees charged to auditees and the fees for non-audit services (“AS’) which they purchased from their auditor There are two reasons why the audit fee charged may be affected by the provision of NAS Firstly, audits may serve as a ‘loss-leader’ for firms providing both audits and NAS (Hillson and Kennelley, 1988), with auditors recouping audit losses through an upwards adjustment of the NAS fee As a result, it is possible that audit fees are a decreasing function of NAS fees (Simon and Francis, 1988) Secondly, there may be a link between NAS and the audit fee caused by the presence of spillovers of knowledge from the NAS area to auditing and vice versa (Simunic, 1984) According to Simunic (1984), the effect of these knowledge spillovers on the audit fee depends on whether a spillover from NAS to auditing reduces the start-up cost of a n audit or the marginal cost of auditing, the price elasticity of demand for auditing and the effect on NAS costs of a spillover from auditing to NAS

Research Design

To control for the effect of the factors outlined above and to test for price-cutting,

a multiple linear regression model of audit fees is used To avoid the econometric problems that could be caused by pooling observations across time, cross- sectional data is used.” Given the limitations inherent in such a research design, the audit fee model takes the following form:

FEE, = b o + bICHANGE, + b,AUD, + b,,NAS, + b$IZE,

+ bSCOMP, + b6RISK, + b71ND, + p

0

330 BUTTERWORTH AND HOUGHTON

Where, FEE; = Audit Fee charged to auditee i ,

= Hypothesis Variable, first year of engagement,

= Size of auditee i’s auditor (Big 8 or Non-Big 8),

= Amount of NAS provided to auditee i by its auditor,

The measurement of each of these variables is outlined below

Audit Fee: The audit fee charged to auditees is measured by the dollar value

of audit fees paid by the firm to its auditor This dollar value is required to

be disclosed by law.”

Auditor Change: T o test the hypothesis that new auditors charge less than

incumbent auditors in a given year, a dummy variable, coded 1 if the audit

of a firm is the initial engagement for the audit (and zero otherwise), is included

in the regression model If the hypothesis is true, then the regression co-efficient

on this dummy variable should be statistically significant and negative

Auditor Size: As with other studies in this field, auditor size is measured using

a dummy variable coded 1 if the auditor is a member of the first tier of auditors2’ (the Big 8, or, more recently, the Big 6); otherwise coded zero

Provision of Non-Audit Services: In this study, the amount of non-audit services

provided by an auditor to an auditee is measured by the dollar amount of NAS purchased by the auditee from its auditor As with the audit fee, this amount must by law be disclosed by the auditee This continuous variable captures the possibility that the greater the amount of non-audit services provided by the auditor, the greater will be the potential spillover effects or the greater will

be the potential for audit fee/non-audit services fee trade-offs, and the greater

Client Size: Consistent with other researchers, (see for example, Francis and

Simon, 1987; and Gerrard et al., 1986) the size of an auditee is measured using the book value of the auditee’s total assets

Client Complexty: Most researchers in this field have measured the complexity

of an auditee by using a continuous measure of its legal/organisational dispersion such as the number of consolidated subsidiaries in the auditee’s group of companies The rationale behind these measures is that the greater the dispersion that is shown by a firm’s structure, the more intricate and sophisticated become the legal and operational forms adopted by the firm (Taylor and Baker, 1981) Consistent with this, the present study uses the number of companies in the auditee’s group (holding company plus subsidiaries)

as the measure for complexity.21

will be the effect on the audit price

0 Basil Blackwell

AUDITOR SWITCHING: T H E PRICING OF AUDIT SERVICES 33 1

Client Risk: Following the practice of previous studies," a dummy variable

is used to control for risk It is coded 1 if a n audit qualification is given by the auditor.23 T h e rationale behind this measure is that some form of audit qualification may signal the possibility of 'significant uncertainties which may result in future losses to the auditee' (Simunic, 1980), consequently exposing the auditor to a higher risk of being successfully sued as a result of the audit

It has the further advantage that, in contrast to the risk measures used in some other studies, there appears to be no a priori reason for believing that it is a

proxy for the size or complexity of the a ~ d i t e e ' ~ T w o such measures of risk are used in this study: audit qualification in the year the auditor is changed

in this study (1988), and the year preceding (1987) The first measure represents

a contemporaneous measure of risk, the second a lagged measure Note that these variables have the disadvantage of not capturing the effect of gradations

in auditee risk on the audit fee.25

Client Industry: The companies in the sample were divided into their industry

groups as classified by the Australian Stock Exchange Industry influences were then controlled for by using a dummy variable coded 1 where a company was involved in the extractive industry and 0 if not

Company Selection and Data

T h e data used in this study came from a computerised data-base of the Annual

Reports of 433 Western Australian-headquartered companies maintained by the (then) Department of Corporate Affairs in Perth, Western Australia, now part of the Australian Securities Commission Unlike other studies, this represents the population of all publicly listed companies in the region.26

RESULTS This section consists of four parts T h e first considers the results of the data selection process T h e second part provides results of univariate tests on the sample data, and the third shows the results of the regression analysis T h e final part carries out further tests on the data

1 Total companies on file;

2 Active (i.e trading) Companies (neither failed in 1988

3 T h e company had filed its 1988 annual report and it

0

332 BUTTERWORTH AND H O U G H T O N

had been processed by the regulators at the time of

4 Information on the company’s subsidiaries1

5 The company’s auditors and audit fee could be

Of the 433 companies on the database, 268 fulfilled the selection criteria Of these, 37 (13.8%) changed auditors between 1987 and 1988 Changes to Big

8 firms totalled 29 (including ten changing from one Big 8 firm to another, and nineteen changes from Non-Big 8) There were 8 which changed to Non-

Big 8 firms (including one change within this sector, and seven changes from Big 8 firms) As noted above, these changes are exclusive of changes caused

by auditor name changes and mergers Table 2 shows the distribution of companies in the sample across industry groups and across the Main and Second boards of the Australian Stock Exchange in Perth

Table 2

Summary of Sample Coverage by Industry and Board

Panel A: Industry Coverage

Panel B: Board Classification

AUDITOR SWITCHING: T H E PRICING OF AUDIT SERVICES 333

Mean Std Deu Mean Std Deu M e a n Std Deu

Per Cent of Auditees -

Descriptive Statistics

Descriptive statistics for the price model variables are provided in Table 3

Although some differences in means were apparent, no statistically significant differences for the control variables were observed between the change and non- change groups.” This is consistent with the findings of other studies in this area

Regression Analysis

T h e basic version of the regression model is tested by modelling untransformed audit fees against the untransformed independent variables Residuals from this model are found to violate the OLS regression assumptions As with previous studies of audit pricing using Australian data, (Francis, 1984; and Gerrard et al., 1986) a loglo transformation of the dependent and continuous independent variables utilised in the study is used to address this problem The results for the model using the transformed data are provided in Column

A of Table 4 This table shows that the model is significant and explains over 70% of the variance in the transformed audit fees Most of the control variables are significant (at the 5 % level or better), and, where relevant, have the predicted signs Further, the regression assumptions of normally distributed