Sie sind auf Seite 1von 24

Accounting and Finance 47 (2007) 423446

Note on audit fee premiums to client size and


industry specialization

E.
Carson
and
N.Finance
Fargher/Accounting
Blackwell
Oxford,
Accounting
ACFI

0810-5391
Journal
ORIGINAL
XXX
The
Authors
compilation
UK
Publishing
and
ARTICLES
Ltd
2007 AFAANZ and Finance 47 (2007) xxxxxx

Elizabeth Carsona, Neil Fargherb


b

a
School of Accounting, University of New South Wales, Sydney, 2052, Australia
Department of Accounting & Finance, Macquarie University, Sydney, 2109, Australia

Abstract
This research note examines the impact of client size on the estimation of audit fee
premiums in the Australian market for audit services. Previous research suggests
that higher audit fees are expected for both larger clients and for industry
specialization. We find that in the Australian market for audit services, the fee
premium attributed to industry specialist audit firms is concentrated in the audit
fees paid by the largest clients in each industry. One reason for higher fees paid
by larger clients is the demand for additional audit services. We find higher fees
for companies cross-listed on US exchanges. We also find that fee premiums to
auditors that are city-industry leaders are strongly related to client size.
Key words: Audit fees; Auditing; Industry specialization
JEL classification: M42
doi: 10.1111/j.1467-629x.2007.00213.x

1. Introduction
This research note examines the association between client size and audit fee
premiums. This topic is important in helping to understand the extent to which fee
premiums attributable to auditor characteristics, such as industry specialization,
We gratefully acknowledge the use of data provided by the Centre for Audit and Assurance
Research at the University of New South Wales, the Securities Industry Research Centre of
Asia-Pacific (SIRCA) from Aspect Financial, the Faculty of Economics and Business,
University of Sydney, and Allen Craswell. We appreciate the helpful suggestions from
Asher Curtis, Chris Hogan, Rajib Doogar, Brian Mayhew, Renee Radich, Peter Roebuck,
Roger Simnett, Mike Stein, Stephen Taylor, Stuart Turley, Arnie Wright, Mike Wilkins,
participants at the American Accounting Association (AAA) Auditing Section meeting 2004,
the International Symposium on Audit Research 2004, the AAA Annual meeting 2004, and
at workshops at the University of Connecticut, Boston College and Northeastern University.
Received 21 December 2005; accepted 29 September 2006 by Gary Monroe (Deputy Editor).

The Authors

Journal compilation 2007 AFAANZ

424

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

can be distinguished from fee premiums attributable to client characteristics,


such as client size.
Prior studies of industry specialization have observed that the results for
tests of industry specialization are sensitive to test specification with respect to
client size. Craswell et al. (1995) report that the industry specialist premium
only occurs for auditees in the top half of the sample based on auditee total
assets. Ferguson and Stokes (2002) and Ferguson et al. (2003) report that the
premium for industry specialists increases as client size increases. As client
size can proxy for client characteristics influencing the scope and complexity of
an audit, the present study examines client size as a determinant of audit fee
premiums. Specifically, we investigate whether specialist auditors obtain a fee
premium for clients of all sizes. As large clients pay higher fees, and auditors
of larger clients are designated as the industry specialist, the research question
of interest is whether specialist auditors earn fee premiums for audits of smalland medium-size clients.
We find that the national industry specialist premium documented by Ferguson
et al. (2003) is found only in the highest quintile of client size. The fee premium
to industry-specialist auditors is not significantly different from zero for the
lower and middle quintiles of clients based on total assets. The fee premium to
city-industry lead auditors (Ferguson et al., 2003) is also strongly related to
client size.
The present study contributes to the prior literature by documenting the sensitivity
of premiums to industry specialization with respect to client size. Although this
sensitivity is noted in the seminal research, it is frequently not cited by subsequent
papers assuming a positive association between specialization and fees. In
examining this sensitivity, we also document three potentially omitted variables
in fee models using Australian data: cross-listing on a US exchange, location of
the client head office and client bargaining power. Of particular importance is
the need to control for larger client demand for assurance services related to US
cross-listing when estimating fee models using Australian data.
2. Previous research
The general audit fee model common to most of the studies in the literature
represents audit fees as a function of client size, client complexity, client and
auditor risk, and audit quality (e.g. Craswell and Francis, 1999). Early research
in this paradigm (e.g. Simunic, 1980; Francis, 1984; Francis and Stokes, 1986;
Palmrose, 1986; Francis and Simon, 1987) makes it clear that client size is a
significant issue with respect to the overall fee model being estimated. Specifically, the fee premium to auditors is expected to be different depending on the
market segment in which the auditor is competing, where market segment is
based on client size.
Recent studies that have considered the role of client size include Reynolds
and Francis (2001), who find no evidence that Big Five auditors report more
The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

425

favourably for larger clients, and Chung and Kallapur (2003), who find no
support for the contention that important clients have higher abnormal accruals.
Sullivan (2000) finds that following the Big Eight to Big Six mergers there is
evidence of reduced marginal costs of auditing large clients as measured by
patterns of client switches following the mergers. DeFond et al. (2000) find that
the fee premium to Big Six industry specialist is robust across the spectrum of
company size in the Hong Kong market.
Prior studies of industry specialization using Australian data have observed
that the results for tests of industry specialization are sensitive to test specification with respect to client size. Craswell et al. (1995) report that the industry
specialist premium only occurs for larger auditees, where large is defined as
being in the top half of the sample based on auditee total assets . Craswell et al.
(1995) argue that larger-sized companies in general have greater agency problems and, hence, are more likely to benefit from the additional audit quality
of a Big Eight industry specialist. Ferguson et al. (2003) also report that the
industry specialist premium increases as client size increases. Recent research
also shows evidence consistent with demand for industry specialist auditors
from Australian companies with characteristics that are likely to be associated
with client size. Chen et al. (2005) find that companies with an audit committee
and companies with a higher proportion of non-executive directors are more
likely to be audited by an industry specialist.
Ferguson et al. (2003, 2006) provide evidence that national rankings are
driven by city-industry leaders. In a small market like Australia with few large
cities, there is the distinct potential that conditioning on industry and city is
identifying the largest, more complex audit clients. Using data from the larger
US market, Francis et al. (2005) find evidence of audit fee premiums being
associated with joint national and city specialist auditors. Francis et al. (2005)
argue that the results are generally consistent for the upper and lower halves of
their sample by client size; however, they also note that the estimated premium
for joint nationalcity leadership is 22 per cent for larger clients and only 7 per
cent for smaller clients (Francis et al., 2005). Therefore, in our study, we investigate the concentration of the observed fee premiums with respect to client size
for both national-level and city-level industry specialists.
3. Client size, industry specialization and fee premiums
The client size and auditor specialization explanations for audit fees both
predict a fee premium for the industry specialist on the largest audits in the
industry. However, the difference is with respect to whether fee premiums are
earned by the specialist auditor for audits other than the largest clients that
dominate the industry. If the audit fee premium is attached to the industry
specialist auditor, then we would expect the premium to attach to all audits
undertaken by that specialist auditor in that industry. If these premiums are
attached only to audits of the largest clients in an industry, then we would not
The Authors

Journal compilation 2007 AFAANZ

426

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

expect to observe significant fee premiums for audits beyond the premiums
earned for audits of the largest clients.
To the extent that the largest clients endogenously choose the auditor best
suited to their business, client size and industry specialist are inherently related
and cannot be separated. However, empirical evidence can provide a basis for
understanding the extent of specialist audit fee premiums across clients of
varying size and the ability of the specialist to earn a fee premium beyond the
premium earned from the largest clients.
There is evidence that differences in client size are related to audit fees.
Mayhew and Wilkins (2003) find that audit firms that possess significantly
higher market share than their competitors earn higher audit fees. Audit firms
with higher market share also typically have a higher proportion of the larger
clients. In contrast, Casterella et al. (2004) examine client bargaining power
where power is measured as each clients sales relative to the sales to all
clients in the industry audited by the companys auditor. They find that audit
fees are lower as the client size increases relative to their auditors industry
clientele.
Consistent with prior studies, we do not have sufficient data to directly examine the quantity or nature of assurance services demanded by specific clients.
However, we can attempt to carefully model the relation between the client
attributes and audit fees, and then consider the incremental fees attributable to
industry specialization for clients of differing size. Overall, in considering the
relation between client size and industry specialization, there are at least three
aspects of client size that can be considered: absolute client size, client size
within industry and client size relative to other clients.
With respect to absolute client size, we consider two aspects of the association between client size and audit fees. The audit fee model is generally well
specified. However, it is known from prior research that the coefficient on log
of assets varies with client size (Bell et al., 1994). Therefore, we consider nonlinearity in the relation between audit fees and client size. Palmrose (1986)
observes that if larger clients require more audit services than smaller clients,
then we would expect that these large clients pay relatively higher fees per dollar of size relative to smaller clients. To examine possible sources of increased
demand for audit services from larger clients, we identified companies crosslisted on a US exchange as a client characteristic correlated with client size and
likely to increase the scope and complexity of audit procedures required by
large clients.1
With respect to size within industry we focus on the premium to industry
specialization to clients of varying size. We empirically examine the extent to
1
From an econometric perspective, cross-listing could be considered an omitted variable
from a typical audit fee model, but only to the extent that size and other variables do not
fully capture the variation in fees attributable to the need to provide assurance on the
reconciliation to US GAAP and other issues relating to dual listing status.

The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

427

which the audit fee premium attached to the industry specialist auditor is
earned across clients of varying size.
Casterella et al. (2004) argue that relative client size within industry results
in clients having greater bargaining power and that, ceteris paribus, this will
result in lower audit fees. In contrast, Mayhew and Wilkins (2003) argue that the
client can only bargain for a lower fee where the auditor has not successfully
differentiated itself from competitors. They find evidence that audit firms that
have successfully differentiated themselves earn a fee premium. This evidence
is consistent with the successfully differentiated audit firm retaining a stronger
bargaining position with their clients. To the extent that larger clients have
larger reputational capital at risk, and tend to select a differentiated leading
auditor in the industry, we would expect higher fee premiums (Mayhew and
Wilkins, 2003). We include a measure for client bargaining power as measured
by Casterella et al. (2004); however, based on the prior research using US data
the premium predicted is ambiguous.
We specifically consider the following aspects of client size in audit fee
models: (i) additional variables to capture non-linearity in the client size to fee
relation; (ii) an additional variable to capture variation in fees arising from
large client demand for additional audit services associated with size (crosslisting on a US stock exchange); (iii) a variable to control for client bargaining
power (POWER); and (iv) interaction variables to capture the variation in fee
premium to industry specialization across clients of varying size.
4. Sample
4.1. Sample selection
For comparability to Ferguson et al. (2003), the sample comprises audit
engagements in Australia for the fiscal year 1998. It should be noted that the Price
waterhouse Coopers and Lybrand merger occurred in late 1997 and potentially
impacts the industry reputation and pricing during this period. However, to address
the issue of the sensitivity of the results to the year selected, we have also estimated
the models for 1999 and 2004, and reported the results as sensitivity analysis.
The 1998 and 1999 samples are drawn from the Who Audits Australia database that contains audit fees and non-audit services fees data for the population
of Australian listed companies. The 2004 sample is drawn from the University
of New South Wales audit fee database. The audit fee data were then matched
to the Aspect database of financial statement information. The data were
reviewed for reasonableness and identified errors corrected by reference to
annual reports. Only observations with all available data, matching total assets
on both databases, and two prior years of financial statement information were
included. We further excluded companies with financial periods of other than
12 months (including one large company WMC), companies reporting in a
foreign currency and audits where the audit office is outside Australia. All
The Authors

Journal compilation 2007 AFAANZ

428

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

monetary units are in Australian dollars. For the city-level analysis, audit fees
are examined for companies headquartered in the five largest Australian cities:
Adelaide, Brisbane, Melbourne, Perth and Sydney.
We refer to the Big Six auditors because of the use of fiscal year end 1998
data in which the firm Coopers and Lybrand was noted to be still issuing audit
opinions. We have not redefined five Coopers and Lybrand audits as Price waterhouse Coopers because of the problematic issue of treating a lower market share
auditor as an industry leader by arbitrary reclassification.2
Between 1999 and 2004 the primary industry code used by the Australian
Stock Exchange (ASX) was redefined. This creates an interesting problem for
researchers. We have used the new Global Industry Classification Standard
(GICS) industry codes for 2004 as any reputation effect based on industry
would arguably be better measured by the industry codes actually in use. For
2004, industry is measured as the four digit GICS code. Two industries with
fewer than 10 companies audited by Big Four auditors were included in larger
industry groupings at the three digit level (GICS 3030 is included with 3020,
and GICS 4530 is included with 4520). This resulted in 18 distinct non-financial
industry groupings. Despite the lack of direct comparability between the 1998
and 2004 industry groupings, the results for 2004 are generally consistent with
those for 1998.
5. Empirical model
We use as a basis for our analysis the audit fee model from Ferguson et al. (2003).
We then extend the basic model to capture variations in the association between
client characteristics and fee premium. We specifically consider the following
aspects of client size: (i) allow the coefficient on client size to vary by quintile of
total assets; (ii) include an additional variable to capture variation in fees arising
from larger clients cross-listing on US markets; (iii) a variable to control for
client bargaining power (POWER); and (iv) interaction variables to capture the
variation in fee premium to industry specialization across clients of varying size.
5.1. Industry specialist auditors
For comparability to prior research, we follow the methodology of Ferguson
et al. (2003) in designating the first and second ranked auditors, based on the
proportion of industry audit fees, as the specialist auditors (SPECIALIST). We
extend the analysis to consider alternate measures of specialization in auditor
reputation, such as the auditor that is the leader in both the industry nationally
and the city office (Ferguson et al., 2003; Francis et al., 2005). The city-industry

Ferguson et al. (2003) refer to Big Five audits and treatment of these observations is
not reported.

The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

429

leader indicator variable takes the value of 1 when the auditor is the number
one auditor, based on proportion of audit fees, in the city and 1 of the top 2
firms nationally (CITYIL).
5.2. Non-linearity in the client size to audit fee relation
Prior research indicates that the coefficient on log of assets varies with client
size (Bell et al., 1994). We model this aspect by inclusion of indicator variables for
each quintile by total assets to partially capture the effect of client size on audit
fees. We interact an indicator variable for the quintile of total assets with the log of
total assets. This allows the coefficient on log of total assets to vary with client size.3
5.3. Large client demand for assurance services: US cross-listing
To examine possible sources of increased demand for audit services from larger
clients, we identified companies required to complete reconciliations between
US Generally Accepted Accounting Principles (GAAP) and Australian GAAP. To
examine this possible source of relatively higher fees for large clients, we include
an indicator variable (ADR) for companies with Level II or Level III American
Depository Receipts (ADR). An ADR is a dollar-denominated negotiable
certificate that represents ownership of shares in a non-US company. ADRs
were identified from the JPMorgan database with an effective date prior to the
end of the 1999 financial year. Level II and Level III ADRs are required to complete reconciliations between US GAAP and Australian GAAP, and file audited
financial statements with the US Securities and Exchange Commission. For the
1998 sample, 13 of the 24 largest clients in the 24 industry categories have ADRs.
5.4. Specialist premiums earned on clients of differing size
Finally, we partition the specialist premium by clients of different size. We
report the results using quintiles of total assets. We calculate the quintile of firm
size within the sample of Big Six clients and interact an indicator variable for
the largest quintile, medium quintiles and small quintile with the indicator
variable for audit by a specialist auditor. Similar results are obtained using
deciles with slightly stronger reductions in the specialist premium when deciles
are used. When size within industry is used, the empirical relations suggest that
the association is stronger between fee premiums and size within industry than
between absolute size and the fee premium.

3
We also allowed both intercept and slope to vary with the quintile of total assets. The
results are generally consistent with those reported, although the combinations of intercept
and slope for each quintile varied less systematically. The inferences for the industry
specialist premium are not altered.

The Authors

Journal compilation 2007 AFAANZ

430

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

5.5. Relative client size


Relative client size is measured following Casterella et al. (2004). POWER is
measured for each company as the natural log of company sales divided by the
sum of logged sales for all firms in the industry audited by the companys
auditor. Because of missing sales data available for 2004 from Aspect, total
revenue is used for sales in 2004.
5.6. Client size and measures of office level industry expertise
Ferguson et al. (2003) find that the premium to industry expertise in Australia
is driven by industry leadership at a city level. In the city-level analysis, we
include an additional control for variation in city-level costs. As audit practices
in major cities (Sydney and Melbourne) are more likely to have higher overheads than practices in smaller cities (Adelaide and Perth), it is necessary to
control for the higher cost of audit production in these cities. As the major audit
cost driver is labour, we use the annual salary cost for an experienced auditor
in each of the major cities from Hays Personnel Services for the relevant year.
Higher salaries are reported for Sydney and Melbourne relative to the other
capital cities. Audits outside the major capital cities are allocated the salary
cost for Adelaide. Although we recognize that an audit production cost would
not be expected to enter linearly into a production cost function, we use the
linear approximation as a parsimonious model allowing comparison to prior
research.
5.7. Model
The resulting ordinary least squares regression model for examining the
impact of size not conditioned on industry is specified as follows (ignoring the
company subscript):
LAF = + q=1,5 1,q(q * LTA) + 2 LSUB + 3 CATA + 4 QUICK
+ 5 DE + 6ROI + 7FOREIGN + 8OPINION + 9 YE
+ 10 LOSS + 11ADR + 12POWER + 13 SPECIALIST*LARGE
+ 14 SPECIALIST*MEDIUM + 15 SPECIALIST*SMALL + ,

(1)

where LAF is the natural log of total audit fees paid to the auditor ($A thousands);
q is an indicator variable for each quintile of client size to allow the slope to
vary with client size; LTA is the natural log of total assets ($A millions);
LSUB is the natural log of the number of subsidiaries; CATA is the ratio of
current assets to total assets; QUICK is the ratio of current assets less inventories
to current liabilities; DE is the ratio of long-term debt to total assets; ROI
is the ratio of earnings before interest and tax to total assets; FOREIGN is the
proportion of subsidiaries that represent foreign operations; OPINION is an
The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

431

indicator variable with a value of 1 for a qualified or modified opinion; YE is


an indicator variable with a value of 1 for non-June fiscal year end; LOSS is an
indicator variable with a value of 1 for a loss in any of the last 3 years; ADR is
an indicator variable taking the value 1 where the client is cross-listed on a
US exchange as a Level II or Level III ADR; POWER is a measure of client
bargaining power relative to the auditors total clientele in the industry as
measured by the natural log of company sales, divided by the sum of logged
sales for all firms in the industry audited by the companys auditor; SPECIALIST
is an indicator variable with a value of 1 for the first and second ranked
auditor in the industry based on proportion of industry audit fees; SPECIALIST
*LARGE is an indicator variable for the two specialist auditors in the industry
interacted with a value of 1 if the client is in the largest quintile of clients
based on total assets and 0 otherwise; SPECIALIST*MEDIUM is an indicator
variable for the two specialist auditors in the industry interacted with a value of
1 if the client is in quintile 2, 3 or 4 of clients based on total assets and 0
otherwise; SPECIALIST*SMALL is an indicator variable for the two specialist
auditors in the industry interacted with a value of 1 if the client is in the
smallest quintile of clients based on total assets and 0 otherwise; and = an
error term.
5.8. Descriptive statistics
Table 1 reports descriptive statistics for the sample of firms with Big Six
auditors. As in Craswell et al. (1995) and Ferguson et al. (2003), to avoid the
confounding effect of brand name on audit fees, the regression models are
estimated only for the firms having Big Six auditors in fiscal year 1998. We
also exclude the financial services and property development industries
(ASX industry codes 16, 17, 19 and 20; GICS codes starting with 40)
because of their dissimilar nature relative to other industries (Fields et al.,
2004).
As is common in studies using Australian data, there is a broad variation in
client size. Consistent with previous research the distributions of the client size,
subsidiaries and audit fee variables are highly skewed. To reduce the impact of
outliers on the residuals, the natural log of size (total assets), log of subsidiaries
and the natural log of audit fees are used in the subsequent analysis. To reduce
the impact of outliers on the distribution of the regression residuals, the
QUICK, DE and ROI variables are winsorized (consistent with Ferguson et al.,
2003) at the 5 per cent level.
Table 2 reports the Spearman correlations between audit fees, client size,
auditor and client characteristics. Indicator variables used for industry specialist
(SPECIALIST) and city-industry specialist (CITYIL) are all significantly positively
correlated with both audit fees and client size. Of interest to this study is
whether industry specialist auditors still earn higher fees after controlling for
any higher fees earned on the largest audits.
The Authors

Journal compilation 2007 AFAANZ

432

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Table 1
Descriptive statistics for sample of Big Six audits of non-financial Australian companies
Sample

Operating companies with Big Six Auditors


1998
N = 558

1999
N = 543

2004
N = 611

Variable

Mean

SD

Mean

Mean

Audit fees ($A thousand)


Log of audit fees (LAF)
Total assets ($A million)
Log of total assets (LTA)
Number of subsidiaries
Log of subsidiaries (LSUB)
CATA
QUICK
DE
ROI
FOREIGN
CITYCOST
POWER
Categorical variables
OPINION
YE
LOSS
ADR
NATIONAL #1 or #2 (SPECIALIST)
CITY #1 AUDITOR AND
INDUSTRY #1 or 2 AUDITOR (CITYIL)

241.21
4.13
658.85
10.80
20.28
0.51
0.38
4.11
0.16
0.11
0.21
46.47
0.17

736.40
1.46
3359.70
2.25
53.27
4.54
0.25
8.41
0.16
0.32
0.27
4.53
0.18

273.04
4.24
708.68
10.93
21.75
0.52
0.40
3.41
0.16
0.07
0.20
46.54
0.12

318.182
4.58
773.09
10.91
21.86
0.75
0.45
4.44
0.13
0.09
0.23
60.58
0.11

12%
23%
64%
9%
57%
29%

13%
22%
61%
7%
61%
31%

13%
23%
66%
9%
57%
28%

Sample excludes banks, financial and property segments (industry codes 16, 17, 19, and 20 for
1998 and 1999; GICS code 40 for 2004). LAF is log of total audit fees paid to the auditor ($A
thousands); LTA is log of total assets ($A thousands); LSUB is the log of the number of subsidiaries;
CATA is the ratio of current assets to total assets; QUICK is the ratio of current assets less inventories
to total assets; DE is the ratio of long-term debt to total assets; ROI is the ratio of earnings before
interest and tax to total assets; FOREIGN is the proportion of subsidiaries that represent foreign
operations; OPINION is an indicator variable with a value of 1 for a modified orqualified opinion; YE
is an indicator variable with a value of 1 for non-June fiscal year end; LOSS is an indicator variable
with a value of 1 for a loss in any of the last 3 years; CITYCOST is the use the annual salary cost for
an experienced auditor in each of the major cities from Hays Personnel Services; ADR is an indicator
variable taking the value 1 if the firm is cross-listed as an American Depository Receipt; POWER is a
measure of client bargaining power relative to the auditors total clientele in the industry as measured
by the natural log of company sales, divided by the sum of logged sales for all firms in the industry
audited by the companys auditor; SPECIALIST is an indicator variable with 1 if the client is audited by
the specialist auditor in the industry, where specialist auditor is identified as the two auditors with the
largest market share based on the highest proportion of industry audit fees; and CITYIL is an indicator
variable with 1 if the company is audited by the number one auditor in the industry and the city, and
number one or two in the industry nationwide based on total audit fees. SD, standard deviation.

The Authors

Journal compilation 2007 AFAANZ

Variable

SPECIALIST

LAF

LTA

LSUB

CATA

QUICK

DE

ROI

FOREIGN

OPIN

LOSS

ADR

POWER

CITYCOST

SPECIALIST
LAF
LTA
LSUB
CATA
QUICK
DE
ROI
FOREIGN
OPINION
LOSS
ADR
POWER
CITYCOST
CITYIL

1.00
0.13
0.12
0.14
0.02
0.01
0.06
0.07
0.07
0.11
0.09
0.05
0.09
0.05
0.55

1.00
0.86
0.80
0.10
0.34
0.59
0.51
0.42
0.19
0.50
0.21
0.52
0.33
0.25

1.00
0.69
0.04
0.29
0.62
0.57
0.29
0.26
0.52
0.22
0.53
0.25
0.24

1.00
0.08
0.26
0.48
0.33
0.53
0.11
0.41
0.22
0.41
0.19
0.22

1.00
0.25
0.13
0.15
0.06
0.17
0.09
0.08
0.13
0.08
0.06

1.00
0.30
0.21
0.07
0.11
0.21
0.02
0.25
0.08
0.05

1.00
0.37
0.16
0.18
0.35
0.12
0.39
0.12
0.14

1.00
0.09
0.35
0.60
0.02
0.43
0.17
0.15

1.00
0.03
0.17
0.15
0.13
0.06
0.07

1.00
0.26
0.01
0.15
0.09
0.13

1.00
0.05
0.40
0.07
0.16

1.00
0.07
0.02
0.17

1.00
0.20
0.04

1.00
0.11

All correlations greater than 0.10 are significant at the 0.01 level. LAF is log of total audit fees paid to the auditor ($A thousands); LTA is log of total assets
($A thousands); LSUB is the log of the number of subsidiaries; CATA is the ratio of current assets to total assets; QUICK is the ratio of current assets less inventories
to total assets; DE is the ratio of long-term debt to total assets; ROI is the ratio of earnings before interest and tax to total assets; FOREIGN is the proportion
of subsidiaries that represent foreign operations; OPINION is an indicator variable with a value of 1 for a modified or qualified opinion; LOSS is an indicator
variable with a value of 1 for a loss in any of the last 3 years; ADR is an indicator variable taking the value 1 if the firm is cross-listed as an American Depository
Receipt; POWER is a measure of client bargaining power relative to the auditors total clientele in the industry as measured by the natural log of company sales,
divided by the sum of logged sales for all firms in the industry audited by the companys auditor; CITYCOST is the use the annual salary cost for an experienced
auditor in each of the major cities from Hays Personnel Services; and CITYIL is an indicator variable with 1 if the company is audited by the number one auditor
in the industry and the city, and number one or two in the industry nationwide based upon total audit fees.

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Journal compilation 2007 AFAANZ

The Authors

Table 2
Non-parametric correlations between variables

433

434

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

6. Results
6.1. Client size and audit fees
We initially document specification issues regarding the association between
client size and audit fees in Australian audit fee data. Table 3 documents the
influence of scale on the Studentized residuals for a typical fee model. We
estimate typical fee models, excluding the industry specialist variable, for each
quintile of client size and calculate the Studentized residual for each observation. As the Studentized residual is expected to be normally distributed (Easton
and Sommers, 2003), we count the number of observations with the absolute
value of Studentized residuals greater than 1.96 and compare the frequency of
Studentized residuals across size quintiles.
In Panel A of Table 3, the median absolute value of the Studentized residual
ranges from 0.538 for quintile 3 of client size to 0.832 for quintile 1. The mean
and median absolute value for the largest client quintile, quintile 1, is the
largest with a value of 0.973. The proportion of observations greater than 1.96
is less than 5 per cent of the sample indicating that the audit fee model is
generally well specified. However, the extreme observations are clustered in the
largest and smallest client quintiles. All the observations with Studentized
residuals greater than 1.96 are in the extreme quintiles. The extreme observations represent 3.6 and 5.4 per cent in the small and large client quintiles,
respectively. There are no extreme observations in quintiles 2 through 4.
Furthermore, while less than 5 per cent of the observations are potentially
influential, 16 per cent of the largest clients in each industry and 17 per cent of
the second largest clients in each industry have Studentized residuals greater
than 1.96. (For 1999, not reported, 20 per cent of the largest clients and 17 per
cent of the second largest clients have Studentized residuals greater than 1.96.)
The large clients in each industry are overrepresented in the potentially influential
observations.
After deleting any potentially influential observations (with absolute value of
the Studentized residuals greater than 1.96), we consider the pattern in the coefficient on log of total assets when audit fee models are estimated on subsamples
of clients by size quintile. The results are reported in Panel B of Table 3. The
coefficients on log of total assets increase with client size. Large audit clients
pay more per dollar of assets than small audit clients. The coefficient on size is
no longer monotonically increasing across size quintiles, but varies depending
on the size of the intercept estimated for the model. Panel B of Table 3 also
reports the same data reported sorted by industry-size quintile; that is, the
quintile of size of total assets within industry. In this case, the coefficient on
size is monotonically increasing, suggesting that size within industry is a determinant of audit fees.
To provide further evidence of the difference between audit fee model
coefficients for small and large client segments, we divided the sample into
The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

435

Table 3
Client size effects on residuals and coefficients in standard audit fee models
Panel A: Client size effects in Studentized residuals: 1998 Dataa

Variable
Size
Size
Size
Size
Size

quintile
quintile
quintile
quintile
quintile

1
2
3
4
5

Mean of
total assets
($A million)

Mean of
absolute
value of
Studentized
residuals

Median of
absolute
value of
Studentized
residuals

Proportion
extreme
observations
(5% level)b %

Median of
Studentized
residuals

3078.420
169.798
44.530
13.755
3.501

0.973
0.695
0.702
0.757
0.812

0.832
0.612
0.538
0.667
0.667

3.604
0
0
0
5.405

0.387
0.084
0.287
0.145
0.302

Panel B: The association between client size and audit fees by quintile of client size after removing
potentially influential observationsc

Variable
(1)

(2)

Rank on client total assets


Size quintile 1
Size quintile 2
Size quintile 3
Size quintile 4
Size quintile 5
Rank on size within industry
Size quintile 1
Size quintile 2
Size quintile 3
Size quintile 4
Size quintile 5

Mean of total assets


($A million)

Intercept

Coefficient on log
of total assets: 1

3010.941
162.991
44.337
13.703
3.635

3.737
1.841
5.363
0.879
1.386

0.646
0.478
0.835
0.436
0.362

2298.216
644.629
320.023
71.428
13.922

3.888
2.970
2.638
1.863
1.602

0.671
0.596
0.591
0.533
0.469

Studentized residuals from the basic audit fee model excluding industry specialist auditor. The sample
excludes audits of financial and property trusts (industry codes 16, 17, 19, and 20).
Model: LAF = + 1LTA + 2LSUB + 3CATA + 4QUICK + 5DE + 6ROI + 7FOREIGN + 8OPINION
+ 9YE + 10LOSS + .
LAF is log of total audit fees paid to the auditor ($A thousands); LTA is log of total assets ($A
thousand); LSUB is the log of the number of subsidiaries; CATA is the ratio of current assets to total
assets; QUICK is the ratio of current assets less inventories to total assets; DE is the ratio of long-term
debt to total assets; ROI is the ratio of earnings before interest and tax to total assets; FOREIGN is the
proportion of subsidiaries that represent foreign operations; OPINION is an indicator variable with a
value of 1 for a modified or qualified opinion; YE is an indicator variable with a value of 1 for non-June
fiscal year end; and LOSS is an indicator variable with a value of 1 for a loss in any of the last 3 years.
b
Proportion of extreme observations is proportion of quintile observations with the absolute value of
Studentized residuals greater than 1.96. cThis table reports coefficients for fee models estimated on
subsamples partitioned by client size excluding all potentially influential observations identified as
observations with the absolute value of Studentized residuals greater than 1.96.

The Authors

Journal compilation 2007 AFAANZ

436

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

client segments below and above the median size (not reported). We can reject
the hypotheses that the coefficients on total assets, number of subsidiaries,
current ratio, quick ratio, and number of foreign subsidiaries in the large client
subsample are equal to the value of the coefficients in the small client subsample.
The coefficient on the log of total assets is 51 per cent higher for the large client
sample than for the small client sample.
Overall, the results so far suggest that the audit fee model is reasonably well
specified; however, there are distinct patterns in the association between client
size and audit fees in the extreme client size quintiles. The observations that
potentially have significant influence are clustered in the quintiles for the smallest
and largest companies. As prior research does not exclude these observations,
and exclusion would remove the largest clients in each industry from the sample,
we use the full sample in the subsequent analysis but we allow the intercepts
and slopes to vary with client size.
6.2. The impact of client size on industry specialist fee premiums
Table 4 estimates fee models with variables capturing the various aspects of
client size predicted to impact fees. Column 3 provides results for a basic fee
model with an indicator variable for the two largest industry specialist auditors
based on market share for comparison with prior research. The model is well
specified with an adjusted R2 of 83.2 per cent. The coefficient on specialist is
positive and significant (coefficient 0.12, t = 2.26).
Column 4 of Table 4 includes the additional variables allowing the coefficient
on log of total assets to vary with client size. Consistent with the results from
Table 3, there is a U-shaped pattern in the coefficients with slightly higher
coefficients for the largest and smallest clients. The differences between coefficients are not statistically significant; however, even small differences when
multiplied by log of total assets predict significantly different fees in the
extreme quintiles. The indicator variable for cross-listing in the USA is positive
and significant (coefficient 0.22, t = 2.27). With respect to potential client
bargaining power, the coefficient is positive and significant (coefficient 0.33,
t = 2.78). This is consistent with the view that clients that are relatively large in
their industry tend to choose an auditor that has successfully differentiated
themselves from their competitors and, hence, retains strong bargaining power
with the client. We do not find any evidence of the clients that are relatively
larger in the industry obtaining any fee discount. The coefficient on specialist
remains positive and significant (coefficient 0.12, t = 2.30) similar to the initial
model estimate.
Column 5 of Table 4 partitions the specialist premium by quintile of client
size. The premium to specialization for the largest quintile of clients is 0.28
compared to 0.08 for the middle quintiles and 0.11 for the smallest quintiles.
The coefficient on specialization is significant for the largest clients but not
significant for the majority of clients. A hypothesis of equal coefficients for
The Authors

Journal compilation 2007 AFAANZ

Year:

(2)

Premiums to
industry specialist
(3)

INTERCEPT
LTA
LTA*QUINTILE 1 (large)
LTA*QUINTILE 2
LTA*QUINTILE 3
LTA*QUINTILE 4
LTA*QUINTILE 5 (small)
LSUB
CATA
QUICK
DE
ROI
FOREIGN
OPINION
YE
LOSS
ADR
POWER
SPECIALIST
SPECIALIST and LARGE
SIZE QUINTILE 1
SPECIALIST and MEDIUM
SIZE QUINTILES 2 4

(.)
(+)

1.86 (8.88)
0.51 (28.98)***

(+)
(+)
()
(+)
()
(+)
(+)
()
()
(+)
(+/)
(+)
(+)
(+)

0.04
0.80
0.02
0.36
0.52
0.60
0.07
0.11
0.18

(5.24)***
(7.25)***
(4.76)***
(1.86)
(5.10)***
(5.82)***
(0.83)
(1.82)
(2.70)***

0.12 (2.26)**

Expanded model
to allow fees to
vary with size
(4)

Expanded model
to allow specialist premium
to vary by client size 1998
(5)

Expanded
model 1999
sample
(6)

Expanded model
2004 sample revised
GICS industries
(7)

1.22 (3.08)

1.15 (2.90)

1.04 (2.50)

0.34 (0.91)

0.46
0.44
0.43
0.43
0.45
0.04
0.82
0.01
0.32
0.36
0.56
0.09
0.09
0.15
0.22
0.44
0.12

0.45
0.44
0.42
0.42
0.45
0.04
0.82
0.02
0.33
0.35
0.55
0.09
0.09
0.14
0.21
0.45

0.44
0.42
0.40
0.40
0.41
0.03
0.89
0.02
0.58
0.19
0.76
0.13
0.03
0.02
0.25
0.41

0.42
0.39
0.39
0.38
0.40
0.03
0.91
0.03
0.54
0.27
0.11
0.09
0.07
0.01
0.26
0.24

(15.78)***
(13.21)***
(11.61)***
(10.49)***
(9.53)***
(5.64)***
(7.44)***
(5.40)***
(1.72)
(3.49)***
(5.64)***
(1.12)
(1.55)
(2.32)***
(2.27)**
(3.05)***
(2.30)**

(14.72)***
(13.08)***
(11.46)***
(10.31)***
(9.22)***
(5.67)***
(7.38)***
(5.40)***
(1.75)**
(3.38)***
(5.54)***
(1.08)
(1.44)
(2.24)***
(2.18)**
(3.05)***

0.28 (2.18)**
(n = 77)
0.08 (1.26)
(n = 179)

(14.13)***
(12.01)***
(10.60)***
(9.49)***
(8.43)***
(4.47)***
(8.50)***
(5.46)***
(3.29)***
(1.79)**
(7.75)***
(1.64)
(0.32)
(0.30)
(2.74)***
(3.01)***

0.23 (1.84)**
(n = 75)
0.16 (2.55)***
(n = 178)

(15.26)***
(12.68)***
(11.53)***
(10.11)***
(9.33)***
(5.27)***
(9.14)***
(9.76)***
(2.71)***
(2.35)***
(2.19)**
(1.11)
(1.30)
(0.18)
(2.64)***
(1.03)

0.31 (2.66)***
(n = 86)
0.12 (1.92)**
(n = 217)

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Journal compilation 2007 AFAANZ

The Authors

Table 4
Audit fee premiums and client size

437

438

Year:

(2)

SPECIALIST and SMALL


SIZE QUINTILE 5
N
Adjusted R2 (%)

(+)

Premiums to
industry specialist
(3)

558
83.2

Expanded model
to allow fees to
vary with size
(4)

Expanded model
to allow specialist premium
to vary by client size 1998
(5)

Expanded
model 1999
sample
(6)

Expanded model
2004 sample revised
GICS industries
(7)

558
84.3

0.11 (0.94)
(n = 62)
558
84.3

0.12 (1.10)
(n = 57)
543
84.9

0.02 (0.19)
(n = 72)
611
82.5

*** and ** indicate significance at the 1 and 5 per cent levels (one-tailed), respectively. LTA is log of total assets ($A thousand); LSUB is the log of the number
of subsidiaries; CATA is the ratio of current assets to total assets; QUICK is the ratio of current assets less inventories to total assets; DE is the ratio of longterm debt to total assets; ROI is the ratio of earnings before interest and tax to total assets; FOREIGN is the proportion of subsidiaries that represent foreign
operations; OPINION is an indicator variable with a value of 1 for a qualified opinion; YE is an indicator variable with a value of 1 for non-June fiscal year end;
LOSS is an indicator variable with a value of 1 for a loss in any of the last 3 years; ADR is an indicator variable taking the value 1 if the firm is cross-listed as
an American Depository Receipt; POWER is a measure of client bargaining power relative to the auditors total clientele in the industry as measured by the
natural log of company sales, divided by the sum of logged sales for all firms in the industry audited by the companys auditor; and SPECIALIST is an indicator
variable with 1 if the client is audited by the specialist auditor in the industry, where specialist auditor is identified as the two auditors with the largest market
share based on the highest proportion of industry audit fees.

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Journal compilation 2007 AFAANZ

The Authors
Table 4 (continued)

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

439

the largest size quintiles of clients and for size quintiles 24 cannot, however,
be rejected at an alpha level of 1 per cent (F = 1.81, p = 0.18) due to large
standard errors for this test.
Overall, the results suggest that the typical audit fee model using Australian
data needs to include an indicator variable for cross-listing to capture the
demand for increased audit services by this group of large clients. Furthermore,
the premium to specialization appears to be clustered in the fees paid by the
largest clients. This confirms the seminal research in the area but also documents
the extent to which fee premiums to specialization are conditional on client size.
6.3. Client size and measures of office level industry expertise
Table 5 provides a similar fee analysis using an audit fee model with an
indicator variable for the city-industry leader (Ferguson et al., 2003). Only
audits in the five largest metropolitan areas are included in this analysis. The
city-industry leader variable takes the value of 1 when the auditor is the number
one auditor in the city and one of the top two firms nationally (CITYIL). Three
partitions of the city-industry lead auditor variable are also included in the
analysis partitioning the specialization measures by client size quintiles. The
indicator variables take the value of 1 if the auditor is a city-industry leader
(CITYIL) and the client is in the highest quintile of the full sample by total
assets, quintiles 2 to 4, or quintile 5, respectively.
Column 3 of Table 5 reports the coefficients for a basic fee model. The
coefficient on city cost is positive and significant. Audits where the head office
of the client is in a higher-cost city have higher fees than audits of clients with
head offices in lower-cost cities. The premium on the city-industry leader in is
0.16, which is a little lower than Ferguson et al. (2003) but still a very large
premium. Column 4 allows the coefficient on total assets to vary with client
size, includes the indicator variable for cross-listing in the USA and allows
the city-industry premium to vary with client size. Consistent with the
industry-level analysis the coefficient on cross-listing in the USA (ADR) is
large and significant (coefficient 0.19, t = 2.06). The coefficient on the
city-industry leader indicator variable remains positive and significant
(coefficient 0.33, t = 2.78). As for industry the coefficient on industry specialization for clients in quintiles 24 is not significantly different from 0. As
discussed below, the coefficient on the medium quintiles is, however, significantly greater than 0 for 1999 and 2004. The pattern of coefficients and the
resulting inferences are generally consistent with those reported for national
industry-level analysis.
7. Sensitivity analysis and further discussion
In this section, we consider the sensitivity of our results to the period examined and alternate definitions of industry specialist (Gramling and Stone, 2001).
The Authors

Journal compilation 2007 AFAANZ

440

Year:

(2)

Premiums to
city-industry
specialist
(3)
3.41 (10.75)

Expanded model
to allow specialist premium
to vary by client size 1998
(4)
2.86 (6.11)

Expanded
model
1999
(5)
2.28 (4.65)

Expanded model
2004 revised
GICS industries
(6)

INTERCEPT

(.)

LTA (Average coefficient in columns 4, 5, 6)

(+)

0.48(27.32)***

0.43 (11.84)***

0.39(10.56)***

0.37(10.84)***

LSUB

(+)

0.04 (5.32)***

0.04 (5.73)***

0.02 (4.66)***

0.03 (5.04)***

CATA

(+)

0.74 (6.83)***

0.77 (7.10)***

0.81 (7.72)***

0.88 (8.72)***

QUICK

()

0.02 (4.99)***

0.02 (5.58)***

0.02 (5.11)***

0.03 (9.55)***

DE

(+)

0.42 (2.21)**

0.40 (2.16)**

0.63 (3.60)***

0.54 (2.71)***

ROI

()

0.57 (5.70)***

0.40 (3.98)***

0.23 (2.14)***

0.24 (2.15)**

FOREIGN

(+)

0.62 (6.16)***

0.58 (5.94)***

0.76 (7.75)***

0.09 (1.78)**

OPINION

(+)

0.05 (0.65)

0.08 (0.94)

YE

()

0.06 (0.98)

0.04 (0.67)

LOSS

()

0.22 (3.42)***

0.20 (3.11)***

CITYCOST

(+)

0.04 (6.71)***

0.04 (6.75)***

0.16 (1.99)**

1.02 (2.03)

0.09 (1.13)

0.03 (0.63)

0.09 (1.47)

0.04 (0.70)

0.06 (0.85)

0.03 (5.31)***

0.02 (2.95)***

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Journal compilation 2007 AFAANZ

The Authors
Table 5
Fee premiums to city-industry leaders

(2)

ADR

(+)

POWER

(+/)

CITYIL

(+)

CITYIL and LARGE


SIZE QUINTILE 1

Expanded model
to allow specialist premium
to vary by client size 1998
(4)

Expanded
model
1999
(5)

Expanded model
2004 revised
GICS industries
(6)

0.19 (2.06)**

0.22 (2.41)***

0.25 (2.56)***

0.37 (2.53)***

0.36 (2.60)***

0.08 (0.33)

(+)

0.33 (2.78)***
(n = 46)

0.40 (3.33)***
(n = 46)

0.52 (4.57)***
(n = 58)

CITYIL and MEDIUM


SIZE QUINTILES 2 4

(.)

0.12 (1.51)
(n = 88)

0.22 (2.76)***
(n = 80)

0.16 (2.19)**
(n = 99)

CITYIL and SMALL SIZE QUINTILE 5

(.)

0.05 (0.33)
(n = 18)

0.17 (1.24)
(n = 20)

0.18 (1.47)
(n = 30)

0.08 (1.29)

0.08 (1.34)

0.10 (1.59)

0.09 (1.45)

0.08 (0.77)
546
84.3

0.08 (0.84)
546
85.3

0.06 (0.69)
526
85.6

NATIONAL 1 or 2 but not CITY 1


CITY 1 not NATIONAL 1 or 2
N
Adjusted R2 (%)

0.16 (2.48)***

0.26 (2.13)**
594
83.1

*** and ** denote significance at the 1 and 5 per cent levels (one tailed), respectively. LTA is log of total assets ($A thousand); LSUB is the log of the number
of subsidiaries; CATA is the ratio of current assets to total assets; QUICK is the ratio of current assets less inventories to total assets; DE is the ratio of longterm debt to total assets; ROI is the ratio of earnings before interest and tax to total assets; FOREIGN is the proportion of subsidiaries that represent foreign
operations; OPINION is an indicator variable with a value of 1 for a qualified opinion; YE is an indicator variable with a value of 1 for non-June fiscal year end;
LOSS is an indicator variable with a value of 1 for a loss in any of the last 3 years; CITYCOST is the use the annual salary cost for an experienced auditor in
each of the major cities from Hays Personnel Services; ADR is an indicator variable taking the value 1 if the firm is cross-listed as an American Depository
Receipt; POWER is a measure of client bargaining power relative to the auditors total clientele in the industry as measured by the natural log of company sales,
divided by the sum of logged sales for all firms in the industry audited by the companys auditor; and CITYIL is an indicator variable with 1 if the company is
audited by the number one auditor in the industry and the city, and number one or two in the industry nationwide based on total audit fees.

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Journal compilation 2007 AFAANZ

The Authors

Year:

Premiums to
city-industry
specialist
(3)

441

442

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

We specifically expect that definitions of industry specialist will affect the estimation of relations between client size, identification of the industry specialist,
and the resulting measures of fee premiums.
7.1. Sensitivity to period examined
We chose the 1998 year to facilitate comparison with research by Ferguson
et al. (2003). To examine the robustness of the results to the period examined,
we re-estimated the results using equivalent data for 1999 and for 2004. In
1999, for the base model, the coefficient on specialist is positive and significant
(coefficient 0.17, t = 3.16). Consistent with the results for 1998, there is a Ushaped pattern in the coefficients when the coefficient is allowed to vary by
quintile of total assets. For the complete model (reported in column 6 of
Table 4), the coefficient on the indicator variable for cross-listing in the USA is
positive and significant (coefficient 0.25, t = 2.74). The premium to specialization for the largest quintile is 0.23 compared to 0.16 for the middle quintiles
and 0.12 for the smallest quintile. The coefficient for the interaction between
specialist premium and medium size of client is significantly greater than 0;
however, there is a consistent decline in premium between the largest clients
and the medium size of client. Overall, the results are consistent with those
reported using the 1998 data.
With the demise of Arthur Andersen, there has been a further concentration
of suppliers of audit services. To establish whether our results are generalizable
to this period, we re-estimate our models for 2004. For the base model, reported
in column 7 of Table 4, the ADR variable continues to be positive and significant (coefficient of 0.26, t = 2.64). We also find evidence of premiums to
national specialist for the largest quintile of clients (0.31, t = 2.66), which
decline across the medium and small quintiles. In column 6 of Table 5, we
report a similar pattern of results for city-industry leaders. Overall, we find evidence that the pattern of premiums to specialization for large firms is robust to
time period despite the significant change in industry designations adopted by
the ASX before 2004. However, in the post-Arthur Andersen market, we no longer
find support for a higher fee paid by relatively large clients in an industry as we
found in earlier periods. For 2004, we do find a premium to the city, but not
national leader (0.26, t = 2.13) that appears to clustered in industry groupings
for GICS code 1510 (materials) and GICS code 3020 (food) where larger industry
groupings than previously result in some large companies employing city leading
but not industry leading auditors. However, we have not partitioned this group
by client size because of the small number of companies involved.
7.2. Number of clients in the industry
A concern in use of the market share measure of specialization in prior literature is the intuition that an industry must have a minimum number of clients
The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

443

in an industry for an auditor to be designated a specialist in that industry.


Although Ferguson et al. (2003) assign specialists in all industries, earlier
research by Craswell and Taylor (1991) and Craswell et al. (1995) were careful
in only identifying specialists in industries with greater than 30 observations
and only identifying auditors as specialists where the auditor has a threshold of
10 per cent of market share on either the number of clients or the percentage of
total audit fees. Replicating our models on the nine industries that previous
research has identified as having industry premiums (0104, 07, 09, 11, 13 and
19), we find similar results to those reported. Influential observations are
clustered in the largest and smallest quintiles by asset size. For the full model
(column 5 of Table 4), fee premiums to specialization are found for ADR
(coefficient 0.21, t = 2.18). The premiums to specialization for the large, medium
and small clients were 0.28, 0.08 and 0.11, respectively.
Replicating our models on the 7 industries that previous research has identified
as having industry premiums and that have 30 clients in our sample (1, 2, 4, 9,
11, 13 and 19), we again find results consistent with those reported above.
7.3. Number of auditors in the industry and identifying a specialist
Seminal research on specialization (e.g. Craswell et al., 1995) examines a
market with eight large international audit firms. Given the small number of
auditors in the current market we consider the ability of the typical methodology
to distinguish between alternate explanations for specialist fee premiums. To
determine whether the fee premium attributed to specialization can also be
explained by client attributes, we examine the observed fee premium when
clients are assigned to five auditors at random. That is, we randomize clients
within industry with respect to the reputation of the auditor selected. We initially allocated all clients to five auditors using a random draw from a uniform
distribution. The industry specialist auditor is then identified based on the highest share of audit fees in an industry and the basic fee model estimated (as per
column 3 of Table 4). We repeated this procedure 1000 times. The mean coefficient on the industry specialist auditor is 0.12 with a range of 0.06 to 0.19.
The largest clients in each industry pay fee premiums in the Australian market,
and whichever auditor has these clients is designated the industry specialist and
receives an apparent fee premium. Fee premiums of the magnitude typically
observed for industry specialist auditor can be explained by the fee premiums
paid by the largest clients, and whether this premium is due to industry specialization or other client attributes requires additional information not typically
available in this type of study.
As a further sensitivity, we also re-estimated the regressions reported as
Table 4 removing clients from one auditor from each estimation. For 1998, only
the test of the significance of the coefficient on the ADR variable is not significant when KPMG is excluded (KPMG audits 13 of 49 clients with ADRs in
1998). The coefficient on the SPECIALIST and LARGE SIZE QUINTILE 1
The Authors

Journal compilation 2007 AFAANZ

444

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

interaction is also not significant when Price waterhouse Coopers clients are
excluded (166 clients are excluded from the sample). However, the coefficients
are consistently in the direction of those reported and most statistical inferences
are consistent with those reported.
8. Conclusion
Previous research suggests that industry specialist auditors earn higher fees.
Previous research has also found that, in the Australian environment, higher
fees are concentrated in clients of larger size. This research attempts to provide
further insight into the association between the client characteristic of size and
audit fee premiums.
Premiums to national industry specialists are concentrated in the fees paid by
the largest clients in each industry. The premium to industry specialist auditors
is not significantly different from 0 for the middle and smaller quintiles of clients
based on total assets. The findings are consistent with auditor specialization
only benefiting large clients. Another possible explanation is that the client
characteristics of large clients impact audit fees and the auditors of the largest
clients are designated the industry specialist. Given the differences in apparent
premiums to large and medium clients it appears important to partition the
specialist premium by client size, or to report results by subsamples partitioned
by client size (e.g. above and below the median). Given the strong relationship
between client size and audit fees, and that audit fees do not vary strictly linearly
with client size, we suggest that it is important to examine the robustness of any
observed fee premium across clients of varying size.
We document a variable that can be considered an omitted variable in typical
audit fee models. Companies that are cross-listed on US exchanges pay higher
audit fees. As these large companies have more complex filing requirements it
is to be expected that they pay higher audit fees. The premium to this factor is
large relative to premiums attributed to industry specialization and should be
considered in audit fee models using Australian data.
We also document three smaller effects. There is non-linearity in the association between client size and audit fees. For the period examined, both smaller
and larger clients tend to pay higher audit fees. We also observe that audit costs
vary by the city of the client head office. Clients located in higher-cost cities
pay higher audit fees. We also find evidence that clients that are large relative
to their industry paid higher fees in the earlier years of our sample period. We
find no evidence consistent with larger clients being able to negotiate lower
audit fees.
The issue of fee premiums paid by large clients is of particular concern
because of the possible explanation that there is a potentially less competitive segment for large auditees that is dominated by a few large audit firms
(Simunic, 1980; Francis and Stokes, 1986). In the USA, the SarbanesOxley Act
explicitly refers to the reduction in the number of firms capable of providing
The Authors

Journal compilation 2007 AFAANZ

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

445

audit services to large businesses (SarbanesOxley Act, US Congress, 2002,


section 701). This paper highlights the higher fees received from larger clients
in each industry with indirect implications for the importance of these clients to
the auditor and auditor independence. However, we find evidence consistent
with a higher demand for more complex accounting and auditing issues from
these clients that needs to be considered before attributing fee premiums to a lack
of competitive pressure. Because of the changes observed in the audit market,
future research could also consider these issues in a dynamic market setting
with a declining pool of large auditors available to audit large companies.
References
Bell, T. B., W. R. Knechel, and J. J. Willingham, 1994, An exploratory analysis of the
determinants of audit engagement resource allocations, Proceedings of the Deloitte &
Touche / University of Kansas Symposium on Auditing Problems, Auditing Symposium
XII, 4967.
Casterella, J. R., J. R. Francis, B. L. Lewis, and P. L. Walker, 2004, Auditor industry
specialisation, client bargaining power and audit pricing, Auditing: A Journal of Practice
& Theory 23, 123140.
Chen, Y., R. Moroney, and K. Houghton, 2005, Audit committee composition and the use
of an industry specialist audit firm, Accounting and Finance 45, 217239.
Chung, H., and S. Kallapur, 2003, Client importance, nonaudit services and abnormal
accruals, Accounting Review 78, 931955.
Craswell, A., and J. Francis, 1999, Pricing initial audit engagements: a test of competing
theories, Accounting Review 74, 201216.
Craswell, A., and S. Taylor, 1991, The market structure of auditing in Australia: the role of
industry specialization, Research in Accounting Regulation 5, 5577.
Craswell, A., J. Francis, and S. Taylor, 1995, Auditor brand name reputations and industry
specialisations, Journal of Accounting and Economics 20, 297322.
DeFond, M. L., J. R. Francis, and T. J. Wong, 2000, Auditor industry specialisation and
market segmentation: evidence from Hong Kong, Auditing: A Journal of Practice &
Theory 19, 4966.
Easton, P. D., and G. A. Sommers, 2003, Scale and the scale effect in market-based
accounting research, Journal of Business, Finance & Accounting 30, 2555.
Ferguson, A., J. R. Francis, and D. Stokes, 2003, The effects of firm-wide and office-level
industry expertise on audit pricing, Accounting Review 78, 429448.
Ferguson, A., J. R. Francis, and D. Stokes, 2006, What matters in audit pricing: industry
specialization or overall market leadership? Accounting and Finance 46, 97106.
Ferguson, A., and D. Stokes, 2002, Brand name audit pricing, industry specialisation, and
leadership premiums post-Big 8 and Big 6 mergers, Contemporary Accounting Research
19, 77100.
Fields, L. P., D. R. Fraser, and M. S. Wilkins, 2004, An investigation of the pricing of
audit services for financial institutions, Journal of Accounting and Public Policy 23,
5377.
Francis, J., 1984, The effect of audit firm size on audit prices, Journal of Accounting and
Economics 6, 133151.
Francis, J., and D. Simon, 1987, A test of audit pricing in the small-client segment of the
U.S. audit market, Accounting Review 62, 145157.
Francis, J., and D. Stokes, 1986, Audit prices, product differentiation, and scale economies:
further evidence from the Australian market, Journal of Accounting Research 24, 383 393.

The Authors

Journal compilation 2007 AFAANZ

446

E. Carson and N. Fargher /Accounting and Finance 47 (2007) 423446

Francis, J., K. Reichelt, and D. Wang, 2005, The pricing of national and city-specific reputations for industry expertise in the U.S. audit market, Accounting Review 80, 113136.
Gramling, A., and D. Stone, 2001, Audit firm industry expertise: a review and synthesis of
the archival literature, Journal of Accounting Literature 20, 127.
Hays Accounting Personnel Salary Survey, 1998, 1999, 2004, Hays Accounting Services
[online; cited 1 December 2005]. Available: www.hays.com.au/salary/.
JPMorgan, Listing of American Depositary Receipt Issuers, JPMorgan, New York [online;
cited 12 May 2006]. Available: www.adr.com.
Mayhew, B. W., and M. S. Wilkins, 2003, Audit firm industry specialisation as a differentiation strategy: evidence from fees charged to firms going public, Auditing: A Journal of
Practice & Theory 22, 3352.
Palmrose, Z., 1986, Audit fees and auditor size: further evidence, Journal of Accounting
Research 24, 97110.
Reynolds, J. K., and J. R. Francis, 2001, Does size matter? The influence of large clients on
office-level auditor reporting decisions, Journal of Accounting and Economics 30, 375
400.
SarbanesOxley Act, 2002, Corporate Responsibility, Public Law 107204 (107th Congress
of the United States, Washington, DC).
Simunic, D., 1980, The pricing of audit services: theory and evidence, Journal of Accounting Research 18, 2735.
Sullivan, M. W., 2000, The Effect of Big Eight Accounting Firm Mergers on the Market for
Audit Services, working paper (US Department of Justice, Washington, DC).
Who Audits Australia database, 1998/1999, Faculty of Economics and Business, University
of Sydney, Sydney, NSW.

The Authors

Journal compilation 2007 AFAANZ

Das könnte Ihnen auch gefallen