Transaction Costs and Competition among Audit Firms in Local Markets *
Ling Chua, Dan A. Simunicb, Minlei Yec, Ping Zhangd
a
Laurier School of Business and Economics, Wilfred Laurier University, 75 University Avenue
West, Waterloo, ON, Canada N2L 3C5
b
Sauder School of Business, University of British Columbia, Vancouver, BC, Canada, V6T1Z2
Rotman School of Management, University of Toronto, 105 St George Street, Toronto, ON,
Canada, M5S 3E6
c, d
July 2015
Abstract
A large body of academic literature interprets the dominant audit firms in audit markets - as
defined by client companies operating in specific client industries - to be high quality,
differentiated suppliers who command higher audit fees. The dominant audit firms are normally
defined by their large client-industry market shares. However, in the industrial organization
literature, dominant suppliers with large market shares are normally viewed as having market
power, which can result in those firms earning economic rents. So the interpretation of any audit
fee premium to dominant audit firms in a market is ambiguous. In this paper, we develop a
measure for audit firm competition in local audit markets based on the transaction costs of
changing audit firms included in DeAngelo’s (1981) multi-period audit pricing model. Our
competition measure reflects the relative size difference between the largest audit firm in a
market and the other audit firms operating in that market. We find that audit fees decrease as the
size difference between the largest audit firm in a market and a client’s incumbent audit firm
increases. The evidence suggests that dominant audit firms charge higher audit fees because of
their significant local competitive advantage over smaller audit firms. Our study advances
understanding of audit market competition and provides an alternative explanation for the excess
audit fees earned by the largest audit firm(s) in local audit markets.
Keywords: Audit pricing, competition, transactions cost
*
We appreciate the helpful comments from participants in accounting research workshops at the
University of British Columbia, University of Pittsburgh, the University of Toronto, Lehigh University,
the University of Technology – Sydney, and conference participants at the 2014 Auditing Section
Midyear Meeting, the 2014 European Accounting Association Conference, and the 2014 Canadian
Academic Accounting Association Conference. Dan Simunic thanks the Social Sciences and Humanities
Research Council of Canada for financial support. Ling Chu and Ping Zhang also thank the Social
Sciences and Humanities Research Council of Canada for financial support.
1
Transaction Costs and Competition among Audit Firms in Local Markets
1. Introduction
The nature of competition among audit firms has been of concern to regulators (e.g.
Subcommittee on Reports, Accounting and Management of the Commission on Government
Operations U.S. Senate 1977; Government Accountability Office (GAO) 2003, 2008) and of
interest to researchers (e.g. Dopuch and Simunic 1980) for many years. These concerns center
around the possibility that a lack of competition among audit firms will lead to higher audit
prices, lower audit quality, and hence a lower quality of financial reporting by companies.
Auditor competition is therefore an important issue.
Unfortunately, competition is a dynamic process that is difficult (impossible) for regulators
or researchers to observe directly. As a result, the nature of competition in a market has typically
been examined empirically using the structure → conduct → performance paradigm of industrial
organization. The observable structure of an industry is normally measured by the degree of
supplier concentration (e.g. the market share of the largest k firms, where k = 1, 2, 3, etc. or the
Herfindahl index 1). Conduct, on the other hand is not observed but is inferred from performance,
using measures such as prices, gross margins, or profitability. The empirical correlation between
variations in structure and measures of performance across markets is then used to make
inferences about the degree of competition in an industry, with higher supplier concentration
commonly found to be associated with poorer industry performance (e.g. higher average prices).
1
The Herfindahl index is a standard measure of concentration and potential market power
market and is calculated as:
of firms in a
𝑁𝑁
𝐻𝐻 = � 𝑠𝑠𝑖𝑖2
𝑖𝑖=1
where si is the market share of firm i in the market, and N is the number of firms. H ranges in value from 1/N to 1,
where a small index suggests a competitive industry with no dominant firms, while a large index value indicates a
market with a dominant player(s).
2
Audit markets in the U.S. and elsewhere have clearly become more concentrated since the late
1980’s after several rounds of consolidation among the large public accounting firms which
reduced them from the Big 8 to the Big 4. The increased level of auditor market concentration is
therefore a concern to regulators and market participants.
While supplier concentration is conceptually linked to market power in the industrial
organization literature, the view of supplier concentration in the economics of auditing literature
is different. Beginning with the influential paper by Craswell, Francis, and Taylor (1995), a large
stream of auditing papers (e.g. Ferguson and Stokes 2002, Ferguson, Francis, and Stokes 2003,
Francis, Reichelt, and Wang 2005, Fung, Gul, and Krishnan 2012, and many others) have used
the market shares of auditors in servicing clients within specific industries in a country (e.g.
mining, banking, insurance firms, etc. in Australia) or clients within industries in local
geographic areas (e.g. mining, banking, insurance firms operating in specific metropolitan areas
in Australia, U.S., etc.) not as a measure of concentration and possible market power, but rather
as a measure of auditor-industry specialization.
In this literature, auditors with high market shares are viewed as “industry specialists” who
have incurred costs to develop industry specific expertise, and therefore produce a systematically
higher level of audit quality and charge a systematically higher audit fee for their higher quality
of service. Craswell, Francis and Taylor’s findings (in an Australian context) that audit firms
with national client-industry market shares ≥ 10% - their definition of industry specialists earned a 34% audit fee premium relative to Big 8 non-specialist auditors has not been replicated
in subsequent research in Australia or other countries. However, researchers continue to use
market share measures to define industry specialists, and routinely include a control variable to
“capture industry specialization” in audit fee regression models where the specialist variable is
3
the market share of the single largest auditor (termed the “industry leader”) or the market share
of several dominant audit firms.
Whether any audit fee premium paid to a dominant audit firm(s) is an economic rent
arising from competitive dominance, or a voluntary payment by clients for a higher quality of
service, is an interesting and important question. Conceptually, the idea that differentially higher
quality auditors who incur higher costs and therefore charge a higher audit fee will necessarily
also enjoy a higher market share is dubious.
For consumer goods, if both a higher and lower quality of product are available in a
market, the market shares of the two products will depend upon their relative prices, consumers’
tastes and preferences (indifference curves), and income (wealth) constraints. Depending upon
these factors, either the higher or lower quality product could enjoy the larger market share.
While audits are not consumer goods but intermediate services purchased by client companies
for the purpose of maximizing firm value, the factors that drive client demand for higher vs.
lower quality audits are analogous. That is, a client firm will voluntarily choose between higher
vs. lower audit quality depending upon their relative prices, firm specific factors that influence
the relative values of higher vs. lower audit quality (e.g. are the firm’s shares publicly traded vs.
closely held), and any budget constraint (e.g. is the firm facing a high bankruptcy risk and
therefore needs to conserve cash). Given these determinants of client firm demand, it seems
inappropriate to assume that audit firms that enjoy high client-industry market shares are
necessarily higher quality auditors. This could be the case in some contexts but is almost
certainly not the case in all contexts.
We believe that many auditing researchers recognize this basic conceptual weakness that
underlies the “industry specialization” literature. However, they are willing to interpret the
4
higher audit fees paid to dominant “industry specialist” auditors as payment for higher audit
quality rather than some form of economic rent because of empirical evidence that dominant
audit firms are also associated with higher quality financial reporting by client companies, thus –
by implication – higher audit quality.
In their recent comprehensive review of archival auditing research, DeFond and Zhang
(2014) point out that many studies have indeed found that national-level industry specialist
auditors are, on average, associated with several high audit quality proxies, including
discretionary accruals, earnings response coefficients, going concern opinions, analyst forecast
accuracy, and other measures. Moreover, there is also more limited evidence that city-level
industry specialists provide higher audit quality (e.g. Balsam et al 2003, Dunn and Mayhew
2004, Lim and Tan 2008, Behn et al 2008, Reichelt and Wang 2010). However, other researchers
(Chin and Chi 2009) have found that industry expertise is primarily a characteristic of the audit
partner who signs the audit report, and that audit firm level industry specialization effects are
insignificant after the industry expertise of the signing partner is included in an analysis.
Cadman and Stein (2007) investigated the relationships among audit fee premiums, auditor
market shares, and audit quality measured a number of ways, including the absolute value of
client discretionary accruals, and the ability of recorded discretionary accruals to predict future
cash flows, and found little evidence that high market share auditors provide increased audit
quality. Indeed, they found that most auditors with high local client-industry market shares do
not charge a fee premium at all, and that findings of audit fee premiums in pooled (across
industry tests) are primarily attributable to a small set of industries in which the high market
share auditors have a very dominant position. They conclude that the available evidence is more
supportive of the hypothesis that high market share audit firms are extracting rents, rather than
5
providing quality differentiated products. Finally, a recent paper by Minutti-Meza (2013)
suggests that the relation between audit quality proxies and auditor industry specialists can be
explained by self-selection. Minutti-Meza shows that after matching clients of specialist and
non-specialist auditors on a number of dimensions, there are no statistically significant
differences in the audit quality proxies between the two groups of auditors. In their review paper,
DeFond and Zhang conclude that self-selection is indeed a legitimate concern; that it is
premature to draw definite conclusions based on available empirical evidence; and they call for
further research. Given the weak conceptual underpinnings linking audit firm client-industry
dominance to industry expertise and higher audit quality, and the mixed empirical results testing
these links, we believe it is appropriate to take a fresh look at these issues.
As noted earlier, in the economic literature on industrial organization, supplier dominance
in a market is usually measured by the degree to which supply is concentrated in the hands of a
few dominant suppliers. In markets for physical goods, and perhaps most services, there will be a
uniform market price at which a good or service can be purchased, with this price increasing if
dominant suppliers are able to extract economic rents from customers. However, the situation is
more complex in the market for audit services. Audit production is very much client-specific and
the characteristics of the audit service and audit fees are known to vary greatly with the size,
complexity, and cash-flow risk of clients. In addition, characteristics of the audit firm (e.g. is it a
Big 4 or non-Big 4 firm) will also affect the audit service and audit fees. As a result overall
industry dominance (e.g. high concentration ratio) by a subset of suppliers may not allow them to
charge higher audit fees for all clients. This is consistent with available evidence. For example,
using market concentration (i.e., the Herfindahl index) as a proxy for the overall level of
competition, Pearson and Trompeter (1994) find that higher industry concentration negatively
6
affects audit fees, and Bandyopadhyay and Kao (2004) do not find that average audit fees are
higher in more concentrated markets. However, Feldman (2006) documents that audit fees have
increased with increased market concentration after the demise of Arthur Anderson. Such mixed
results lead to questions concerning the adequacy of using overall market concentration as a
proxy for auditor competition.
Consistent with this perspective, Numan and Willekens (2012) argue that it is not often the
case that all firms in an industry face the same level of competition. Based on spatial competition
theory of oligopolistic pricing with differentiated products as analyzed in Chan, Ferguson,
Simunic and Stokes (2004), Numan and Willekens (2012) predict and find that, ceteris paribus,
audit fees increase in the “spatial distance” between the incumbent auditor office and that
auditor’s closest competitor, where distance is measured by the relative size of industry market
shares. They find that competitive pressure from a “nearby” competitor has a negative effect on
audit fees for all audit firms, even for the dominant firms who may be considered “industry
specialists”. Thus, Numan and Willekens (2012) make some progress in the measurement of
market competition by testing whether or not price competition is indeed “local”, being both
auditor-specific and client-specific.
In this paper, we extend the analysis of local competition in auditing, focusing on the
effects of variations in the size of suppliers (auditors) in a market on audit service production and
pricing. It is well known that client companies retain their audit firms for multiple years, rather
than changing auditors annually, presumably because the transaction costs of auditor change are
non-trivial. This leads to the multi-period pricing of audit services which was analyzed in
DeAngelo (1981), Magee and Tseng (1990), and Sabac and Simunic (2001). We argue that the
competitive pressure on any incumbent auditor’s fees basically depends upon the ease with
7
which the auditor’s clients can switch to a competing audit firm. If the transaction costs of
auditor change are low, the ability of the incumbent auditor to extract economic rents from
clients is limited 2; conversely, high costs of changing auditors give an incumbent auditor greater
pricing power. We conjecture that transaction costs are associated with the relative size
differences between competing audit firms in a market and the incumbent firm. That is, the
larger the size difference between the largest available supplier and the incumbent supplier, the
lower the transaction cost for a client to switch to the largest supplier.
We combine these arguments with the simple multi-period audit pricing model in
DeAngelo (1981) that includes transaction costs in the form of both auditor learning and clientincurred switching costs, to develop empirical predictions concerning audit fees as a function of
the size of the conjectured costs of auditor change. Our arguments essentially predict that, ceteris
paribus, audit fees charged by an incumbent audit firm that is not the largest auditor office in a
market is a decreasing function of the size of that office. That is, for any audit firm operating in a
market, ceteris paribus, audit fees will decrease as the relative size of the office decreases. We
test this pricing hypothesis using U.S. public company audit fee data for local audit markets
(Standard Metropolitan Area by client 2-digit SIC industries) from 2000-2011 and find evidence
that is strongly consistent with our prediction.
Our tests provide an alternative explanation as to the underlying cause of higher audit fees
paid to city-level dominant audit firms, who are typically considered to be “industry specialists”.
That is, a market leader (i.e., usually considered a “city-level client- industry specialist”) is able
to charge a price premium (relative to smaller firms) due to its market power because it is
relatively more costly for its clients – relative to the clients of smaller audit firms in the market -
2
In the limit, if the transaction costs of auditor change are zero, then audits would simply be purchased and priced
independently each period.
8
to switch to an alternative supplier. Moreover, this fee premium is a continuous function of
market power as measured by the differences in audit firm size - not dichotomous as would be
the case if there was a fixed quality difference between the high market share specialist auditors
vs. the lower market share non-specialist auditors.
Our measurement of competition among auditors in a market contrasts with that of Numan
and Willekens (2012). They find audit fees increase as the absolute difference (termed
”distance”) between the incumbent auditor’s market share and the closest non-incumbent’s
market share increases, whether the non-incumbent’s share is smaller or larger than the
incumbent’s share. We argue that the price pressure in a market comes from the largest audit
firm in the market – rather than from firms of similar size or smaller firms. This implies that
taking the absolute value of market share differences as in Numan and Willekens is
inappropriate. In fact, in sensitivity tests, we find that the pricing effect of the Numan and
Willekens “distance” variable is not statistically significant after controlling for the competitive
advantage of the dominant firm in a market.
The remainder of the paper is organized as follows. In section II, we develop our economic
arguments concerning the crucial role of transaction costs in audit market competition. Section
III lays out the empirical test design, including definitions of key variables. Data are described
and results of the hypothesis tests as well as various additional (sensitivity) tests are reported in
Section IV. Section V summarizes and concludes the paper.
2. Transaction costs and audit pricing
DeAngelo (1981) develops a simple multi-period (perpetuity) audit pricing model that
incorporates the effects of the various determinants of an audit firm’s fee, including start-up
costs of auditor learning and client incurred switching costs. In this paper, we consider both the
9
start-up costs and the switching costs that are incurred if a client changes auditors to be the
transaction costs of audit firm change. The model describes an audit firm’s net present value or
profit (𝜋𝜋) from a client engagement as a function of the initial year’s audit fee (F1), normal
annual production costs (A), first period start-up costs (K), a recurring audit fee in the second and
subsequent year (F), and a discount rate (r), where:
𝜋𝜋 = (𝐹𝐹1 − 𝐴𝐴 − 𝐾𝐾) +
𝐹𝐹 − 𝐴𝐴
𝑟𝑟
In addition, the recurring audit fee, F, is limited by a client’s ability to change audit firms
in order to secure a lower fee while incurring the costs of changing audit firms, denoted CS, such
that:
𝐹𝐹 +
𝐹𝐹
≤ 𝐴𝐴 + 𝐾𝐾 + 𝐴𝐴/𝑟𝑟 + 𝐶𝐶𝐶𝐶
𝑟𝑟
This implies that the entry deterring (audit firm change deterring) audit fee in the second
and subsequent years can be written as:
𝐹𝐹 = 𝐴𝐴 +
𝑟𝑟(𝐶𝐶𝐶𝐶 + 𝐾𝐾)
1 + 𝑟𝑟
DeAngelo (1981) uses these pricing relations to examine the phenomenon termed “lowballing” and argues that competition among audit firms will drive an audit firm’s economic profit
(𝜋𝜋) to zero, yielding an audit fee in the first period (F1) that is less than first period costs (A + K),
because quasi-rents (F > A) can be earned in the second and subsequent periods so long as the
transaction costs of audit firm change are non-zero (CS + K > 0). However, the pricing model is
still relevant if one does not assume that audits are competitively priced. If competition in a
market is weak such that 𝜋𝜋 > 0 and the audit firm earns monopoly rents, the entry deterring
annual audit fee in the second and subsequent periods (F) still depends on the transaction costs of
auditor change (CS + K). Moreover, by inspection of the fee function (the third equation above),
10
F increases monotonically in the value of these transaction costs. The important conclusion from
this pricing model is that positive transaction costs give an audit firm pricing power, that is, the
ability to raise audit fees over the avoidable costs incurred in a period (A) and thus earn quasirents and perhaps monopoly rents. Moreover, an incumbent audit firm’s ability to raise audit fees
increases as the size of transaction costs increases. Note that the pricing power exists even if all
audit firms are the same with respect to the quality of audits performed and/or the recurring costs
of performing audits, which is an assumption in DeAngelo (1981) and which we also assume.
Specifically, there is no auditor-client industry specialization assumed to exist. The only
distinction among audit firms is between the incumbent audit firm and that audit firm’s potential
competitors.
It is useful to note that formally in DeAngelo’s pricing model, a lack of sufficient
competition only affects the value of F1 which would be “too high” and yield π > 0, since
bidding for the client only occurs at the beginning of the engagement. Once an incumbent auditor
is in place, audit fees only depend on the value of A and the transaction costs of auditor change,
with the incumbent auditor pricing the audit so as to deter entry. More generally, however, weak
competition among auditors in the real world may allow a dominant incumbent firm(s) to earn
monopoly rents in all periods – not just in the first period. In our tests, these two possible effects
are combined as we are not able to identify whether higher audit fees represent rents or quasirents.
2.1 Nature of the transaction costs of auditor change
In the pricing model, start-up costs (K) are incurred by the audit firm in the first period of
an engagement and are related to the audit firm learning about the characteristics of a new audit
11
client, while the costs of changing auditors (CS) are potential costs borne by the client only if it
chooses to switch to a new audit firm. In an empirical test using recurring (second and
subsequent period) audit fees, the distinction between K and CS is not important as both
components have the same effect on F. These two costs are conceptually very broad and for a
potential audit firm may include various costs of learning about a new audit client’s business,
transaction flows, accounting, and internal control systems; assessing the abilities and integrity
of key client personnel; assessing the risks of various kinds of material financial statement
misstatements, and so forth. For the client who changes audit firms these costs could include the
time needed to assess competing audit firm bids; adjustment costs of learning to work with and
“training” a new audit team in the details of the company’s operations and systems; and so forth.
To our knowledge, no prior research has measured these transaction costs directly and we
do not attempt to do so in this paper. Rather, our main argument is that these costs are associated
with the relative size of the client portfolios of the audit firms in a market. Specifically, the costs
will be positively related to the incumbent auditor’s client portfolio size and inversely related to
the size of competing audit firms’ client portfolios. For example, all of the existing clients of a
small audit firm in a market can potentially switch to the largest firm operating in the market at
relatively low cost. Conversely, the existing clients of the largest firm in a market would incur
relatively higher transaction costs in attempting to switch to alternative, smaller suppliers. Note,
however, that no actual auditor change is expected to occur. Rather, the relative size of
transaction costs of auditor change affects the pricing power of the incumbent auditor. Returning
to the example, because of the relatively low transaction costs, small audit firms have low
pricing power (K + CS is relatively low). Conversely, the largest incumbent audit firm in a
12
market has high pricing power (K + CS is relatively high). The following paragraph further
explains the rationale for our argument.
The production of audit services is highly labor intensive (O’Keefe, Simunic and Stein
1994) with labor-embodied human capital (e.g. knowledge) playing an important role in audit
production. In a given auditor-client market (e.g. a specific industry in a metropolitan area), if a
client decides to switch from an incumbent to a competitor, then the competitor needs to have the
capacity, such as appropriate personnel, to staff the new engagement. For a profit maximizing
audit firm in a steady state, a local audit office’s existing personnel will mainly be employed to
service current engagements and the slack in staff is likely to be minimal. Given the potential
new demand, the office would have to hire additional personnel to staff its now larger portfolio
of engagements, and in addition to sheer number of staff, the potential competitor audit firm will
have to obtain (i.e. hire or re-assign) staff with the appropriate knowledge to serve a new client,
including the knowledge of the industry in which the new client operates. It is reasonable to
argue that larger auditor offices have greater elasticity in resources to deal with potential new
demand when the slacks are minimal, through reshuffling existing personnel and hiring new
staffs. As a result, larger auditor offices in a local market are likely to be more efficient and have
lower cost than smaller competitors with respect to the transaction costs of auditor change.
In sum, we assume that these transaction costs are a monotonically decreasing function of
the size of auditor offices in a local audit market (i.e. the transaction costs of auditor change are
highest for the clients of the largest audit firm and decrease as the size of the incumbent firm
decreases). Note that other than differences in transaction costs of auditor change, we assume
that the recurring costs of performing an audit are the same across all audit firms.
13
2.2 The incumbent auditor office’s pricing of audit services
The incumbent auditor office receives differential competitive pressures from various
potential suppliers operating in a market. As per the pricing model, the incumbent auditor office
has to price its continuing audits correctly (limit pricing) such that its clients will not switch to
any of its competitors. Consequently, an incumbent auditor office’s fees have to reflect the
lowest transaction costs of changing auditors by its clients, which are most likely the transaction
costs of changing to the largest audit firm (office) in the market. Therefore, although auditors in
a market may have many competitors, the highest competitive pressure comes from the largest
supplier who has the lowest transaction costs. In other words, the largest auditor office is a
reference point for the level of competitive pressure faced by all other auditors in the same
market.
To illustrate, suppose there are three auditor offices in a market, say Big, Middle, and
Small. Then both the Middle and Small offices must price their continuing engagements so as to
deter client switching to Big and their audit fees will be:
F = A + f (CSB + KB)
where the subscript B denotes the costs associated with switching to the Big audit firm. The
larger is the office, Big, the lower the audit fees of its smaller competitors. What about the fees
charged by the Big office? By our assumption that the transaction costs of switching to Big are
the lowest of the firm offices currently in the market, the audit fees of Big will be constrained by
possible client switching to Middle and Small and Big’s fees will be greater than the fees
14
charged by both Middle and Small. Also, the greater the size difference between Big and the next
largest office in the market, the greater will be the fee difference. 3
Taking the argument one step further, it is reasonable to assume the smaller is an auditor
office, the easier (cheaper) it would be for the largest auditor office in the market to
accommodate any client of the small auditor. Therefore, the competitive pressure on an auditor
office would be an increasing function of the size difference between that office and the largest
audit office in the market. As a result, the pricing of the existing clients by an incumbent auditor
office would depend not only on the size of the biggest competitor but the difference in office
size between the incumbent and the largest competitor. Returning to the example, if it is cheaper
for Big to absorb all the clients of Small than it is to absorb all the clients of Middle, then ceteris
paribus, client audit fees would be:
FS = A + f (CS S-B + K S-B)
for the Small office and
FM = A + f (CS M-B + K M-B)
for the Middle office where FM > FS , and finally FB > FM > FS . These arguments lead directly to
the following hypothesis in alternative form:
Hypothesis: The audit fee charged by an incumbent auditor office that is not the largest in
a market is a decreasing function of the difference between the size of the largest auditor and the
size of that office.
The tests of this hypothesis provide an alternative explanation as to the underlying cause of
higher audit fees for “city industry specialists”, which are the dominant auditor offices in audit
markets. We argue that the intensity of competitive pressure is inversely related to the size of the
3
One could argue that competitive pressure on the Big audit office comes from potential new entry outside the
current market structure. However, a new audit firm entrant to a client industry will likely incur additional start-up
costs. This new entry effect is outside the scope of this paper.
15
incumbent auditor office. The bigger the dominant auditor office in a market, the lower is the
transaction cost for a potential client to switch to this office and the larger the transaction cost for
its own clients to switch to a smaller auditor office. Therefore, a market leader (i.e., city industry
specialist) is able to charge a price premium due to its market power, and this fee premium is a
continuous function of market power. Moreover, the fee premium is an economic rent – either a
quasi-rent associated with the size of (CS + K) and/or it could be a monopoly rent. 4 Note that our
hypothesis implies that the distinction between “city industry specialists” and “non-specialists”
does not represent a difference in audit quality and that this difference is not dichotomous but
rather represents a continuous difference in pricing power as a function of the relative size of the
incumbent audit firm. Our approach essentially removes the possibility that the premium in audit
fees charged by larger auditors is related to higher audit quality because it is unlikely that the
proportion of auditees who require certain audit quality matches the relative size of the auditors
while the auditor size is also positively associated with audit quality.
3. Definition of variables and design of our tests
3.1 Empirical model
Building on prior audit fee research (Simunic 1980; Francis et al. 2005; Hay, Knechel, and
Wong 2006; Numan and Willekens 2012), we use the following basic empirical model to test our
hypothesis:
𝐿𝐿𝐿𝐿𝐿𝐿 = 𝛽𝛽0 + 𝛽𝛽1 𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽2 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽3 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 + 𝛽𝛽5 𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄 + 𝛽𝛽6 𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽7 𝑅𝑅𝑅𝑅𝑅𝑅
+ 𝛽𝛽8 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 + 𝛽𝛽9 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝛽𝛽10 𝑌𝑌𝑌𝑌 + 𝛽𝛽11 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽12 𝐵𝐵𝐵𝐵𝐵𝐵
+ 𝛽𝛽13 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹_𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 + 𝜀𝜀,
(1)
4
As mentioned earlier, our empirical tests are not able to identify whether quasi-rents or monopoly rents are
included in audit fees.
16
where the following are control variables:
LAF = natural log of audit fees;
LTA = natural log of total assets;
LBSEG = natural log of the number of unique business segments;
LGSEG = natural log of the number of unique geographic segments;
CATA = ratio of current assets to total assets;
QUICK = ratio of current assets excluding inventory to current liabilities;
LEV = ratio of long-term debt to total assets;
ROI = ratio of earnings before interest and tax to total assets;
FOREIGN = an indicator variable that equals 1 if revenue from foreign operation is
reported, 0 otherwise;
OPINION = an indicator variable coded 1 for modified audit report, and zero otherwise;
YE = an indicator variable that equals 1 for December 31 year-end and 0 otherwise;
LOSS = an indicator variable that equals 1 if loss in current year, and 0 otherwise;
BIG = an indicator variable that equals 1 for Big N auditors and 0 otherwise;
FIXED_EFFECTS = year, industry indicator variables, and MSA indicator variables; and
𝜀𝜀 = random-error term
The control variables included in the audit fee model are based on numerous previous
studies such as Dao, Raghunandan, and Rama (2012), Francis and Yu (2009), Fung et al. (2012),
Hay et al. (2006), Numan and Willekens (2012), and Simunic (1980). We control for client size
(LTA), complexity (LBSEG, LGSEG, FOREIGN), and bankruptcy risk (CATA, QUICK, LEV,
ROI, LOSS). The coefficients of LTA, LBSEG, LGSEG, FOREIGN, CATA, DE, and LOSS are
expected to be positive. We expect the coefficients of QUICK and ROI to be negative. Following
17
Francis et al. (2005) and Fung et al. (2012), we include audit opinion (OPINION) which is a
client risk measure and may also measure the need for additional work, and a variable indicating
a December fiscal year-end (YE), which may also capture a difference in audit costs, hence fees.
Extant literature has shown that Big audit firms earn fee premiums (Hay et al. 2006) and thus we
also control for Big audit firms (BIG) in the regression. We expect positive coefficients for
OPINION, YE, and BIG. Finally, indicators for year, industry effects, and MSA effects (e.g. to
control for different cost levels across MSAs) are included in all tests.
3.2 Test variable
An audit market is defined as a two-digit client SIC industry in a U.S. Metropolitan
Statistical Area (MSA, U.S. Census Bureau definition). Our basic test variable is:
DIFFERENCE = [(The sum of the largest auditor’s audit fees in a market – The sum of this
auditor’s audit fees in a market) ÷ Total audit fees in a market].
The variable DIFFERENCE captures the competitive disadvantage of other auditors
relative to the largest auditor in a specific market, and we expect β13 < 0. Note that the variable
DIFFERENCE is not just a measure of differential pricing by the dominant audit firm(s) in a
market but measures the pricing power of all suppliers in a market, relative to the largest
supplier. This measure can therefore distinguish between the two conceptual “stories”
concerning the nature of audit fee premiums to dominant firms, namely, are these payments for
higher audit quality or some form of economic rent (quasi-rents or monopoly rents)? If audit fees
decrease as the relative size of a supplier decreases, then the declining fee premium either
measures market power (our interpretation) or there are multiple audit quality levels available in
markets and these quality levels correspond to the market shares of auditors. That is, the largest
18
supplier sells the highest quality level, the second largest supplier sells the second highest level,
and so forth down to the lowest quality level sold by the smallest supplier. Obviously, a
differential audit quality interpretation of the variable DIFFERENCE is highly implausible as
there is no economic reason why varying audit quality levels offered for sale and the market
shares of the suppliers of these differential quality levels should be perfectly aligned!
4. Data and results
4.1 Sample and data
The sample is selected from two sources: the Audit Analytics database and the Compustat
data base. Panel A of Table 1 presents the sample screening procedures. We start with a total
number of observations of 99,800 with audit fee and MSA data for 2000-2011 on Audit
Analytics, and use this data to calculate the values of DIFFERENCE. We then merge the data
with the Compustat variables. We lose 37,304 observations in the merging process. We then
exclude companies in the financial sector, because the audit fee model for these firms are
different from other industries due to their special characteristics (Fields, Fraser, and Wilkins
2004; Kanagaretnam, Krishnan, and Lobo 2010). Furthermore, 5,166 observations do not have
values for all control variables. To focus on the issues of interest in this study, we exclude audit
markets that have only one active audit firm. 5 Recall that the transaction costs and resulting rents
and quasi-rents in the pricing model describe continuing audit engagements, so we exclude 5,223
observations with one or two year auditor tenure. The final sample consists of 26,876 firm-year
observations.
Descriptive statistics for all variables we use are reported in panel B of Table 1, and the
correlations between variables are reported in Panel C of Table 1. Note that the variables listed
5
Similar results are obtained when audit markets with only two or fewer auditors are excluded.
19
after DIFFERENCE in Panel B of Table 1 are defined and explained later in the paper. The data
presented in these tables are comparable to those reported in previous studies (e.g., Fung et al.
2012). Four ratio variables (QUICK, CATA, LEV and ROI) are winsorized in both extreme ends
to 0.5 percentile level. 6
4.2 Basic test results
Our basic results for the test of the differential pricing hypothesis are shown in Column 1
of Table 2 where the estimated coefficient of the DIFFERENCE variable is negative (as
hypothesized) and statistically significant. This suggests that there is an “audit firm effect” such
that small audit firms have limited pricing power relative to larger competitor firms.
All estimated coefficients of the control variables in the regression are also statistically
significant and have the expected sign. Moreover the estimated value of the coefficient of the
natural log of total assets (LTA) is expected to be about 0.5 when U.S. data is used in estimating
the audit fee model (Simunic, 1980), and its value is approximately 0.47. Also as expected, the
regression model has high explanatory power (R2 ≈ 0.86).
These results show that the relative size of the auditors in a market appears to be an
important, previously unrecognized determinant of audit fees, and that, ceteris paribus, audit fees
are not uniform across suppliers, even in the absence of any audit quality differences.
4.3 Additional tests
4.3.1
Effect of ‘Specialists’ in an audit market
As noted earlier, our hypothesis test provides a basis for distinguishing between two
underlying conceptual reasons why audit fees in a local market are highest for the dominant audit
6
The main results are not changed if the data are not winsorized.
20
firm. We argue that audit fees increase because of differences in supplier pricing power.
However, the auditor specialization, quality differentiation literature that originates with the
paper by Craswell et al. (1995) would interpret the differences as a premium (discount) for
higher (lower) audit quality. For the results of our hypothesis test to be consistent with a “quality
differentiation” story, there must be multiple levels of auditor specialization and industry
expertise sold in local client-industry defined markets, and the lower an audit firm’s market
share, the lower must be its level of audit quality. To our knowledge, no one has argued that this
rather unlikely situation is, in fact, the case. Rather, that literature has argued that there are some
client industry specialist auditors, whose services command a pricing premium relative to nonspecialist auditors; that is, client industry specific markets have two levels of audit quality
available – specialist quality and non-specialist audit quality.
We conduct two tests to investigate whether the differential fees captured by our
competition measure, DIFFERENCE, reflect the fee premium of quality differentiated specialists
relative to non-specialists, and whether the classification of specialists and non-specialists is
dichotomous. In the first test, we exclude the dominant auditor (specialist) in each audit market.
If the remaining auditors who are non-specialists have similar quality or competitive power, the
coefficient of DIFFERENCE should not be significant. The second test is to add an indicator
variable, SPECIALIST, in our regression model. We expect that the variable will fully capture
the effect of DIFFERENCE if there are only two types of auditors in a market (i.e., specialist vs
non-specialists and the coefficient of SPECIALIST should be positive). 7
7
By definition, SPECIALIST and DIFFERENCE must be highly correlated. SPECIALIST is a measurement of the
largest auditor’s dominance relative to all other auditors while treating all other auditors equally. SPECIALIST
captures the average fee premium between the largest auditor and all other auditors. In contrast, DIFFERENCE
captures the differences between the SPECIALIST and all other auditors individually. Our test that excludes all
specialists fully addresses the multicollinearity issue between SPECIALIST and DIFFERENCE.
21
SPECIALIST = an indicator variable that equals 1 if an auditor has the highest
market share in a local client industry market, 0 otherwise.
Column (2) of Table 2 presents the results of our main analysis when all specialists and
their clients are excluded from the analysis. The overall regression results using these nonspecialist observations are similar to those obtained using the full sample, and the variable
DIFFERENCE is significant (-0.331, t=-16.30). Column (3) of Table 2 reports full sample results
when SPECIALIST is included in the audit fee model along with DIFFERENCE and the control
variables. Note that the estimated coefficient of DIFFERENCE is unchanged relative to column
1 and the estimated coefficient of SPECIALIST is essentially zero. 8 These results suggest that
the variable SPECIALIST is not a good measure of the relative market power of the dominant
auditor office relative to every other auditor office, and when used alone probably does not
capture a quality differentiation price premium.
4.3.2
Effect of auditor size outside the client-industry local market
Given that auditor dominance within a client-industry local MSA market appears to be
associated with pricing power rather than quality differentiation, it is useful to investigate how (if
at all) prices are affected by the size of auditors outside of the client-industry local MSA market
but still within the same MSA. To do this, we construct a variable SUPERSIZE, where:
SUPERSIZE = an indicator variable that equals 1 if the auditor is the largest both
within the local client-industry market and also the largest for all other client
industries within an MSA.
8
If the variable DIFFERENCE is excluded from the regression while the variable SPECIALIST is included, the
estimated coefficient of SPECIALIST is positive and statistically significant, which replicates “specialist” results
normally reported in the literature. See column 1 of Table 3.
22
Note from the descriptive statistics in Table 1, that 41.1% of the observations (clientyears) in the sample use SPECIALIST auditors, while 21.7% of the observations use
SUPERSIZE auditors. The pricing results are shown in Table 3 where column 1 documents the
usual finding that specialist auditors command higher audit fees. However, when SUPERSIZE is
added to the regression, the coefficients of both SPECIALIST and SUPERSIZE are positive and
significant, suggesting that the fee premiums charged by dominant auditors incrementally
increase, if the auditor is largest in both the local MSA client-industry market and also the largest
in that local area (i.e. MSA). Finally, when DIFFERENCE is included in the analysis (column 3)
the SPECIALIST effect disappears, the SUPERSIZE effect remains, and (as in Table 2) the
coefficient of DIFFERENCE is negative and significant. Overall, these results suggest that
overall auditor size matters in audit pricing, and that the pricing effects reflect market power
rather than audit quality differences since the variables SUPERSIZE and DIFFERENCE simply
measure (absolute and relative) audit firm size and are unrelated to audit quality.
To obtain further insight as to how differential auditor size outside a client-industry local
market affects audit pricing within a market, we next construct three variables to determine if the
estimated coefficient of DIFFERENCE varies with auditor dominance outside a market. These
are:
MSA_DOMINANCE = DIFFERENCE if an MSA-client industry market leader
also has the largest other operations in the MSA, zero otherwise.
IND_DOMINANCE = DIFFERENCE if an MSA-client industry market leader
has the largest client-industry operations outside the local MSA, zero
otherwise.
23
NATION_DOMINANCE = DIFFERENCE if an MSA-client industry market
leader has the largest operations outside the local MSA, zero otherwise.
The basic idea motivating these variables is that the transaction costs of auditor change
and the cost (pricing) advantage of the dominant firm (and the disadvantages of smaller audit
firms) in a local client-industry market may depend not only on the dominant firm’s size within a
local client industry market, but also its dominance outside of that market. Say, for example, that
Ernst & Young is the dominant auditor of wineries headquartered in the San Francisco Bay area
MSA, then the variable DIFFERENCE captures the transaction cost (pricing) disadvantage of all
other incumbent audit firms of San Francisco Bay area winery clients. But if Ernst & Young is
also the dominant auditor of all other types of Bay area clients (MSA_DOMINANCE = 1 x
DIFFERENCE) then this may confer a further cost advantage on Ernst & Young in potentially
taking over the winery clients of other audit firms. If this is true, then the estimated coefficient of
MSA_DOMINANCE will be negative. Furthermore, if Ernst & Young is the dominant auditor of
winery clients throughout the United States (IND_DOMINANCE = 1 x DIFFERENCE) then this
fact may also enhance E & Y’s cost (pricing) advantage of Bay area winery clients. Finally, if
E& Y has the largest operations outside of the Fan Francisco Bay area irrespective of client
industry (NATION_DOMINANCE = DIFFERENCE x 1) then this may also reduce the
transaction costs and therefore the threat of potential takeover of Bay area winery clients of other
auditors.
Turning to results, note from Table 1 that 10.2% of auditors who are dominant in a clientindustry local MSA market are simultaneously the dominant audit firms of all types of clients in
that MSA (n = 2741). However, they are rarely also the dominant audit firm for all clients in the
2-digit SIC industry (n = 54) or the dominant firm throughout the country (n = 108). Table 4
24
displays the results of including DIFFERENCE along with the three ‘dominance’ variables in
our analysis. Note that all estimated coefficients have the (expected) negative signs and are
weakly statistically significant. Overall, these results suggest that auditor dominance and the
relative disadvantage of non-dominant audit firms is largely a local MSA, client industry
phenomenon, with the size of an audit firm’s other operations playing a role – but not a critical
role – in the determination of the transaction costs of auditor change and audit pricing.
4.3.3
Replication of auditor ‘Location’ test reported by Numan and Willekens (2012)
As mentioned earlier, in a recent paper, Numan and Willekens report a test of auditor
pricing as a function of an audit firm’s location in space relative to its clients and its competitors.
The test is motivated by the spatial competition model developed in Chan, Ferguson, Simunic
and Stokes (2004). That paper models a non-cooperative oligopoly where clients and suppliers
are located in a product-characteristics space (which could include physical space as in Hotelling
1929) and the supplier located “closest” to a client enjoys a cost advantage in supplying the audit
service relative to competitors. Holding audit quality constant, a client is motivated to purchase
an audit from the closest (cheapest) supplier. Moreover, that supplier’s price is limited by how
close the nearest alternative supplier is to the client. Thus the least-cost (closest) supplier
becomes the incumbent audit firm but does not price the audit at its own marginal (avoidable)
cost but rather will use a limit pricing strategy to deter entry and price the audit at the cost facing
the next-closest supplier. As a result, the incumbent audit firm will earn rents (quasi-rents and
possibly monopoly rents).
While the spatial model of auditor competition is conceptually quite appealing,
operationalizing the notions of “space” and “distance” to test the pricing implications of the
25
model is a challenge. In their paper, Numan and Willekens define distance in space by the
difference in market shares of suppliers, where markets are defined the same way as in our paper
(i.e. client industries in U.S. MSAs), and market shares are also measured the same way (specific
audit firm’s audit fees relative to total audit fees in a market). They argue that the greatest pricing
pressure comes from the supplier whose market share is “closest” (i.e. most similar value) to the
incumbent audit firm. They hypothesize that audit fees increase as the difference or distance
between the incumbent’s market share and the closest non-incumbent’s market share increases,
whether the non-incumbent’s share is smaller or larger than the incumbent’s share. They use the
absolute value of this difference to compute a variable called DISTANCE and find that it has a
statistically significant positive coefficient. That is, the greater the DISTANCE the greater is the
pricing power of the incumbent audit firm.
Since we argue that the greatest pricing pressure in a market comes from the largest audit
firm in the market – not from smaller firms – we believe that the variable DISTANCE is not a
good measure of competitive pressure on the incumbent audit firm. If the incumbent firm is a
small firm, the greater the distance to its closest competitor (larger firm), the lower the fee it can
charge. We believe that the reason a positive coefficient on DISTANCE was observed in Numan
and Willekens is because it is an average effect, as most of the clients in their sample utilize Big
audit firms.
Thus it is interesting to determine whether Numan and Willekens’ reported findings are
robust after controlling for our competition measure, DIFFERENCE. The results of this test are
shown in Table 5 where the variables DISTANCE, as well as Numan and Willekens’ variable
termed PORTFOLIO (the proportion of revenues an audit firm generates from an SIC industry
relative to its total client revenues earned in an MSA), and DISTANCE_PORTFOLIO (i.e.,
26
DISTANCE × PORTFOLIO) are included in the regression. Column (1) replicates their results;
column (2) shows results after controlling for DIFFERENCE; column (3) adds SPECIALIST to
their model; while column (4) includes SECIALIST, the Numan and Willekens variables, and
DIFFERENCE. Note that Table 5 presents results using a sample of only Big 4 audit firms and
the years 2005 and 2006, to more closely approximate the sample used in Numan and
Willekens. 9
Column (1) shows that the coefficient of variable DISTANCE is positive and statistically
significant as reported by Numan and Willekens. But once we include DIFFERENCE, our
measurement for competition, the coefficient of the variable DISTANCE becomes statistically
insignificant in columns (2) and (4). Note also, that simply controlling for SPECIALIST (in
column (3), does not affect their findings. However, our test variable DIFFERENCE has
consistent coefficient estimates as those shown in Table 2, suggesting that this measure is more
reliable for capturing competition in a market and supports our argument that the most
competitive pressure comes from the dominant auditor office in a market. 10
In their analyses, Numan & Willekens include the Herfindahl index and find that, contrary
to normal expectations, average prices are lower in more concentrated markets as the estimated
coefficient of Herfindahl in their paper negative. The same result is seen in Table 5, where:
HERFINDAHL = Herfindahl concentration index for each audit market, where
the Herfindahl index is calculated as the sum of si 2 and i is an audit office in
an audit market
and s is a market share of an office in a local client-
industry market based on audit fees.
9
Using our full sample yields similar results (not shown).
Though smaller clients may not engage large audit firms, they always have the option of doing so and the
incumbent small audit firm must recognize this possibility in pricing audits.
10
27
Note, however, that in columns (2) and (4) which include the DIFFERENCE variable, the
coefficient of HERFINDAHL becomes insignificant. As discussed in Numan and Willekens, this
implies that measures of overall industry concentration do not explain the pricing of audit
services where prices vary with both clients and suppliers, an effect which is incorporated into
our measure of local market competition (i.e. DIFFFERENCE).
Finally, the analysis reported in Table 6 further investigates the validity of Neuman and
Willelkens’ DISTANCE measure using our full sample. Column (1) of Table 6 replicates Numan
and Willekens’ basic finding that audit fees increase with the distance of the incumbent auditor
from the closest supplier. 11 However, we argue that it is the largest supplier – not suppliers that
are smaller than the incumbent auditor – who constrain the pricing power of the incumbent.
Therefore, we separate the sample into cases where the closest supplier (in terms of market
share) is smaller than the incumbent auditor (n = 19,624) and cases where the closest supplier is
larger than the incumbent auditor (n = 7,252). The results are shown in columns (2) and (3). Note
that when the closest alternative supplier is smaller than the incumbent, the incumbent has
pricing power, charges a higher audit fee, and this fee increases in DISTANCE. However, when
the closest alternative supplier is larger, the incumbent lacks pricing power and audit fees
decrease in distance (i.e. the smaller is the incumbent auditor the lower the audit fee). Colum (4)
includes both positive and negative distance variables and demonstrates that these differential
pricing results continue to hold. These results imply that the Numan and Willekens’ spatial
completion explanation for the positive coefficient of DISTANCE is invalid, and the results
reported in column (1) are simply an average effect that happens to be consistent with their
argument.
11
We do not include variable HERFINDAHL in the tests because we do not think this variable can be easily
interpreted. However, including this variable does not affect the results on all other variables.
28
5. Summary and Conclusions
In this paper, we develop a measure of the degree of audit firm competition in audit
markets based on a theory of transaction costs associated with auditor change. We argue that the
transaction costs of changing auditors are a function of both the incumbent audit firm’s size and
the size of the largest audit firm operating in a market. For a given client, the lowest transaction
costs occur if that client switches to the largest auditor office operating in the market. We argue
that the competitive pressure on an incumbent auditor’s fees basically depends upon the ease
with which the auditor’s clients can potentially switch to a competing auditor. If the transaction
costs of switching auditors are low, the ability of the incumbent auditor to extract economic rents
– either monopoly rents or quasi-rents - from its clients is limited; conversely, high costs of
auditor change give an incumbent auditor greater pricing power. That is, the size of transaction
costs of auditor change affects the relative competitive position of suppliers. We posit that the
larger is an auditor’s operation in a market, the lower are its transaction costs of taking on new
business. As a result, the largest auditor can put the greatest competitive pressure on all other
auditors operating in a market, and the pressure is a function of the size difference between the
largest auditor’s operations and the size of the incumbent audit firm’s operations in a market.
Note that our arguments only concern the relative pricing power of auditors and no actual auditor
changes are expected to occur if audits are “properly priced”.
Based on these arguments, we develop a variable termed DIFFERENCE, which is a
measure of the size difference between the largest auditor’s operations in a market and
incumbent auditor’s size. We predict that audit fees decrease as the size difference between
dominant auditor in a market and incumbent auditor increases.
29
We test our pricing hypothesis using U.S. data for the years 2000-2011 classified by 2-digit
Standard Industrial Classification client industries for Metropolitan Statistical Areas. Our basic
measure of the size of an audit firm’s operations in a market is the market share of an auditor’s
office, where market share is the ratio of an office’s audit fees for clients operating in a 2-digit
SIC industry to the total audit fees of all clients in a 2-digit industry located in a Metropolitan
Statistical Area.
Using the standard audit fee model that has been well calibrated in previous research, we
find strong support for our basic prediction that audit fees depend upon the relative competitive
position of the incumbent supplier. This finding is robust across various sensitivity tests.
Moreover, if we include the usual measure of auditor specialization (i.e. the market share of the
dominant audit firm operating in a local client 2-digit SIC industry) with our test variable, our
results are unchanged while the specialist measure does not have a statistically significant
positive coefficient. This result indicates that competitive pressure is a continuous function of the
auditor’s relative size, and that the traditional dichotomous variable (SPECIALIST) cannot
accurately capture this effect.
In conclusion, variations in the transaction costs of changing auditors for the clients of
incumbent audit firms appear to be an important, hitherto unrecognized determinant of the
pricing of audit services. This is not too surprising since these transaction costs of changing
auditors are an important element of DeAngelo (1981) multi-period audit pricing model. While
the implications of that model have been used to investigate the phenomenon of “lowball”
pricing in the first (early) period(s) of an audit engagement, to our knowledge, the implications
of the model for the pricing of continuing audit engagements where the level of transaction costs
30
determine the level of quasi-rents, and perhaps monopoly rents, an auditor can incorporate into
audit fees has not been investigated.
Finally, our arguments and evidence challenge the usual view that client industry market
shares are appropriate measures of quality differentiated auditors, a perspective that has
dominated the literature for the last 18 years. Whether or not there are systematic industry
specific audit quality differences across audit firms remains to us an open question. One
possibility is that all audit firms who operate in a client market are quality differentiated, relative
to audit firms that do not operate in the market. Another possibility is that audit firms that have
market power in a particular context may also perform a higher quality audit in that context,
since a client’s threat to terminate an audit relationship if the auditor “looks too hard” or fails to
report as the client wishes is less credible (not credible) if the client has few (no) alternative
supplier choices. In other words, these audit firms have bargaining power over the clients. Thus
higher market power could be correlated with both a higher audit fee and a higher quality audit.
However, such higher audit quality and higher audit fee would be client specific and depend on
the competitive context. We leave the further investigation of these interesting possibilities to
future research.
31
References
Bandyopadhyay, S., Kao, J., 2004. Market structure and audit fees: a local analysis.
Contemporary Accounting Research 21 (3), 529–561.
Cadman, Brian, Michael T. Stein, 2007. Industry Specialization and Auditor Quality in U.S.
Markets, Working paper, Northwestern University and Old Dominion University.
Chan, D., Ferguson, A., Simunic, D. A., Stokes, D., 2004. A Spatial analysis and test of
oligopolistic competition in the market for audit services. Working paper, University of
British Columbia.
Craswell, A., Francis, J., Taylor, S., 1995. Auditor brand name reputation and industry
specialization. Journal of Accounting and Economics 20 (3), 297–322.
Dao, M., Raghunandan, K., Rama, D. V., 2012. Shareholder voting on auditor selection, audit
fees, and audit quality. The Accounting Review 87 (1), 149–171.
DeAngelo, L. 1981. Auditor independence, ‘low balling’, and disclosure regulation. Journal of
Accounting and Economics 3 (2), 113-127.
Dopuch, N., Simunic, D. A., 1980. The nature of competition in the auditing profession: A
descriptive and normative view. Regulation and the Accounting Profession, edited by J.
Buckley and F. Weston, Lifetime Learning Publications, Belmont, CA.
Feldman, E.R., 2006. A basic quantification of the competitive implications of the demise of
Arthur Anderson. Review of Industrial Organization 29 (3), 193–212.
Ferguson, A., Francis, J., Stokes, D., 2003. The effect of firm-wide and office-level industry
expertise on audit pricing. The Accounting Review 78 (2), 429–448.
32
Ferguson, A., Stokes, D., 2002. Brand name audit pricing, industry specialization and industry
leadership premiums post Big 8 and Big 6 mergers. Contemporary Accounting Research
19 (1), 77–110.
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 (1), 53-77.
Francis, J., Reichelt, K., Wang, D., 2005. The pricing of national and city-specific reputations for
industry expertise in the U.S. audit market. The Accounting Review 80 (1), 113–36.
Francis, J. R., Yu, M. D., 2009. Big 4 office size and audit quality. The Accounting Review 84 (5),
1521–1552.
Fung, S. Y. K., Gul, F. A., Krishnan, J., 2012. City-level auditor industry specialization,
economies of scale, and audit pricing. The Accounting Review 87 (4), 1281-1307.
Government Accountability Office (GAO), 2003. Mandated study on consolidation and
competition. Report 03-864. Washington, D.C.: Government Printing Office.
Government Accountability Office (GAO), 2008. Continued concentration in audit market for
large public companies does not call for immediate action. Report 08-163. Washington,
D.C.: Government Printing Office.
Hay, D. C., Knechel, W. R., Wong, N., 2006. Audit fees: a meta-analysis of the effect of supply
and demand attributes. Contemporary Accounting Research 23 (1), 141–91.
Hotelling, H. 1929. Stability in competition. The Economic Journal 39 (153), 623-40.
Kanagaretnam, K., G. V. Krishnan, and G. J. Lobo. 2010. An empirical analysis of auditor
independence in the banking industry. The Accounting Review 85 (6), 2011-2046.
Magee, R., M. Tseng, 1990. Audit pricing and auditor independence, The Accounting Review 65
(2), 315-336.
33
Minutti-Meza, M. 2013. Does auditor industry specialization improve audit quality? Journal of
Accounting Research 51 (4), 779-817.
Numan, W., Willekens, M., 2012. An empirical test of spatial competition in the audit market.
Journal of Accounting and Economics 53 (1-2), 450-465.
O’Keefe, T., Simunic, D. A., Stein, M. T., 1994. The production of audit services: Evidence
from a major public accounting firm. Journal of Accounting Research 32 (2), 241-261.
Pearson, T., Trompeter, G., 1994. Competition in the market for audit services: The effect of
supplier concentration on audit fees. Contemporary Accounting Research 11 (1), 115-135.
Sabac, F., D.A. Simunic, 2001. A rational expectations model of auditor change: Tenure, pricing,
and auditor independence, Working paper, University of Alberta and UBC.
Simunic, D. A. 1980. The pricing of audit services: Theory and evidence. Journal of Accounting
Research 18 (1), 161–190.
Subcommittee on Reports, Accounting and Management of the Commission on Government
Operations, United States Senate. 1977. The Accounting Establishment: A Staff Study,
Washington, D.C., Government Printing Office.
34
Appendix Variable Definitions
LAF = natural log of audit fees;
LTA = natural log of total assets;
LBSEG = natural log of the number of unique business segments;
LGSEG = natural log of the number of unique geographic segments;
CATA = ratio of current assets to total assets;
QUICK = ratio of current assets excluding inventory to current liabilities;
LEV = ratio of long-term debt to total assets;
ROI = ratio of earnings before interest and tax to total assets;
FOREIGN = an indicator variable that equals 1 if revenue from foreign operation is reported,
0 otherwise;
OPINION = an indicator variable coded 1 for modified audit report, and zero otherwise;
YE = an indicator variable that equals 1 for December 31 year-end and 0 otherwise;
LOSS = an indicator variable that equals 1 if loss in current year, and 0 otherwise;
BIG = an indicator variable that equals 1 for Big N auditors and 0 otherwise;
DIFFERENCE = [(The sum of the largest auditor’s audit fees in a market – The sum of this
auditor’s audit fees in a market)/Total audit fees in a market]. An audit market is defined as a
two-digit SIC industry in a U.S. Metropolitan Statistical Area (MSA, U.S. Census Bureau
definition);
SPECIALIST = an indicator variable that equals 1 if the auditor is the largest in the MASIndustry market, and 0 otherwise;
SUPERSIZE = an indicator variable that equals 1 if the auditor is the largest in the MASIndustry market and the largest outside the MAS_Industry market, and 0 otherwise;
35
MSA_DOMINANCE = equals DIFFERENCE if an MAS-Industry market leader has the largest
other operation, zero otherwise.
IND_DOMINANCE = equals DIFFERENCE if an MAS-Industry market leader has the largest
industry operations outside the MAS-Industry market, zero otherwise.
NATION_DOMINANCE = equals DIFFERENCE if an MAS-Industry market leader has the
largest operations outside the MAS market, zero otherwise.
HERFINDAHL = Herfindahl concentration index per audit market, where the Herfindahl index
is calculated as the sum of si2, where i is an audit office in an audit market and s is market share
in an audit market based on audit fees. An audit market is defined as a two-digit SIC industry in
a U.S. Metropolitan Statistical Area (MSA, U.S. Census Bureau definition);
DISTANCE = smallest absolute fee market share difference between the incumbent auditor and
his closest competitor in an audit market. An audit market is defined as a two-digit SIC industry
in a U.S. Metropolitan Statistical Area (MSA, U.S. Census Bureau definition);
DISTANCE_smaller=DISTANCE if the nearest neighbor is smaller than the incumbent auditor.
DISTANCE_smaller=0 if the nearest neighbor is larger than the incumbent auditor.
DISTANCE_larger=DISTANCE if the nearest neighbor is larger than the incumbent auditor.
DISTANCE_larger=0 if the nearest neighbor is larger than the incumbent auditor.
PORTFOLIO= the relative revenue share an audit firm generates in a 2-digit SIC industry
relative to the total revenue generated by an audit firm in an MSA.
.
36
Table 1 Description of Data
Panel A: Sample Selection
Observations in USA with audit fee and MSA data for 2000-2011 on Audit
Analytics
Less:
Observations not on Compustat
(37,304)
Financial Sector (SIC 6000-6999)
Missing control variables
Audit market with only one auditor
Audit engagements in the first or second year
(17,637)
(5,166)
(7,594)
(5,223)
Final Sample (sample for regression analysis)
26,876
99,800
Panel B: Descriptive Statistics (N=26,876)
Variable
Audit_Fees ($)
Total Assets ($mil)
LAF
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
BIG
DIFFERENCE
DISTANCE
PORTFOLIO
HERFINDAHL
SPECIALIST
SUPERSIZE
MSA_DOMINANCE
IND_DOMINANCE
NATION_DOMINANCE
Mean
1,468,094
2,870
13.221
5.695
0.445
0.606
0.510
2.673
0.259
-0.010
0.413
0.056
0.706
0.406
0.770
0.233
0.239
0.178
0.475
0.411
0.217
0.102
0.002
0.004
Median
553,125
286
13.223
5.657
0.000
0.000
0.512
1.523
0.178
0.054
0.000
0.000
1.000
0.000
1.000
0.127
0.100
0.107
0.430
0.000
0.000
0.000
0.000
0.000
Standard
Deviation
3,112,548
11,499
1.392
2.213
0.628
0.704
0.263
3.732
0.279
0.222
0.492
0.229
0.456
0.491
0.421
0.265
0.282
0.203
0.197
0.492
0.412
0.221
0.033
0.049
P25
195,619
63
12.184
4.149
0.000
0.000
0.295
0.932
0.035
-0.053
0.000
0.000
0.000
0.000
1.000
0.000
0.024
0.038
0.315
0.000
0.000
0.000
0.000
0.000
P75
1,424,495
1,400
14.169
7.244
1.099
1.099
0.722
2.836
0.407
0.108
1.000
0.000
1.000
1.000
1.000
0.426
0.386
0.237
0.592
1.000
0.000
0.000
0.000
0.000
37
Panel C: Spearman Correlation (Bold denotes significant at 5% or lower)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
LAF
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
BIG
DIFFERENCE
DISTANCE
PORTFOLIO
HERFINDAHL
SPECIALIST
SUPERSIZE
MSA_DOMINANCE
IND_DOMINANCE
NATION_DOMINANCE
1
1.00
0.82
0.31
0.43
-0.26
-0.15
0.37
0.33
0.46
-0.20
0.06
-0.28
0.46
-0.41
0.30
0.09
0.05
0.33
0.17
-0.25
-0.06
-0.05
2
1.00
0.33
0.32
-0.45
-0.20
0.49
0.43
0.35
-0.27
0.07
-0.38
0.52
-0.41
0.34
0.04
0.09
0.34
0.18
-0.24
-0.05
-0.03
3
1.00
0.18
-0.19
-0.14
0.19
0.20
0.14
-0.11
0.01
-0.19
0.13
-0.11
0.13
-0.03
0.08
0.10
0.06
-0.05
0.00
-0.02
4
1.00
0.09
0.09
-0.03
0.19
0.60
-0.14
-0.05
-0.14
0.24
-0.17
0.08
0.11
-0.07
0.12
0.01
-0.10
-0.04
-0.03
5
1.00
0.52
-0.57
-0.18
0.04
-0.04
-0.13
0.16
-0.09
0.12
-0.12
-0.04
-0.08
-0.10
-0.07
0.07
-0.02
0.00
6
1.00
-0.43
-0.10
0.04
-0.22
0.00
0.06
0.06
0.03
-0.09
0.01
-0.13
-0.05
-0.04
0.01
-0.01
0.00
7
1.00
0.15
0.04
-0.03
0.12
-0.10
0.19
-0.18
0.22
-0.01
0.17
0.17
0.13
-0.09
-0.02
-0.01
8
1.00
0.21
-0.26
-0.06
-0.74
0.17
-0.12
0.14
-0.08
0.08
0.11
0.06
-0.08
-0.01
0.00
9
1.00
-0.13
-0.06
-0.16
0.22
-0.17
0.09
0.08
-0.06
0.12
0.03
-0.12
-0.02
-0.02
10
11
12
13
14
15
16
17
18
19
1.00
0.00
0.24
-0.23
0.13
-0.12
0.03
-0.01
-0.09
-0.05
0.08
0.03
0.02
1.00
0.04
0.07
-0.05
0.05
0.05
0.04
0.05
0.04
-0.02
0.00
-0.01
1.00
-0.15
0.11
-0.13
0.05
-0.08
-0.10
-0.05
0.07
0.01
0.00
1.00
-0.45
0.38
-0.23
0.07
0.36
0.21
-0.28
-0.06
-0.04
1.00
-0.53
-0.19
0.02
-0.88
-0.56
0.51
0.05
0.08
1.00
0.05
0.59
0.61
0.46
-0.20
-0.01
0.00
1.00
-0.09
0.16
-0.04
-0.04
-0.02
-0.02
1.00
0.26
0.31
0.07
0.02
0.04
1.00
0.63
-0.46
-0.06
-0.07
1.00
-0.29
-0.03
-0.04
20
1.00
0.05
0.04
38
21
1.00
0.16
Table 2
Relationship between Audit Fees in an MSA-Industry Market and Auditors’ Market Positions
VARIABLES
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
BIG
DIFFERENCE
(1)
LAF
(2)
LAF
(3)
LAF
0.468***
(16.53)
0.104***
(18.83)
0.128***
(20.81)
0.415***
(20.37)
-0.034***
(-28.75)
0.096***
(6.42)
-0.253***
(-11.68)
0.240***
(27.94)
0.119***
(6.79)
0.101***
(13.71)
0.088***
(9.88)
0.317***
(29.19)
-0.340***
(-24.52)
0.384***
(8.29)
0.103***
(13.77)
0.121***
(15.39)
0.399***
(15.86)
-0.033***
(-24.07)
0.065***
(3.55)
-0.234***
(-9.01)
0.208***
(19.11)
0.097***
(4.86)
0.095***
(10.24)
0.076***
(6.87)
0.328***
(27.02)
-0.331***
(-16.30)
9.521***
(108.40)
8.687***
(65.00)
0.468***
(16.53)
0.104***
(18.83)
0.128***
(20.81)
0.415***
(20.38)
-0.034***
(-28.74)
0.096***
(6.42)
-0.253***
(-11.68)
0.240***
(27.93)
0.119***
(6.79)
0.101***
(13.71)
0.088***
(9.88)
0.317***
(29.20)
-0.343***
(-17.67)
-0.002
(-0.22)
9.523***
(107.49)
26,876
0.866
YES
YES
YES
Full
15,828
0.850
YES
YES
YES
Only non-SPECIALISTs
26,876
0.866
YES
YES
YES
Full
SPECIALIST
Constant
N
Adjusted R2
MSA FE
Industry FE
Year FE
Sample
39
This table presents results from three basic regressions. In the first regression, only variable
DIFFERENCE is included, in addition to all control variables. The second regression is the same
as first regression on the subsample excluding the specialists. The third regression includes both
DIFFERENCE and SPECIALIST. Variable descriptions are presented in the Appendix. *, **,
and *** represent significance at the 10%, 5%, and 1% levels, respectively, using two-tailed tests.
Robust t-statistics are presented in parentheses.
40
Table 3
Relationship between Audit Fees in an MSA-Industry Market and Auditors’ Operation Size
VARIABLES
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
BIG
SPECIALIST
(1)
LAF
(2)
LAF
(3)
LAF
0.477***
(16.83)
0.104***
(18.79)
0.130***
(21.00)
0.416***
(20.30)
-0.035***
(-28.88)
0.098***
(6.48)
-0.266***
(-12.22)
0.239***
(27.64)
0.118***
(6.70)
0.101***
(13.63)
0.090***
(9.98)
0.352***
(32.52)
0.117***
(16.49)
0.479***
(16.91)
0.104***
(18.78)
0.130***
(21.09)
0.416***
(20.31)
-0.035***
(-28.90)
0.097***
(6.44)
-0.266***
(-12.20)
0.239***
(27.65)
0.118***
(6.72)
0.101***
(13.62)
0.090***
(9.99)
0.351***
(32.44)
0.101***
(11.22)
0.031***
(3.01)
9.275***
(101.72)
9.254***
(101.93)
0.469***
(16.60)
0.104***
(18.82)
0.128***
(20.88)
0.415***
(20.38)
-0.034***
(-28.75)
0.096***
(6.39)
-0.253***
(-11.67)
0.240***
(27.94)
0.119***
(6.80)
0.101***
(13.70)
0.088***
(9.90)
0.316***
(29.15)
-0.014
(-1.26)
0.025**
(2.38)
-0.341***
(-17.56)
9.506***
(107.42)
26,876
0.864
YES
YES
YES
Full
26,876
0.864
YES
YES
YES
Full
26,876
0.866
YES
YES
YES
Full
SUPERSIZE
DIFFERENCE
Constant
N
Adjusted R2
MSA FE
Industry FE
Year FE
Sample
41
This table presents results from three regressions. In the first regression, only variable
SPECIALIST is included, in addition to all control variables. The second regression adds the
variable SUPERSIZE to the first regression. The third regression adds DIFFERENCE to the
second regression. Variable descriptions are presented in the Appendix. *, **, and *** represent
significance at the 10%, 5%, and 1% levels, respectively, using two-tailed tests. Robust tstatistics are presented in parentheses.
42
Table 4
Relationship between Audit Fees in an MSA-Industry Market and Dominance of the Largest
Auditor
VARIABLES
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
BIG
SPECIALIST
SUPERSIZE
DIFFERENCE
MSA_DOMINANCE
(1)
LAF
(2)
LAF
(3)
LAF
0.469***
(16.60)
0.104***
(18.81)
0.129***
(20.91)
0.416***
(20.39)
-0.034***
(-28.76)
0.095***
(6.37)
-0.253***
(-11.68)
0.240***
(27.92)
0.119***
(6.81)
0.101***
(13.72)
0.088***
(9.90)
0.315***
(28.91)
-0.012
(-1.04)
0.023**
(2.24)
-0.322***
(-14.92)
-0.036*
(-1.91)
0.469***
(16.58)
0.104***
(18.84)
0.129***
(20.89)
0.416***
(20.37)
-0.034***
(-28.75)
0.095***
(6.37)
-0.253***
(-11.66)
0.240***
(27.97)
0.119***
(6.80)
0.101***
(13.71)
0.088***
(9.88)
0.315***
(28.92)
-0.012
(-1.04)
0.023**
(2.23)
-0.320***
(-14.85)
-0.034*
(-1.80)
-0.272**
(-2.29)
9.510***
(107.26)
9.510***
(107.18)
0.469***
(16.57)
0.104***
(18.82)
0.129***
(20.89)
0.416***
(20.41)
-0.034***
(-28.77)
0.095***
(6.36)
-0.253***
(-11.67)
0.240***
(27.98)
0.119***
(6.81)
0.101***
(13.68)
0.088***
(9.87)
0.315***
(28.95)
-0.011
(-1.00)
0.023**
(2.21)
-0.317***
(-14.67)
-0.034*
(-1.81)
-0.218*
(-1.76)
-0.137*
(-1.90)
9.510***
(107.17)
IND_DOMINANCE
NATION_DOMINANCE
Constant
43
Observations
Adjusted R2
MSA FE
Industry FE
Year FE
Sample
26,876
0.866
YES
YES
YES
Full
26,876
0.866
YES
YES
YES
Full
26,876
0.866
YES
YES
YES
Full
This table presents results from three regressions. In the first regression, the variable
MSA_DOMINACE is included, in addition to all control variables. The second regression adds
the variable IND_DOMINACE to the first regression. The third regression adds
NATION_DOMINACE to the second regression. Variable descriptions are presented in the
Appendix. *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively,
using two-tailed tests. Robust t-statistics are presented in parentheses.
44
Table 5
Relationship between Audit Fees in an MSA-Industry Market and two Market Position
Measurements
VARIABLES
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
HERFINDAHL
DISTANCE
PORTFOLIO
DISTANCE_PORTFOLIO
(1)
LAF
(2)
LAF
(3)
LAF
(4)
LAF
0.461***
(5.88)
0.115***
(7.59)
0.130***
(7.06)
0.553***
(8.54)
-0.030***
(-5.28)
-0.007
(-0.15)
-0.509***
(-6.46)
0.252***
(9.97)
0.333***
(3.85)
0.067***
(3.13)
0.123***
(4.02)
-0.535***
(-6.13)
0.362***
(5.86)
0.307***
(2.96)
-0.178
(-0.85)
0.465***
(5.91)
0.116***
(7.70)
0.128***
(6.98)
0.543***
(8.43)
-0.030***
(-5.34)
-0.012
(-0.24)
-0.501***
(-6.38)
0.255***
(10.08)
0.330***
(3.79)
0.066***
(3.06)
0.124***
(4.08)
-0.173
(-1.56)
0.093
(1.17)
0.256**
(2.47)
-0.174
(-0.84)
-0.306***
(-5.47)
0.463***
(5.90)
0.117***
(7.68)
0.129***
(7.02)
0.550***
(8.49)
-0.030***
(-5.28)
-0.009
(-0.18)
-0.504***
(-6.41)
0.253***
(10.00)
0.332***
(3.82)
0.067***
(3.12)
0.123***
(4.03)
-0.452***
(-4.91)
0.248***
(3.37)
0.270**
(2.57)
-0.147
(-0.70)
0.464***
(5.91)
0.116***
(7.66)
0.128***
(6.98)
0.543***
(8.43)
-0.030***
(-5.35)
-0.012
(-0.25)
-0.502***
(-6.38)
0.255***
(10.09)
0.330***
(3.79)
0.066***
(3.05)
0.124***
(4.09)
-0.155
(-1.37)
0.098
(1.24)
0.262**
(2.52)
-0.184
(-0.88)
-0.347***
(-4.56)
-0.026
(-0.79)
11.591***
(43.01)
DIFFERENCE
SPECIALIST
Constant
11.568***
(46.50)
11.581***
(42.60)
0.071***
(2.94)
11.547***
(43.78)
45
N
Adjusted R2
Industry FE
Year FE
Sample
3,626
3,626
3,626
3,626
0.751
0.752
0.751
0.752
YES
YES
YES
YES
YES
YES
YES
YES
All four regressions include only observations of Big 4
auditors in 2005 and 2006
This table presents results from four regressions. The first and third regressions include
DISTANCE and other control variables in Numan and Willekens (2012). The first regression
does not include variable SPECIALIST while the third one includes it. The second and fourth
regressions add DIFFERENCE to first and third regressions respectively. Variable descriptions
are presented in the Appendix. *, **, and *** represent significance at the 10%, 5%, and 1%
levels, respectively, using two-tailed tests. Robust t-statistics are presented in parentheses.
46
Table 6
The Relationship between Audit Fees in an MSA-Industry Market by Distinguishing the Direction of the Nearest DISTANCE
VARIABLES
LTA
LBSEG
LGSEG
CATA
QUICK
LEV
ROI
FOREIGN
OPINION
YE
LOSS
BIG
DISTANCE
DISTANCE_smaller
(1)
LAF
(2)
LAF
(3)
LAF
(4)
LAF
0.504***
(17.75)
0.101***
(17.91)
0.132***
(20.61)
0.489***
(23.49)
-0.035***
(-28.25)
0.086***
(5.64)
-0.319***
(-14.33)
0.259***
(29.04)
0.149***
(8.38)
0.112***
(14.64)
0.096***
(10.41)
0.382***
(33.50)
0.058***
(3.35)
0.523***
(16.61)
0.105***
(16.21)
0.136***
(18.25)
0.491***
(20.12)
-0.035***
(-23.82)
0.107***
(6.12)
-0.324***
(-12.38)
0.262***
(24.87)
0.148***
(6.95)
0.121***
(13.30)
0.092***
(8.39)
0.361***
(25.92)
0.459***
(6.99)
0.086***
(7.61)
0.111***
(8.96)
0.465***
(11.71)
-0.036***
(-15.28)
0.022
(0.73)
-0.275***
(-6.55)
0.252***
(15.18)
0.135***
(4.23)
0.081***
(5.77)
0.097***
(5.80)
0.391***
(19.12)
0.502***
(17.65)
0.103***
(18.24)
0.130***
(20.51)
0.485***
(23.32)
-0.035***
(-28.48)
0.084***
(5.52)
-0.311***
(-14.01)
0.259***
(29.38)
0.145***
(8.14)
0.109***
(14.39)
0.093***
(10.11)
0.366***
(32.19)
0.162***
0.150***
47
(8.04)
0.231***
(8.90)
-0.133**
(-2.22)
9.008***
(100.02)
0.242***
(7.68)
-0.263***
(-3.87)
9.008***
(102.19)
-0.339***
(-10.39)
0.173***
(3.78)
0.110
(0.95)
9.199***
(49.56)
26,876
0.854
YES
YES
Full
19,624
0.856
YES
YES
Only smaller neighbours
7,252
0.843
YES
YES
Only larger neighbours
DISTANCE_larger
PORTFOLIO
DISTANCE_PORTFOLIO
Constant
N
Adjusted R2
Industry FE
Year FE
Sample
(8.34)
-0.268***
(-10.80)
0.220***
(8.53)
-0.202***
(-3.45)
9.070***
(107.07)
26,876
0.856
YES
YES
Full
This table presents results from four regressions. The first regression include DISTANCE and other control variables in Numan and
Willekens (2012) using the full sample. The second regression is on the subsample of closest neighbors who are smaller than the
auditor. The third regression is on the subsample of closest neighbors who are larger than the auditor. The fourth regression is on the
full sample while distinguishing the DISTANCE to the larger and smaller neighbors. Variable descriptions are presented in the
Appendix. *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively, using two-tailed tests. Robust t-statistics
are presented in parentheses.
48