Joumal of Accounting Research

Vol. 18 No. 1 Spring 1980

Printed m [Link].

The Pricing of Audit Services:

Theory and Evidence
DAN A. SIMUNIC*

1. Introduction
The question of the existence of competition among auditors has been
the subject of considerable discussion in recent years. More specifically,
the "Big Eight" firms as a group have been accused of monopolizing the
market for audits (Staff Study of the Subcommittee on Reports, Account-
ing and Management of the Senate Committee on Government Opera-
tions [1977]). However, evidence on the issue is scanty and typically
anecdotal (e.g., Bernstein [1978]). The evidence of the Staff Study itself
is limited to concentration statistics, with the allegations relying on what
has come to be called the "concentration doctrine" (Demsetz [1973]).
According to this doctrine, supplier concentration is a reliable indicator
of supplier behavior and performance. In this paper, I provide evidence
from a test of the hypothesis that price competition prevails throughout
the market for the audits of pubUcly held companies, irrespective of the
share of a market segment which is serviced by the Big Eight firms. The
evidence is based on an examination of a sample cross-section of audit
fees.
In order to test the competitiveness of the audit industry using fee
data, it is first necessary to develop a positive model of the process by
which audit fees are determined. Since an audit fee is the product of unit
price and the quantity of audit services demanded by the management of
the audited company (hereafter called the auditee), cross-sectional dif-
• Assistant Professor, University of British Columbia. This paper is based on my doctoral
dissertation (University of Chicago, 1979). I am indebted to Yale Brozen, Sidney Davidson,
Nicholas Dopueh, Roger Kormendi, Shyam Sunder, and William Wecker for their com-
ments. I have also benefited from the comments of participants in various accounting
workshops at which the results of my dissertation were presented. [Accepted for publication
November 1979.]
161
Copyright ©, Institute of Professional Accounting 1980
162 JOURNAL OF ACCOUNTING RESEARCH, SPRING 1980

ferences in fees can represent either the effect of quantity differences or

price differences. A positive model of the determinants of audit quantity
and price suggests factors which need to be controlled before any infer-
ence about competition can be made from observed fee data.
Following Demski and Swieringa [1974], I consider the external audit
to be a subsystem of an auditee's overall financial reporting system. In
this regard, the audit service is viewed as an economic good to the
auditee, which has substitutes and complements in consumption. Thus
the quantity of auditing demanded by an auditee will result from a
conventional equalization of marginal private benefits and costs.' How-
ever, the nature of the benefits which an auditee derives from the audit
service is stiU an open question. I hypothesize that the potential legal
liability of an auditee and auditor to financial statement users (share-
holders, creditors, etc.) drives the design of external financial reporting
systems. That is, the benefits are in the nature of liability avoidance. The
implications of this sort of motivation are discussed later in this paper.
A second requirement for testing competition is the identification of a
competitive benchmark. The typical approach in the industrial organi-
zation literature (see Weiss [1971]) is to make cross-sectional interindus-
try comparisons of market structure (generally measured by a concentra-
tion ratio) with performance (generally measured by an average rate of
return earned by firms within an industry). In such studies, industries
with "low" supplier concentration serve as a benchmark. However, any
interindustry comparison of rates of return involves difficult problems of
control for confounding differences, such as in risk. In this paper, the test
for competition is an intraindustry comparison of prices, where the
competitive benchmark is the market segment for "small" audits. An
intraindustry test is possible because, as shown in table 1, the market
dominance of the Big Eight firms increases significantly with the size of
the audited company. Thus, I assume that price competition prevails in
the submarket for the audits of "small" companies and test for the effects
of increasing Big Eight concentration on prices paid by "large" auditees.
The data for this paper consist of 397 observations on audit fees and
related variables obtained from a sample survey of publicly held corpo-
rations in the United States. The survey was conducted during 1977. The
data were analyzed using a series of least-squares regressions where the
specification of the regression equations was derived from the model of
audit fee determination. Results include the identification of a number of
' The terms "audit" and "audit service" refer to the process of auditing, not the auditor's
certificate which is attached to a set of financial statement* This distinction is necessary
because the audit production fiinction is clearly not single valued—audit processes of many
different specifications can result in the same observed outcome. Much confusion can be
avoided if one is careful to distinguish between these two elements. For example, while
owners may demand that financial statements prepared by management be accompanied
by an audit certificate, under existing institutional arrangements, it is the auditee manage-
ment who demands the quantity of auditing which underlies that certificate.
 PRICING OF AUDIT SERVICES 163

TABLE 1
Big Eight Firm Concentration Ratios in the Market for
the Audits of Publicly Held Companies When Auditees
Are Classified by Size'
Auditee Site as Measured by Sales Big Eight Concentration
(in millions) Ratio
$1 to $25 .59
$26 to $50 .76
$51 to $100 .82
$101 to $250 .88
$251 to $500 .91
Sales > $500 .95
* These concentration ratios are constructed from the data
reported by Harris [1976] which consist of information on auditor
identity, company size, etc. for 8,077 publicly held corporations.
The concentration ratio is simply the number of auditees in each
size class who were audited by one of the Big Eight firms, divided
by the total number of companies in that class. Note that concen-
tration ratios are generally constructed using the sales, assets,
value added, or number of employees or sellers; however, such
information i.^ not available for CPA firms.

significant audit fee determinants, as well as failure to reject the hypoth-

esis that price competition prevails throughout the market for financial
audit services. The demand-based positive model of auditing developed
in the paper also provides insights into the economics of auditing under
current institutional arrangements.

2. Assumptions of the Fee Determination Model

I assume that both the auditee and auditor are risk neutred and seek to
maximize their own expected profits each period. Thus, auditee manage-
ment seeks to maximize the expected profits of the financial reporting
entity, while the auditor seeks to maximize the expected profits of the
auditing firm. Both parties can purchase resources in competitive factor
markets. Further, let:
a = the quantity of resources utilized directly by the auditee in
operating the internal accounting system
q = the quantity of resources utilized by the auditor in performing
the audit examination
V = the per-unit factor cost of internal accounting system resources
to the auditee
c = the per-unit factor cost of external audit resources to the
auditor, including all opportunity costs and therefore a provi-
sion for a normal profit.
I assume that resources are utilized efficiently so that a and q not only
denote inputs to the auditee's financial reporting system but also corre-
spond to unique quantities of output constructs which may be called
internal accounting control and external audit control, respectively. Thus,
164 DAN A. SIMUNIC

a financial reporting system is completely described by the ordered pair

(a,q).
Assume that the auditee and auditor are jointly and severaUy liable to
financial statement users for losses attributable to defects in the audited
financial statements.^ Further, the benefits from the financial reporting
system (a, q), derive solely from reduction of losses tofinancialstatement
users. Let the random variable, d, denote the present value of possible
future losses which may arise from this period's audited financial state-
ments. Thus, E(d) = f(a, q). Assume that the auditee and auditor
identically assess this function which has first- and second-order partial
derivatives where:
aE(d) i'E(d)
<0 - >0
aa aa' aaaq
aE(d) fE(d)
<0 >0
aq aq'
And at any given level of E(d):
da d'a
CO
dq dq'
Since liability is joint and several, actual losses will somehow be divided
between the two parties. Let 6 denote the ex-post fraction of losses borne
by the auditor where:

At the time of the audit, this loss apportionment factor is also a random
variable. Assume that the auditee and auditor identicaUy assess E(d) and
assume that d and 0 are independent.
Finally, let p denote the unit price of external audit services to the
auditee; that is, the price per unit of q purchased by the auditee from the
auditor. The auditor's revenue derived from an audit engagement is
therefore pq, or the audit fee.

3. The Auditee's Problem When the Market for Audits Is

Competitive
An expected profit-maximizing auditee wUl seek to minimize the ex-
pected periodic costs of operating a financial reporting system. Let TC
- Section ll(f) of the Securities Act of 1933 and Section 18(b) of the Securities Exchange
Act of 1934 provide for joint and seversd liability on the part of auditees and auditors. Under
these provisions, the entire amount of damages suffered by a third party can be collected
from any one of the liable persons, with that person generally retaining rights to collect
from all other persons who are also liable.
 PRICING OF AUDIT SERVICES 165

denote the total costs of the system. The auditee's problem can initially
be expressed as an unconstrained minimization of expected total costs:
minE(TC) = va + pq + E(d\a,q)(l - E(d)). (1)
In this problem, the choice variables are the systems design, (a, q),
whereas i' is a market parameter, E(S) is an unconditional expectation,
and E{d) is an expectation conditional on the system's design. The value
of p, on the other hand, will depend on the state of competition in the
market for audit services.
An auditor's minimum supply price per unit of q is marginal cost.
Alternatively, his minimum fee for different levels of audit quantity wiU
be equal to his incremental expected total cost, denoted E(C), where:
E(C) = cq + E(d\a,q)E(e). (2)
Note that the auditor's expected costs are a function of the auditee's
financial reporting system. Since by definition, the parameter c includes
all costs of a unit of q, including a normal return,' when the market for
audits is competitive pq' = E(C), and the auditee's problem becomes:
rmnE(f'O = va + cq + E(d\a, q)E(6) +E{d\a,q)(\ - Eiff)) (3)
or simply:
min£( fC) = va + cq + E(d\a,q). (4)
The necessary condition for the expected cost minimization is:

aE(TC) dE(d) -0 or dE(d)

aa da da

EifC) md) (- r = 0 or
dE(d)
dq dq aq

This condition states that the auditee will demand quantities of a and q
up to the point where the marginal reduction in expected liability losses
is equal to the auditee's marginal resource cost. The solution to this
system of equations under competition is denoted (a, q).
While the solution values (d, q) are obtained simultaneously, the
nature of internal accounting systems and external auditing suggests that
implementation of the solution by the auditee is sequential, that is, d is
followed by q. This is consistent with the normal technical auditing model
wherein an auditor's audit program design is a function of the internal
accounting system. However, to what extent does the model allow for
independent action on the part of auditors in attaining the solution (d,
g)? In other words, given that the auditee has computed (d, q) and
' It is reasonable that an auditor's return to capital be computed per unit of q rather
than per engagement, since an auditor's capital itself is predominantly general rather than
engagement specific.
166 DAN A. SIMUNIC

installed the internal accounting system d, what additional conditions are

necessary for the auditor to perceive that an audit examination of scope
q is in his own best interests? An agreement on the amount of the audit
fee, pq, insures only that q -^ q, since for larger values of q an auditor
would fail to earn a normal return. To insure that audit scope is in fact
equal to q, it is necessary to assume (in a single-period model) that the
auditee is able to evaluate q.
It is also worth noting that the solution (d, q) is invariant to the
incidence of third-peuty liability. Equation (3) will reduce to equation (4)
for all values where 0 :£ E(d) :< 1. This result holds because the auditee
expects to incur aU systems costs and is eissumed to be indifferent about
whether costs occur in the form of the cost of internal accounting systems
(I'd), the external audit fee (cq -I- .E(d | d, q)E(d)), or the expected present
value of the auditee's share of residual liability losses E(d\d, q)(l -
E(6)). However, attainment of the competitive solution when E(d) > 0
requires that the auditee recognize that p is not a fixed value but a
function, p(a, q). In addition, when E(6) > 0, the auditee must be able to
analyze the component terms of equation (2). Thus, if one assumes that
analysis is costly, the imposition of third-party liability upon the auditee
alone is preferable to a regime of joint and several habUity, unless there
are sufficient offsetting benefits associated with existing arrangements.

4. The Auditee's Problem in a Noncompetitive Market

Setting
Suppose that a dominant subset of auditors (e.g., the Big Eight firms),
through collusion, agree to limit price competition so as to introduce an
element of monopoly profit into audit prices. Let m represent the amount
of monopoly rent included in the unit price p.* The fee schedule of an
auditor who was a member of the cartel would be:
pq = (c + m)q + E(d\a, q)E(e).
Substituting into equation (1) and simphfying, an auditee would seek to:
minE( TC) = va + (c + m)q + E(d\a, q).
And the necessary condition for the minimization would be:
dE(d)
da
aE(d)
= c + m.
' In principle, a group of auditors acting as a cartel could he expected to compute and
incorporate into p, a joint profit-maximizing value of m However, to demonstrate the
effects of monopoly pricing, determination of the optimum m is not necessary and I merely
assume that m > 0.
 PRICING OF AUDIT SERVICES 167

By inspection, the revised solution value of q, denoted qm, must

decrease relative to q. In addition, since:

dadq
the reduction oi q makes the value of aE(d)/aa more negative at any a,
and thus the desired quantity of internal accounting under monopoly,
denoted am, must be greater than d. However, because of diminishing
substitutability between internal accounting and external audit resources
in controlling liability losses, it must also be true that:

E(d\an,,qm)>E(d\d,q), and (5)

E(TC\a^,qr,,)>E(TC\d,q). (6)
That is, monopoly pricing induces a substitution away from q toward a
relatively less productive resource, a, which results in an increase in the
desired level of expected residual liability losses (relation (5)). This effect
can be described as a decrease in the quality of the auditee's financial
reporting system because financial statements are now more likely to be
defective, resulting in larger expected liabUity payments to third parties.
In summeu^, auditor monopoly reduces the quantity demanded of exter-
nal auditing jmd results in lower-quality, higher-cost financial reporting
systems.
By definition, monopoly pricing increases the unit price of extemed
auditing, p. However, the effect upon observable audit fees is indetermi-
nate. That is:
pqn, i pq,
depending upon the price elasticity of demand implicit in the auditee's
cost minimization problem at the competitive solution point (d, ^). If the
implicit demand function is inelsistic in the vicinity of the point (c, q),
then for small values of m, the audit fee would increase, while if demand
was elastic in that vicinity, the audit fee would decrease.
While the increase inp has no clear implication for the audit fee itself,
I have shown that the auditee's_total systems costs must increase in a
monopoly setting. However, E(TC) includes an unobservable cost com-
ponent, E(d\ a, q)(l — E(0)), the share of residual losses expected to be
incurred directly by the auditee. Unless the entire increase in E(TC)
occurs in this unobservable component, which is unlikely,^ monopoly
pricing will increase an auditee's observable systems costs, or:

vam + pqm > vd + pq.

' For this to occur, it is necessary that £(fi) = 0 and that the entire increase in E{ fC) be
in the expected residual loss component.
168 DAN A. SIMUNIC

5. The Effects of Auditor Production Economies on

Financial Reporting Systems
While I 2issume that the factor costs of resources are uniform to all
auditors, the production functions of audit firms may nevertheless differ.
In the model, production is measured by reduction of expected liability
losses. Therefore, allowing for production function differences implies
that dE(d)/dq may vary across auditors.
If specific auditors enjoy unique economies in production, then these
firms will earn economic rents, but the characteristics of auditee financial
reporting systems, including audit fees, wiU not be affected. Sucb econ-
omies are not of interest here. Alternatively, there may exist sources of
economies which can be exploited, at least potentially, by more than one
auditor. If there is competition among the auditors who achieve the
economies, then rents would be bid away and audit prices would decrease.
In addition, in equilibrium, auditees would only demand services from
those auditors who fully achieved the available economies.
The necessary condition for the minimization of expected costs with
economies is:
_ dE(d) _
~~da ^
dE(d) dE(d) c
—A = c or = -
dq dq \
where the parameter \ S 1 denotes an auditor's relative efficiency in
reducing expected losses. For auditors who achieve economies, X > 1. I
will denote the auditee's expected cost-minimizing solution with econ-
omies as (ae, qe). The characteristics of this solution are the reverse of
the monopoly results. That is, while p decreases, the sign of the change
in audit fee is indeterminate, since the auditee is motivated to substitute
q for a and the net change in pq depends on the price elasticity of the
imphcit demand function for external auditing. As in the case of monop-
oly, unless the entire change occurs in the auditee's share of residual
losses, which is unlikely, then:
rap -I- pq^ < id + pq.

6. Effects of Variations in Assessed Loss Functions

The loss function, E(d) = f(a, q), wiU vary across audit engagements.
The assessment of this function by the auditee and auditor can be
expected to reflect all available information, including historic loss expe-
riences under similar circumstances. The spatial location of the function
assessed for a specific engagement can be influenced by many factors,
including the leged environment and various internal characteristics of
the engagement, which I call determinants of loss exposure. Possible
 PRICING OF AUDIT SERVICES 169

variables which might be associated with such differences are discussed

in a later section. At this point, I simply note that a uniform upward
displacement of the loss function in the (E(d), a, q) space, that is, greater
loss exposure at any (a, q), increases the marginal benefits from expected
loss reduction and results in expansion of the cost-minimizing financial
reporting system. Note that there is no change in relative prices. Rather,
the expansion of a and q is complementary. As a result, each of the three
components of the auditee's expected total costs, namely, va, pq, and
E(d\ a, q)(l — E(S)), can be expected to increase. Finally, this increase
in demand for both control resources will occur under any and all of the
three industry scenarios—competition, monopoly, and auditor econo-
mies—discussed in the previous sections.

7. Summary of the Model and Development of the Test for

Competition
The implications of the model, in the form of directional changes in
the auditee's decision variables emd related costs, are summarized in table
2. The changes described in the first two columns of the table are
measured against the competitive solution; the last two columns describe
the effects of differences in assessed loss functions and the loss-sharing
ratio. The characteristics of auditee financial reporting systems are
grouped into two categories, unobservable and observable. Included in
the former are the auditee's share of residual losses, the price of audit
services, and the quantities of internal accounting control and external
audit control which the auditee demands.
TABLE 2
Implications of the Model with Respect to Cross-Sectional Differences in the
Characteristics of Auditee Financial Reporting Systems

Monopoly Production
increase in Audi- Increase in
Pricing Economies tor s onaje ot Loss Exposure
Losses
Unobservable:
Auditee's share of residual
losses £ ( d 1 a, 9)(1 - £(#)) -H - - 4-
Unit price of the audit service
(P) -H - -H none
Quantity of internal accounting
control (a) -t- — none +
Quantity of external audit con-
trol (g) — -t- none
Ohservable:
Audit fee (pq) Depends on implicit
price elasticity of de-
mand -t- -1-
Cost of internal accounting con-
trol {va) -^ none -t-
Sum (va +pq) -1-
170 DAN A. SIMUNIC

Recall from the assumptions of the model that a and q either can be
thought of as generalized quantities of factor inputs or as output con-
structs. As outputs, these quantities are not observable. It is the output
construct, external audit control, which is priced in the market for audit
services and the price, p, is likewise not observable. Rather, we can only
observe the fee, pq. In the case of internal accoimting, there is no
intervening output market and thus no market price. However, assuming
efficiency in resource use, va and pq are commensurable.
Based on the model, the hypothesis that price competition prevails
throughout the market for audit services can be tested using the sum of
observable systems costs, va + pq. Before I develop the test, note that
while the Big Eight firms have been accused of monopolizing the market
for audit services, these firms are also the most hkely to have exploited
available economies of scale. Furthermore, scale economies can exist in
either a monopolistic or a competitive market setting. Note also that
audit services may be differentiated. With respect to product differentia-
tion, the market for audits is a hedonic market (see Rosen [1974]). That
is, differentiated products are not observed directly but rather are re-
vealed by differences in prices which are associated with differences in
observed product characteristics. In auditing, the principal differentiating
characteristic of the service is likely to be the identity of the suppUer,
and again it is the Big Eight firms which enjoy visibility and brand-name
recognition among buyers. Thus, the pricing of audit services can be
expected to be complex, and any price differences between Big Eight
firms and other auditors must be interpreted with care.
Given these considerations, to test the hypothesis of competition it is
first necessary to control for cost differences arising from differences in
loss exposure and the expected loss-sharing ratio. Call the cost not
explained by these factors the cost residual. Next, classify auditee cor-
porations into two size categories, "small" and "large,"* and classify
auditors as Big Eight or non-Big Eight firms. The purpose of the auditee
size classification is to partition the market into a segment where auditees
can and do purchase services from a large number of supphers and which
can therefore be assumed competitive, and a segment where the Big
Eight are highly dominant and may behave as a cartel. Differences in the
average cost residuals between auditees using Big Eight and those using
non-Big Eight firms c£m then be interpreted using table 3. Note that the
test requires a joint comparison of differences in average cost residuals in
both market segments; that is, results in the "large" auditee segment can
only be interpreted by reference to the competitive benchmark. Finally,
in interpreting possible findings, I assume that if the Big Eight firms
collude to increase prices in the "large" auditee segment, their non-Big
Eight competitors would seek to expand market share and price consist-
' For the moment, the bound between these categories is not specified.
 PRICING OF AUDIT SERVICES 171
TABLE 3
Test for Competition: Interpretation of Possible Differences in Average Residual Total
Systems Costs of Auditees Using Big Eight versus Non-Big Eight Auditors across
Market Segments
ResultD for the "Small" Auditee Segment
Result* for the "Large"
Auditee Segment \ (CRE\ai<
(CUE 18) (CRE\») (CRE\»)
{CRE\S)>(CRE\8)' Competition with Monopoly pric- Monopoly pric-
differentiated ing by the Big ing by the Big
product to the Eight Eight together
Big Eight with scale econ-
omies to the Big
Eight

(CRE\8) Competition with Competition Monopoly pric-

differentiated without any scale ing by the Big
product to the economies to the Eight together
Big Eight to- Big Eight with scale econ-
gether with dis- omies to the Big
economies to Eight
hon-Big Eight
with "large" au-
ditees"

(CRE\8)<{CRE\S) Competition with Competition with Competition with

differentiated diseconomies to scale economies
product to the non-Big Eight to the Big Eight
Big Eight to- with "large" au-
gether with dis- ditees
economies to
non-Big Eight
with "large" au-
ditees
* (CRE18) denotes the average residual costs of auditees using a big Eight firm and (CRE \ 8) denotes
the average residual costs of audit«es using a non-Big Eight firm.
* * The combinations wherein the costs of large auditees only are lower when a Big Eight firm is the
auditor would be evidence of economies to the Big Eight or diseconomies to the non-Big Eight when
performing large audits. Note that this is not evidence of economies to auditor size, arising, for example,
ftom staff specialization, since the effect is limited to a certain audit context. As a result, I would interpret
such findings as evidence of diseconomies to smaller auditors when servicing large auditees.

ent with their own cost conditions, rather than to maintain the cartel
price.

8. Control Variables for Differences in Loss Exposure

Little is presently known about sources of variations in liability loss
exposure across audit engagements. To obtain some initial insights, I
discussed the question with Chicago-area representatives of each of the
Big Eight firms and with representatives of a number of organizations
172 DAN A. SIMUNIC

writing professional liability insurance coverage for accountants.' From

these discussions, the following general factors were identified as possible
determinants of loss exposure: (a) the size of the auditee, (b) the com-
plexity of the auditee's operations, (c) auditing problems associated with
certain financial statement components, especially inventories and re-
ceivables, (d) the industry of the auditee, and (e) whether the auditee is
a publicly or closely held company. The data for this research were
obtained solely from publicly held corporations and thus were homoge-
neous with respect to the last factor. The variables and their measures
used to control for the remaining factors are described below. Each
variable is assigned a mnemonic name for ease of further reference.

SIZE OF THE AUDITEE

I measured the size of the auditee by the entity's total year end assets
(variable name ASSETS). A priori, the stock of assets seems more closely
related to possible loss exposure than would an accounting fiow measure,
such as revenue, because defective financial statements which result in a
lawsuit frequently involve some deficiency in asset valuation. In addition,
external auditors have traditionally approached the audit process through
the ending balance sheet, relying on the fact that verification of balance
sheet components indirectly verifies reported income. Note also that both
internal accounting and external auditing are sampling-based processes.
To the extent that increases in measured total assets of auditees refiect
increases in the number of individual elements which comprise the
accounting populations of which total assets are composed, then the
sample size required to achieve a given level of control will increase at a
decreasing rate. Thus, I hypothesized that the positive relationship
between ASSETS and both q and a is nonlinear.

COMPLEXITY OF THE AUDITEE'S OPERATIONS

Loss exposure can be expected to increase, the greater the decentrali-
zation and diversification of the financial reporting entity. Both of these
aspects of complexity increase the number of decision centers in an
organization whose activities need to be monitored. The recent contro-
versy over illegal corporate payments and the resulting internal control
requirements imposed by the Foreign Corrupt Practices Act reflect the
overall problem of controlling large decentralized organizations.
In this study, I measured decentrahzation by the number of consoli-
dated subsidiaries which are included in the auditee's financial statements
(variable name SUBS). The diversification of the auditee is measured
using two variables: (1) the number of Standard Industrial Classification
System two-digit industries in which the auditee operates, less one
' Information was obtained from the following insurers: American Home Assurance
Company, the St. Paul Companies, Chubb Custom Market, Inc., Crum & Forester Insurance
Companies, and Lloyd's Underwriters' Non-Marine Association.
 PRICING OF AUDIT SERVICES 173

(variable name DIVERS), and (2) by the ratio of the auditee's foreign to
total assets at year-end (variable name FORGN).
RECEIVABLES AND INVENTORIES
Auditors have long recognized that certain accounting populations
involve potentially greater loss exposure through the loose notion of
"relative audit risk" (see, e.g., Lenhart and Defiiese [1957]). In this sense,
both receivables and inventories are "risky" balance sheet components.
Specific auditing procedures (confirmation and observation) are recom-
mended for these accounts. Moreover, the valuation of these items is a
complex task, requiring a forecast of future events. Liability exposure is
thus expected to vary cross-sectionally with the relative size of receivables
and inventories in different auditee badance sheets. To control for these
differences, I used the two ratio variables: receivables to total assets at
year-end (variable name RECV) and inventories to total assets at year-
end (variable name INV).
PRINCIPAL INDUSTRY OF THE AUDITEE
While loss exposure may well vary with the industry(ies) in which an
auditee operates, there is really no basis to hypothesize any specific
industry effects. However, possible differences in q and a associated with
industry classification were also investigated in the sample data.

9. Control Variables for Differences in the Assessed Loss-

Sharing Ratio
A plausible, and probably the only measurable, determinant of .E(^) is
evidence of auditor or auditee financial distress. Recall that the auditor
and auditee are assumed to assess identical distributions on 6. Evidence
of auditor financial distress increases the probability that the realization
of 0 will be ^ = 0; that is, all losses must be paid by the auditee because
the auditor is insolvent. Thus E(6) would decreeise. Conversely, evidence
of auditee financial distress increases the probability that realized ^ = 1,
due to auditee insolvency, and E(d) would increase.
Because public accounting firms are private partnerships, evidence
which might be used to assess possible financial distress of auditors is
difficult to obtain. As a result, I did not control for this effect.
I used three variables to control for cross-sectional differences in .E(^)
arising from auditee financial difficulty. The first is a measure of the
auditee's accounting rate of return in the current year, the ratio of net
income to total assets at year-end (variable name PROFIT). E(6) is
expected to increase as the rate of return decreases. Note that rate of
return measures have been found useful in bankruptcy studies (e.g..
Beaver [1968] and Altman [1968]) for discriminating between "failed"
and "nonfailed" firms. The second variable is a (0, 1) variable which was
174 DAN A. SIMUNIC

assigned a value of one if an auditee had incurred a net loss during any
one of the current or two prior fiscal years (variable name LOSS). Finally,
I used another (0, 1) variable to identify those auditees who received a
"subject to" qualified audit opinion in the current year (variable name
SUBJ). A "subject to" qualification is given when there exist significant
uncertainties which may result in future losses to the auditee. In the
United Kingdom, Firth [1978] found that the issuance of an opinion form
equivsdent to a "subject to" (namely, "asset valuation" and "going con-
cern" qualifications) by a U.K. auditor had a significant negative impact
on the market price of the auditee's securities. Thus, the conditions which
underlie the issuance of a "subject to" opinion were hypothesized to
affect the assessment of E(6) such that E(0) increases when SUBJ takes
on a vedue of one.

10. Differences in Auditor Production Functions

There has been essentially no previous research in the area of auditor
production functions and sources of production economies. In my study,
I did not investigate directly any specific sources of economies to scale.
Rather I inferred the presence of some unspecified economies to firm size
if, as described in table 3, the residual systems costs of "small" companies
using Big Eight auditors were, on average, lower than the costs of "small"
companies using non-Big Eight auditors.* The argument is analagous to
a test for scale economies based on the survivor principle (see Stigler
[1968]), in that the current state of the industry might arise from any
number of factors which increase the relative efficiency of the Big Eight
firms.
I did, however, control for the possible source of production function
differences across engagements brought about by reductions in q because
of auditor learning over time. The reduction of cumulative average costs
through learning during successive performances of a task has been
observed in several physical production situations. A similar phenomenon
can be expected in auditing if the auditee employs a given auditor for a
number of years. Normally, the effects of learning would be measured
using time series of production and cost data. But, a learning effect can
also be observed in a cross-section of audit fees if the effect is sufficiently
large and the length of auditee/auditor association varies in the sample.
To control for possible differences in q from this source, I included as a
control variable the number of years an auditee has used its current
' To infer the presence of scale economies to the Big Eight, it is both necessary and
sufficient that (CRE \ 8) < (CRE 18) in the competitive market segment. Of course, if the
Big Eight do not price as a cartel, then the same relationship would be obser\ed in the
market segment for "large" audits.
 PRICING OF AUDIT SERVICES 175

auditor (variable name TIME).^ The relationship between TIME and q

was hypothesized to be negative.
11. Measurement of the Dependent Variables
The auditee's total observable systems costs consist of the two com-
ponents, va and pq. Since the audit fee is a payment made in a market
transaction, measurement of pq was relatively straightforward. This
element of total cost is denoted by the variable named FEE.
Identification and measurement of the empirical counterpart of the va
component was substantially more difficult. From the model, note that
internal accounting resources sure only relevant to a test for competition
among external auditors if the (a, q) resources are substitutable. In order
to obtain a relevant measure of va for purposes of this study, it was
sufficient to identify and measure those elements of internal systems
costs which could reasonably be substituted for external audit services.
Based on this argument, I measured va using the salaries paid by the
auditee to its internal auditors. This element of total cost is denoted by
the variable named ICOST.
WhUe the test for competition is based on residual differences in the
sum of FEE + ICOST, the individual components may differ in degree
of measurement error and are each of interest. Therefore, in the analyses
which foUow, I examine the behavior and determinants of the separate
components as well as of the sum.
The variables, their measures, and hypothesized relationships are
summarized in table 4. The average residual difference in the systems
costs of auditees across the two auditor classes is measured by the
coefficient of a (0, 1) variable named AUDITOR which is assigned a
value of one when the auditor is a Big Eight firm. The questionnaire used
to obtain data on these variables is available on request.

12. Design and Results of the Survey

The basis for the survey was the list of publicly held corporations and
their auditors comprising the 1976 edition of Who Audits America? This
list includes 8,077 companies classified by sales volume and auditor
identity. A stratified sample of 1,207 companies was contacted during
1977 througb a questionnaire mailed to a principal financial officer of the
corporation. The size of the sample was based on an informal assessment
of the marginal benefits-costs of sampling and an expected response rate
of 30 to 40 percent.'° The sample is stratified by size of auditee (companies
' As learning occurs, an external auditor becomes relatively more efficient in reducing
Icsses and, from the model, the auditee is motivated to substitute q for a. However,
• adjusting a cont(ol system is probably costly, and I do not expect to observe nor do I control
for this effect in the empincal work.
™ The response rate to the Financial Executive Institute's audit fee survey of its members
was 41 percent (Hobgood and Sciarrino [1972]).
176 DAN A. SIMUNIC
TABLE 4
Summary of Variables and Hypothesized Relationships
Relationship to Auditee's Sys-
Name tenis Cost Component
FEE ICOST
Control variables for differences in loss
exposure:
1. Total assets at year-end ASSETS
2. Number of consolidated subsidiaries SUBS -^
3. Number of two-digit SIC industries DIVERS +
in which auditee operates, less one
4. Foreign assets •>- total assets at year- FORGN
end
5. Accounts, loans, and notes receivable RECV
+ total assets at year-end
6. Inventories -s- total assets at year-end INV

Control vsuriables for differences in the

assessed loss-sharing ratio:
7. Net income -^ total assets PROFITS N/A
8. (0, 1) variable where (1) if auditee LOSS
incurred loss in any of last three fiscal N/A
years
9. (0, 1) variable where (1) if auditee SUBJ N/A
received a "subject to" qualified opin-

Control variable for differences in auditor

production functions:
10. Number of years auditee has used TIME N/A
current auditor

Auditor identity:
11. (0, 1) variable where (1) if auditor is AUDITOR any—see table 3
a Big Eight firm

Dependent variables:
12. Amount ofcurrent year's external au- FEE
dit fee
13. Salaries paid to internal auditors in ICOST
current year

with sales less than $125 million versus those with sales greater than that
amount) and by auditor group (Big Eight firm versus non-Big Eight
firm). The "small" auditee segment was bounded at $125 miUion, since
the marginal Big Eight market share for clients of this size and larger
approaches 90 percent. Thus the hypothesis that competition prevails in
the market for audits of companies greater than about this size becomes
difficult to maintain, a priori."
" That is, a market share this large would surely be sufficient evidence, to a believer in
the "concentration doctrine," that the market segment was not competitive. Even a smaller
 PRICING OF AUDIT SERVICES 177

TABLE 5
Responses by Stratum
Large Auditees
Small Auditees (sales (sales greater than All Auditees
less than $125 miUion) 1125 million)
Big Eight auditor:
Sample requests 333 425 758
Usable responses 117 172 289
Response rate 35% 40% 38%

Non-Big Eight auditor:

Sample requests 326 123 449
Usable responses 70 38 108
Response rate 21% 31% 24%

All auditors:
Sample requests 659 548 1207
Usable responses 187 210 397
Response rate 29% 37% 33%

Within each stratum, sample units were selected at random. Re-

sponse rates by stratum are shown in table 5. FoUow-up second requests
were sent to approximately 50 percent of initial nonrespondents, with an
emphasis on those strata with a low response rate.

73. Test for Nonresponse Bias

I examined the possibility that respondents and nonrespondents were
not homogeneous with respect to relevant characteristics by comparing
the vedues of variables reported by early and late respondents, a standard
test for nonrespondent biases (see Oppenheim [1966]). Means and stan-
dard deviations (where applicable) of the values of reported variables for
the two groups are shown in table 6. The early respondents are those
whose replies were received during the first three weeks (the declared
due date of the mailing), whUe the late respondents consist primarily of
those companies who replied to the second request.
Note from table 5 that nonresponse was more frequent among smaller
companies and those auditees using non-Big Eight auditors. Table 6
shows that while the size of early and late respondents was approximately
the same, a substantially larger proportion of late respondents used a
non-Big Eight auditor. This supports the argument that late respondents
were good proxies for nonrespondents. Since the mean values of the
vsunous other variables were substantially identical, the data obtained do
not seem to be biased. However, I still cannot explain the reluctance of
companies using non-Big Eight auditors to respond to the survey.
market share might he sufficient to support an allegation of monopoly, but there is no basis
to determine a minimum required share. Thus the bound is necessarily somewhat arbitrary.
178 DAN A. SIMUNIC

TABLE 6
Test for Nonresponse Bias
"'Late Respondents

ASSETS: mean $389,125M $384,682M

std. dev. (918,908) (756,725)
SUBS: mean 13 15
std. dev. (20) (44)
DIVERS: mean .89 .91
std. dev. (1.32) (1.47)
FORGN: mean .07 .06
std. dev. (.17) (.13)
RECV: mean .23 .24
std. dev. (.16) (.18)
INV: mean .23 .26
std. dev. (.18) (.28)
PROFIT: mean .06 .06
std. dev. (.07) (.05)
LOSS: percent of companies with net losses 2O''r 21%
SVBJ: percent of companies with a "subject
to" 10% lO^r

TIME: mean 18 19
std. dev. (13) (16)
FEE: mean • $170M $193M
std. dev. (258) (337)
AUDITOR: percent of companies using a Big Eight
auditor 71% 42%

14. Test of the Determinants of Audit Fees

Data received from the respondents are summarized in table 7, using
means and standard deviations (in pju-entheses) for the arithmetic vari-
ables, and the percentage of observations wben the variable takes on a
value of one for the categorical variables.
The hypotheses about the determinants of the audit fee component of
total observable costs were tested by obtaining least-squares' estimates
of the coefficients of the variables in the following linear regression
function:
= bo + biSUBS + biDIVERS + bsFORGN + [Link]
-I- bJNV + bePROFIT + b7L0SS + bsSUBJ
+ bc,TIME + bioAUDITOR + u
where the variables were constructed as described in table 4, and where
the error term, «, was assumed to have the standard properties.'^ The
" Examination of various scatter plots of residuals indicates that the residual variance is
homogeneous when the square-root transformation is used. However, with the cube-root
transformation, as well as with a log transformation of ASSETS, which was also examined,
the error variance is correlated with ASSETS. When the exponent is varied from .33 to .5,
the signs of significant coefficients are unchanged and there are only small changes in the
(-values for all coefficients.
 PRICING OF AUDIT SERVICES 179
TABLE 7
Descriptive Statistics for Variables
By Auditee Size By Auditor Group
397 Total 187 Auditees 210 Auditees 108 Auditees
fiVi AiiHiti^iiR
Ohservation« with Sales with Sales bO^ nuuiiees Using a Non-
Less Than Greater Than Big Eight Audi-
tI25 million $126 million ig u or tor
FEE $206.6M $77.1M $322.0M $226.6M $153.0M
(277.1) (71.4) (355.0) (266.9) (297.2)
ICOST $I12.0M $29.0M $202.8M $141.0 $ 45.5M
(271.0) (83.0) (361.8) (308.2) (134.2)
ASSETS $555.1MM $176.7MM $891.9MM $695.6MM $178.9MM
(1,194.5) (640.7) (1,147.7) (1,311.8) (672.8)
SUBS 16.9 7.0 25.6 18.1 13.4
(30.5) (13.9) (37.8) (31.1) (28.8)
DIVERS .9 .6 1.3 1.0 .7
(1.4) (.9) (1.6) (1.4) (1.1)
FORGN .07 .05 .11 .08 .05
(.15) (.11) (.17) (.15) (.11)
RECV .23 .24 .18 .23 .23
(.17) (.18) (.11) (.17) (.14)
INV .23 .25 .23 .20 .29
(.19) (.21) (.17) (.17) (.23)
PROFIT .06 .06 .06 .06 .07
(.06) (.08) (.04) (.04) (.08)
LOSS 16.3% 25.7% 8.1% 13.8% 23.1%
SUBJ 8.0% 8.5% 7.6^r 8.0% 9.3%
TIME 19.0 yrs. 13.7 yrs. 23.8 yrs. 19.9 yrs. 16.6 yrs.
(15.2) (12.0) (16.1) (15.8) (12.9)
AUDITOR 72.5% 62.0% 82.0% 100.0% 0.%

dependent variable was deflated by some power transformation of AS-

SETS, denoted ASSETS', in order to linearize the FEE = g(ASSETS)
relationship. I used a power transformation because, in the absence of a
theory from which the form of the function, g, can be derived, the
exponent for a power transformation can readily be estimated. Thus, if
FEE = wASSETS'v, then In FEE = In w + e(ln ASSETS) + In v, and
a least-squares estimate of the exponent is simply the e in the regression
of In FEE on In ASSETS.
Note that by using size-defiated audit fees as the dependent variable,
an implicit interaction is assumed between ASSETS' Jind each of the
independent variables and the error term in the determination of the
observed undeflated value of FEE. That is, the effect of each of the
independent variables in the regression function was assumed to be
conditional on auditee size.
Using this approach, the residt of the first step in the data analysis
yielded a regression of In FEE on In ASSETS of:
180 DAN A. SIMUNIC

In FEE = 3.33 -t- .45(ln ASSETS)

^ = 9.1 t = 22.7
R' = .57
n = 397 observations.

Taken alone, the variable ASSETS is a very significant determinant of

the audit fee, with a nonlinear relationship, as hypothesized. While the
estimated exponent for the ASSETS transformation is .45, the test of the
determinants of the ratio, FEE/ASSETS', may be sensitive to the
specific transformation which is used. Therefore, I fitted the regression
function using both a somewhat larger exponent value of .5 (a square-
root transformation) and a somewhat lower value of .33 (a cube-root
transformation). The best results were obtained with a .5 exponent and
are shown in table 8. However, overall results were not sensitive to this
variation in the exponent.'^
Table 8 shows the regression coefficients, their standard errors (in
parentheses), and various regression statistics. Coefficients which are
significant at the .05 level in a one-tail or two-tail t-test (as appropriate)
are marked with an asterisk. The correlation matrix of the variables is
displayed in table 9. Column 1 of table 8 includes aU usable observations,
whUe the last three columns, which are relevant to the test for competi-
tion, exclude twenty-four responses received from banks (SIC code 60).
Bank respondents were excluded from the test both because none of the
banks in the sample was audited by a non-Big Eight firm, and they were
outhers with uniformly very low vedues for the dependent variable. The
coefficient for the variable BANK in column 1 is for a (0, 1) variable
(given assigned a value of one for these twenty-four respondents).
Examination of the data also revesded that the dependent variable was
systematically low for public utility companies (SIC code 49). However,
the twenty-two utilities in the sample were audited both by Big Eight
and non-Big Eight firms. Thus, I included in the regression another (0, 1)
variable, denoted UTILITY, whose observations were retained for the
test of competition. In addition, a power transformation (using a .5
exponent) of the variable SUBS and a log transformation of the variable
TIME helped to linearize the fitted function.
A final modification of the regression function, made after a preliminary
examination of the data, was to partition the Big Eight firms into the two
groups. Price Waterhouse & Co. ((0, 1) variable named AUDITOR-PW)
and the remaining seven firms ((0,1) variable named i4[7D/T'O/?-7). This
was done because, as shown in table 10, the average value of the
dependent variable varied across the Big Eight, with Price Waterhouse
" This is the hypothesized specification of the fimction. As explained below, certain ex-
post modifications of the regression function were made after a preliminary analysis of the
data.
 PRICING OF AUDIT 8ERVICES 181
TABLE 8
Stgression of (FEE/ASSETS') on Explanatory Variables
Excluding Banks
"* 397 Total 373 ObMrva-
Sign of tions Excluding 171 Auditees 202 Auditees
Coeincien Observations Banks with Sales Less with Sales
Than $125MM Greater Than
$125MM
SUBS' .96* 1.01* 1.32* .93*
(.14) (.15) (.37) (.18)
DIVERS + 1.03* .99* 1.82* .72*
(.26) (.27) (.55) (.31)
FORGN 14.88* 14.72* 14.37* 14.61*
(2.45) (2.53) (5.27) (2.88)
RECV -f 9.06* 9.72* 6.15* 18.93*
(2.29) (2.41) (3.06) (4.46)
INV + 7.26* 7.42* 6.46* 9.09*
(1.86) (1.90) (2.57) (2.98)
BANK none -9.79*
(1.63)
UTILITY none -2.97* -2.83* -1.64 -1.62
(1.59) (1.63) (4.01) (1.87)
PROFIT - 2.52 2.58 1.81 1.53
(5.52) (5.64) (6.74) (12.20)
LOSS + 1.83* 1.85* 2.12* .93
(.92) (.94) (1.20) (1.73)
SUBJ 2.71* 2.84* 3.84* .81
(1.19) (1.23) (1.90) (1.73)
LOG(TIME) - -.43 -.23 1.18 -1.58
(.86) (.89) (1.34) (1.28)
AUDITOR-PW none .76 .94 .78 1.20
-(- o r - (1.14) (1.18) (1.88) (1.61)
AUDITOR-! -1.66* -1.69* -1.84 -1.15
(.77) (.79) (1.08) (1.21)
Intercept 5.67 5.14 3.84 5.14
Std. error of the estimate 6.32 6.45 6.57 6.30
Adjusted R' .46 .42 .28 .51
F 26.79 23.6 6.5 18.4
FEE/ASSETS'
mean 13.0 13.61 12.61 14.54
std. dev (8.58) (8.49) (7.74) (8.98)
* An asterisk indicates that the regression coefficient is significant at the .05 level in a one-tail or two-
tail (as appropriate) (-test. Standard errors of the regression coefficients are shown in parentheses.

a high-value outlier both in average deflated fees received and, as will be

shown, in auditees' average deflated internal costs."
" It is interesting to note that the ranking of the Big Eight by average deflated fees in
table 10 corresponds closely to the verbal description of each firm recently given by
Bemstein [1978]. Quoting Bernstein, the firms are described in sequence: PW—"the premier
accounting firm"; AY—"not aggressive, super professional"; E & W—"not on the competi-
tive edge"; DHS—"not aggressive, strong auditors"; AA—"aggTessive... emphasize
growth"; PMM—"very aggressive... price cutter"; TR—"very aggressive... price cutter";
CX—"most aggressive of the eight... price cutter."
182 DAN A. SIMUNIC

r • r

S m -v a ^ m ^ t.-qp
§1 ' ' ' ' \' \ ' \' \'

$
1 1 1 s 1 1

s§ s
I o
1 1
B o
1

I I I I I

w
n .h a
I I

i! cq o o —

CO «3

!§55
 PRICING OF AUDIT SERVICES 183
TABLE 10
Average Fees Paid to Big Eight Auditors Deflated by ASSETS"
(Excluding Bank Respondents)
Number of FEE Mean -^ ASSETS • .Std
Observations Dev
Price Waterhouse & Co 44 16.4 9.1
Arthur Young & Co. 16 15.0 7.7
Non-Big Eight firms 108 14.1 8.8
Emst & Whinney 30 14 8.5
Deloitte Haskins & Sells 27 13 9.8
Arthur Andersen & Co. 57 12.7 8.8
Peat Marwick Mitchell & Co. 30 12..5 8.0
Touche Ross 20 11.8 5.6
Coopers & Lybrand 41 11.6 6.8

Note that the individual regression coefficients across the last three
columns of table 8 are similar. A formal test of the homogeneity of the
"small" and "large" auditee segments yielded an F-statistic of 1.3 with
(13,347) degrees of freedom and resulted in failure to reject (at the .05
significance level) the nuU hypothesis that the overall regression relation-
ship is homogeneous across the two auditee size categories.'^ However,
homogeneity of the audit fee regression across auditee size classes is not,
in itself, a sufficient test for competition, because possible auditee substi-
tution toward or away from internal accounting resources has not been
controlled. For the same reason, the uniformly negative coefficients of
AUDITOR-! and the uniformly positive coefficients of AUDITOR-PW
cannot be interpreted at this point.

15. Test of the Determinants of the Sum of Control Costs

In order to interpret the AUDITOR coefficients, it was necessary to
examine the behavior of total systems costs. Of the 397 observations, 333
respondents provided information on the variable ICOST. Respondents
who failed to provide the information were typically large, decentralized,
consohdated entities for whom the cost of collecting the data might likely
" The test statistic (Johnston [1972])is:
^ _ (SS, - SS2 - SSM
(SS2 -t- SS3)/(m2 -t- ms - 2*)
where
SS, = residual sum of squares for the total observations (excluding banks) regression
SSi = residual sum of squares for Sales < $125MAf regression
SS3 = residual sum of squares for Sales > $125MAf regression, and
* = 13(ntimher of variables); m; = 171; mj = 202.
Substituting:
p ^ (14997 - 6824 - 7488)/13 _ 52.7
(6824 -^ 7488)/(171 -h 202 - 26) " i T S ° ^'^^
184 DAN A. SIMUNIC

be high. Of the 333 complete responses, nineteen were from banks, which
reduced the available observations for the test of competition to 314.
Proceeding as in the analysis of the audit fee data, the regression of
ln(FEE H- ICOST) on in(ASSETS) yielded an estimated slope coeffi-
cient, e, of .55. This vjdue is sufficiently close to one-half to justify use of
the square-root transformation of ASSETS as the size deflator for each
cost component, and thus for their sum.
The least-squares' estimates of the coefficients of the variables in the
linear regression function:
FEE -I- ICOST
= bo -t- biSUBS -(- biDIVERS
ASSETS'
-t- [Link] -I- b,RECV-^• bJNV
H- bePROFIT -)- [Link] -t- bsSUBJ
-H b^TIME -I- buyAUDITOR-PW
-H buAUDITOR-7 -I- bnUTILITY -\- u
are displayed in table 11, together with various regression statistics. As in
table 8, the standard errors of the coefficients are shown in parentheses,
while coefficients which are significant at the .05 level are marked with
an asterisk. In these results, the sets of auditor coefficients are of principal
interest, with the remaining variables included solely for purposes of
control.
Finally, for completeness, I display in table 12 the least-squares' coef-
ficient estimates of the variables hypothesized to be determinants of
ICOST. The regression function is:
ICOST
bo -H biSUBS -I- b2DIVERS -I- b^FORGN
ASSETS'
-t- biRECV-i- bJNV+ bJkUDITOR-PW
+ biAUDITOR-! -\- b^UTILITY + u.
Results are presented using the same format as in table 11. Four obser-
vations were deleted from column 1 as outliers, representing very small
companies who used some internal auditors with very high values for the
constructed dependent variable.
A scan of table 12 shows that the regression results are, on the whole,
unsatisfactory. The low adjusted R' and the lack of significance and
inconsistent signs of many of the control variables suggest that the
determinants of ICOST are not correctly specified and/or that there is
significant error in the measurement of this variable.
 PRICING OF AUDIT SERVICES 185

TABLE U
Regression of {(FEE + ICOST^/ASSETS') on Explanatory Variables
160 Auditees with 154 Auditees with
Sales Less Than Sales Greater Than
Excluding Banks $12SMM $125MM

SUBS' 1.58* 1.34* 1.35*

(.21) (.43) (.28)
DIVERS 1.70* 2.22* 1.35*
(.41) (.66) (.54)
FORGN 12.88* 12.38* 12.09*
(3.41) (6.40) (4.12)
RECV 9.74* 7.39* 25.72*
(3.22) (3.67) . (6.79)
INV 8.73* 9.16* 7.71*
(2.52) (3.03) (4.41)
UTILITY -2.75 -1.36 -2.32
(2.17) (4.60) (2.68)
PROFIT .57 -.63 .37
(7.26) (7.77) (18.11)
LOSS .56 .89 2.18
(1.22) (1.41) (2.53)
SUBJ 2.88* 4.49* .25
(1.63) (2.19) (2.57)
LOG (TIME) 1.41 2.19 -1.54
(1.23) (1.63) (1.93)
5.71* 2.64 8.42*
AUDITOR-PW (1.57) (2.23) (2.34)
-.82 -1.58 -.02
AUDITOR-! (1.04) (1.27) (1.79)
Intercept 4.16 3.35 6.89
Std. error of the estimate 8.00 7.48 8.23
Adjusted R' .44 .26 .51
F 21.9 5.6 14.2
(FEE -t- ICOSTl/ASSETS^
mean 17.14 14.1 20.33
std. dev. (10.75) (8.7) (11.75)
* An asterisk indicates that the regression coefficient is significant at the .05 level in a one-tail or two-
tail (as appropriate) (-test. Standard errors of theregressioncoefficients are shown in parentheses.

16. Discussion and Interpretation of Results

Findings with respect to the hypothesized control variables are sum-
marized in table 13. The significance tests of the FEE determinants are
from column 2 of table 8 (recall that these regression results are generally
homogeneous across auditee size classes), while only those variables
whose coefficients are statistically significant in both columns of table 12
are listed as significant determinants of ICOST (these regression results
are not homogeneous across the two columns).
The contr61 variables for differences in loss exposure describe aspects
of the external audit and internal control environment. Note that the
selection of specific aspects of the environment as control variables
186 DAN A. SIMUNIC

TABLE 12
Regression of (ICOST/ASSETS^) on Explanatory Variables
166 Auditaea with 154 Auditees with
Sales Leaa Than Sales Greater Thwi
$125MM $125MM
SUBS'' .03* .04*
(.01) (.01)
DIVERS .25 .95*
(.19) (.31)
FORGN 4.55* -.11
(1.68) (2.37)
RECV .42 3.97
(1.01) (3.93)
INV -.10 .07
(.84) (2.59)
UTILITY .75 -.88
(1.26) (1.47)
AUDITOR-PW .81 6.03*
(.61) (1.37)
AUDITOR-! -.24 1.57
(.35) (1.04)
Intercept .76 2.24
Std. error of the estimates 2.09 4.86
Adjusted R^ .09 .25
F 2.8 7.5
Dependent variable
mean 1.34 6.77
std. dev. (2.19) (5.63)

T A B L E 13
Summary of Significance Tests on Hypothesized Control Variables
Significant Relationship (at .05 level)
with
Audit Fees Internal Audit Costs
Control variables for differences in loss exposure:
ASSETS Yes Yes
SUBS Yes Yes
DIVERS Yes No
FORGN Yes No
RECV Yes No
INV Yes No

Control variables for differences in expected loss shar-

ing ratio:
PROFIT No Does
LOSS Yes not
SUBJ Yes apply

Control variable for differences in auditor production Does

functions over time: not
TIME No apply
 PRICING OF AUDIT SERVICES 187

follows from the hypothesis that avoidance of third-party liability losses

motivates the design of auditee control systems. The fact that all variables
in this group (namely, ASSETS, SUBS, DIVERS, FORGN, RECV, and
INV) are statistically significant determinants of audit fees supports the
descriptive validity of this hypothesis with respect to the external audit
component of the system. However, the same variables are far less
successful in explaining cross-sectional variation in internal audit costs.
The latter result may indicate either that liability avoidance is not a
primary motivator in the design of internal systems, or the presence of
significant measurement problems.
The quantity of internal accounting control relevant to external finan-
cial reporting is a construct in the model whose empirical counterpart
can only be measured with substantial error. For example, internal
auditors are likely to be involved both in the verification of financial data
for external reporting and in monitoring operating efficiency and the
adherence to general management policies. Since the principal benefit
from these other activities is not liability avoidance, a failure to separate
these costs by various internal audit activities leads to misspecification of
the regression function for observed internal audit costs. Although further
research on this problem is necessary, the overall results reported here
nevertheless do not support a rejection of the hypothesis that liability
avoidance drives the design of financial reporting systems.
The control variables for differences in the assessed loss-sharing ratio
represent alternative measures of auditee financial distress. The insignif-
icance of the PROFIT variable and significance of the two categorical
variables, LOSS and SUBJ, suggest that E(d) and therefore p do not
vary continuously with the profitability of auditees. Rather, the auditor's
expected share of residual liability losses seems to increase only with
evidence of significant deterioration in the auditee's operations or pros-
pects.
The fact that audit fees were not found to vary systematicaUy with
TIME could indicate either that learning effects were "swamped" by the
noise in a cross-section, or that auditors pursue multi-period pricing
policies, in that they average the expected cost reduction of learning over
time. With such policies, lesumng effects could not be observed in fee
data.
For the test of competition, the test statistics are the AUDITOR
coefficients in table 11, interpreted using table 3. For both "large" and
"small" auditees and for the observations in total, the AUDITOR-!
coefficients are not significantly different from zero. Thus, the hypothesis
that price competition prevails throughout the market for audits of
publicly held companies cannot be rejected. Indeed, since average fees
for audits by seven of the dominant firms are lower throughout the
market than the fees of their non-Big Eight competitors (table 8) and the
fact that coefficients of AUDITOR-! in table 11 stiU tend to be negative.
188 DAN A. SIMUNIC

the results suggest that the Big Eight firms enjoy scale economies which
are passed on as lower prices to auditees. This is obviously an important
issue which deserves further research.
Finally, with respect to Price Waterhouse & Co. (PW), the significantly
positive coefficient of AUDITOR-PW in column 1 of table 11, which is
consistent across auditee size classes, indicates that the clients of this
firm, on average, use higher cost control systems. However, this is not
evidence of monopoly pricing by PW, since the obvious and best substi-
tute for a PW audit is not the employment of additional internal auditors,
but rather an external audit by some other Big Eight firm. Note that
both the separate audit fee (from table 8) and internal cost components
(from table 12) are differentially greater for PW auditees. In the regres-
sions, the PW classification vsu-iable may be a proxy for omitted variables
describing certain unknown differentiating characteristics of PW cUents
which affect the quantities (a, q) demanded by these companies. Thus,
auditees using PW may demand control systems of greater than average
qusdity. Alternatively, the significantly positive coefficient of ^4 UDITOR-
PW (relative to A UDITOR-!) in table 8 may represent a price difference
paid by auditees for a differentiated service. That is, a PW audit may
possess some utility-bearing characteristics to auditees which command
a positive implicit price in the market. While it is not possible to
distinguish between the two interpretations, clearly PW, PW chents, or
both are somehow differentiated from other class members.

17. Concluding Comments

One of the impediments to understanding the audit service is the
ambiguity of the relationship between auditors, audited companies, and
external financial statement users. Auditors are exhorted, in their codes
of ethics, for exeunple, to be independent and objective; yet they are hired
and compensated by auditees. The finding of the Commission on Audi-
tors' Responsibilities [1978] that "many users appear to misunderstand
the role of the auditor and the nature of the service he offers" is therefore
not surprising.
In the positive model of auditing developed in this paper, the essential
interdependence of the auditee's and auditor's economic interests is
recognized. The model allows for auditor independence in the sense that
the auditor implements ^ as a complement to the internal accounting
system, a. An auditee demands a positive quantity of auditing because
external auditors have some advantage (relative to internal systems) in
certain aspects of control. In the model, therefore, em auditor is indepen-
dent in the same sense as is any supplier of a service who seeks compe-
tently to perform a task demanded by a customer. The auditee and
auditor are not always adversaries. Although their economic interests
may diverge in ex-post htigation, as each tries to minimize losses, the
hypothesis regarding liability avoidance motivation implies that, at the
 PRICING OF AUDIT SERVICES 189

time of the audit, there is a mutuality in the auditee's and auditor's

private interests vis-a-vis the external world.
The question of whether the avoidance of third-party liability is a
dominant source of expected benefits which drives the design of auditee
financial reporting systems is an important one because of a possible
externality which can thereby result. Expected liability losses are a
private cost, and the legal system is the screen through which losses
suffered by third parties are filtered. An alternative would be to force the
auditee to internalize expected losses to users by relating the quality of
control system design to the market valuation of the auditee's firm. In
that way, variations in the quality of control systems as perceived by the
market would affect the risk-return characteristics of auditees' securities.
Note, however, that currently neither internal control costs nor external
audit fees are information which is publicly available to market agents.
Thus, under current arrangements, market reaction to alternative sys-
tems designs can only be a tenuous auditee motivator, at best. The
Securities and Exchange Commission recently did propose a requirement
that management disclose and discuss weaknesses in its internal controls.
But in the final version of Accounting Series Release No. 250, the SEC
withdrew a proposed rule that external audit fees be disclosed. One
argument in support of required disclosure of audit fees could be the
avoidance of potential externalities, to the extent that the audit fee is a
good measure of the quality of auditing purchased.
Finally, the failure to reject the hypothesis that price competition
prevails throughout the market for audits of publicly held companies
suggests that observed differences in Big Eight concentration across the
market may be essentially irrelevant. That is, concentration statistics, by
themselves, cannot support the allegation that the Big Eight firms are
monopolizing the market for audit services.
REFERENCES
ALTMAN, E. "Financial Ratios, Discriminant Analysis and the Prediction of Corporate
Bankruptcy." Journal of Finance (September 1968): 589-609.
BEAVER, W. Market Prices, Financial Ratios, and the Prediction of Failure." Journal of
Accounting Research (Autumn 1968): 179-92.
BERNSTEIN, P. "Competition Comes to Accounting." Fortune (July 17, 197H): 89-96.
COMMISSION ON AUDITORS' RESPONSIBILITIES. Report. Conclusions, and Recommenda-
tions. New York: Commission on Auditors' Responsibilities, 1978.
DEMSETZ, H. The Market Concentration Doctrine. Stanford, Calif.: American Enterprise
Institute-Hoover Institute on War, Revolution and Peace, 1973.
DEMSKI, J., AND R. SWIERINGA. "A Cooperative Formulation of the Audit Choice Problem "
The Accounting Review (July 1974): 506-13.
FIRTH, M. "Qualified Audit Reports: Their Impact on Investment Decisions." The Account-
ing Review (July 1978): 642-50.
HARRIS, S. P. Who Audits America? Menlo Park, Calif.: Spencer P. Harris, 1976.
HOBGOOD, G., AND J. SCIARRINO. "Management Looks at Audit Services" Financial
Executive (Apnl 1972): 26-32.
JOHNSTON, J. Econometric Methods. 2d ed. New York: McGraw-Hill, 1972.
190 DAN A. SIMUNIC

LENHART, N., AND P. DEFLIESE. Montgomery's Auditing. 8th ed. New York: Ronald Press,
1957.
OPPENHEIM, A. M. Questionnaire Design and Attitude Measurement. New York: Basic
Books, 1966.
ROSEN, S. "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Compe-
tition." Journal of Political Economy (January 1974): 34-55.
STICLEB, G. "The Economies of Scale." Reprinted in The Organization of Industry.
Homewood: Richard D. Irwin, 1968.
SUBCOMMITTEE ON REPORTS, ACCOUNTING AND MANAGEMENT OF THE COMMISSION ON
GOVERNMENT OPERATIONS, UNITED STATES SENATE. The Accounting Establishment:
A Staff Study. Washington, D.C.: Government Printing Office, 1977.
UNITED STATES OFFICE OF MANAGEMENT AND BUDGET. Standard Industrial Classifica-
tion Manual. Washington, D.C.: Government Printing Office, 1972.
WEISS, L. "Quantitative Studies in Industrial Organization." In Frontiers of Quantitative
Economics, edited by M. D. Intriligator. Amsterdam: North-Holland Puhlishing Co.,
1971.