Legal Liabilities, Audit Accuracy
and the Market for Audit Services
Sankar De and Pradyot K. Sen*
1. INTRODUCTION AND OVERVIEW
In recent years, liability lawsuits against auditors seem to have
reached an epic proportion. The liability claims against the big
six audit firms in the US is estimated to be about $30 billion,
which exceeds their total partner's capital by a sizable amount
(Balachandran, 1993). Several large jury decisions and out of
court settlements often get reported in the financial press. The
seriousness of the problem is appreciated by academics and
practitioners alike.
1Since the early 1990s, the practitioners felt
that the existing liability regime put an ever increasing burden
on the audit firms and their clients (Weinbach, 1993; Cook, 1993;
and Freedman, 1993). The litigation phenomenon is not limited
to the US alone. US audit firms routinely face lawsuits for work
done outside the US.
2Although the US remains the preferred
ßBlackwell Publishers Ltd. 2002, 108 Cowley Road, Oxford OX4 1JF, UK
* The authors are respectively from the Center for Professional Development in Finance, Berkeley, California and the Haas School of Business, University of California, Berkeley; and the University of Cincinnati. They wish to thank Bala Balachandran, Robert Chatov, Srikant Datar, Jere Francis, Angela Gore, Bob Hagermann, Jack Hughes, Bjorn Jorgensen, Ken Klassen, Rick Lambert, Bob Magee, Nahum Melumad, V. G. Narayanan, Zoe-Vonna Palmrose, Victor Pastena, Stefan Reichelstein, Kevin Sachs, Toshi Shibano, Brett Trueman, the workshop participants at University of California at Berkeley, SUNY-Buffalo, University of Iowa, Center for Applied Mathematics ± New Jersey Institute of Technology, New York University, University of Wisconsin-Madison, European Econometric Society Annual Meetings and European Economics Association Annual Meetings, several anonymous referees and the editor of this journal for their comments on different versions of this paper. Part of the work was done when both the authors visited the Indian Institute of Management Calcutta, India. Typing assistance of Rona Velte and Kathy McCord is greatly appreciated. (Paper received May 2000, revised and accepted January 2001)
Address for correspondence:Pradyot K. Sen, Department of Accounting and Information System, College of Business, 302 Lindner Hall, University of Cincinnati, Cincinnati OH 45211-0211, USA.
forum for litigation, audit litigation has begun to become
prevalent in the common law countries such as England,
Caribbean, Canada, Australia, New Zealand, and in many Asian
countries such as Hong Kong and Singapore. Even in countries
such as Japan and Germany, audit related litigation has become
more
common.
3The
Bank
of
Credit
and
Commerce
International (BCCI) litigation, perhaps involving the most
amount of money in any litigation, was filed in both Luxembourg
and the UK. It is believed that in countries such as New Zealand
and Australia, the rate of litigation, considering the size of the
economy,
far
exceeds
professional
malpractice
litigation
anywhere in the world (Grobstein and Liggio, 1993). Given the
differences in the liability levels and regimes in different parts of
the world, it is important to understand how a rational auditor
would respond to changing liability levels and regimes and
modify her efforts level and fees and how such choices would
influence the supply of audit accuracy and the demand for audit
services. This paper is an attempt to study this problem through a
model, where in a market characterized by asymmetric
information between the firm-managers and the shareholders,
firms demand auditing to signal their private information and,
anticipating the demand and responding to various liability levels
and regimes, rational auditors of different operating efficiency
choose the level of care (effort).
loss of business arising from a type 1 audit error, and potential
costs of legal liabilities arising from a type 2 audit error. In our
model, some audit firms are more efficient than others, due to
their operating cost structure, and this advantage in operating
cost structure of some firms is a common knowledge in the
market. However, all auditors are generally efficient in the sense
that they all generate favorable findings for a good client firm
with a higher probability than for a bad client firm. However, no
audit service achieves perfect audit accuracy (zero type I and type
II errors) because the costs involved may be prohibitive. In this
scenario, if more efficient auditors are subjected to a higher level
of liability and business loss arising from type I error, we
characterize an equilibrium where more efficient auditors also
provide a higher quality audit service and receive higher audit
fees.
5of the market, the audit fees, and decide whether to demand
high quality or low quality audit services.
7In this model, we show (in Section 4) that under fairly general
conditions, depending on the level of the differential in fees for
the two audit methods, three Nash Sequential Equilibrium (NSE)
outcomes, which are robust to refinements
8are possible:
1. For cost differential below a certain level, a pooling
outcome in which all firms choose high quality auditing
(DP).
2. For the differential exceeding that level, a semi-separating
outcome in which all good firms opt for high quality
auditing while bad firms randomize between the two audit
methods (DSS).
3. If the differential is too high, a third possibility exists where
both types forego high quality auditing.
We focus on the second equilibrium where all good and either
some or all bad client firms demand high quality audit services in
equilibrium, unless the differential in fees for high and low
quality services is too much. On the other hand, client firms who
choose low quality audit services are always bad themselves. The
intuition behind the second equilibrium is that as more and
more bad firms choose a high quality audit, the expected value of
such a firm in the pool is diluted and, given sufficiently high fees,
eventually no higher than what could be obtained with the help
of a low quality audit. As a result, in the second equilibrium, on
the margin, the bad firms become indifferent to the choice of
audit method and consequently, randomize between the two
audit methods. Interestingly, the proportion of bad quality firms
which choose high quality auditing in this equilibrium is unique
given the parameters of the system and, further, higher for a low
differential between the respective costs of the two audit
methods.
for audit purposes. Table 1 shows that, during 1982±1987, 82% of
all audited firms in the US used big eight audit firms.
9In the
higher annual sales range of $250 million and more, this
proportion was 95%. In terms of value, the big eight audit firms
audited about 97% of the total sales of all US audited firms in this
period, though the existing empirical evidence suggests that the
bigger firms also charge higher fees. Though several reasons have
been suggested for this overwhelming preference for the services
of the big audit firms in spite of their higher fees, there is no
formal model of demand for audit services that explains this
phenomenon in the existing literature on auditing.
10Description of Clients for Big Six and Non Big Six Audit Firms, 1982±1987
Number of Clients Total Sales Audited Number of Companies Audited
($billions) 1±25 mill.c 26±50 mill. 51±100 mill. 101±250 mill. 250 mill and over
Year B6a NB6b B6 NB6 B6 NB6 B6 NB6 B6 NB6 B6 NB6 B6 NB6
1987 6942 1336 3664 111 2943 911 833 119 765 135 855 88 1546 73
1986 6716 1457 3123 105 2553 862 815 156 776 121 839 94 1433 78
1985 6543 1417 3525 104 2691 986 770 135 791 112 794 104 1488 78
1984 6488 1452 3465 104 2694 912 772 137 764 116 779 107 1479 71
1983 6173 1572 3170 104 2566 1097 766 171 699 115 765 109 1377 71
1982 5727 979 3290 96 1879 815 684 102 638 88 788 39 1468 75
Yearly
Average 6430 1369 3373 104 2554 930 773 137 739 115 689 90 1465 74 Average
% 82 18 97 3 73 27 85 15 87 13 88 12 95 5
Notes:
a B6 represents big six b NB6 represents non big six.
c This represents the range of company size audited in terms of million dollar (US).
Source: Computed using data from various issues ofWho Audits America, The Data Financial Press.
DE
AND
SEN
ß
Blackwe
ll
Publi
shers
Ltd
auditor's efforts; and therefore, provides incremental value to the
client firms. So far, the literature on audit markets seems to have
concluded that the joint effect of legal liabilities, as manifested
through these two factors, is at best, an empirical issue and may
even offset each other in determining the demand for audit
services (see for example, Lys, 1993).
However, since any penalty paid by the auditor can possibly
reduce the penalty to be paid by the insider, auditor penalty has
an insurance role in the Hillegeist model and adversely affects
the owner's disclosure behavior. We do not explicitly model any
penalty to the insiders. Taken together, our study and the
Hillegeist study provide some confirming and complementary
findings on the issue of audit quality choice and auditor liability.
Other studies that have addressed the issue of demand for
auditing, particularly in special contexts such as new issues
markets, include Titman and Trueman (1986), Datar, Feltham
and Hughes (1990), Bachar (1989) and Sarath and Wolfson
(1989). Though they contribute useful insights, the scope of these
studies is different from ours.
13In an insightful study, Melumad
and Thoman (MT) (1990) address some of the same issues
exam-ined in the present study. However, there are important
differences between the two models as well as the results derived
from them. In MT, the manager can disclose the true type apart
from selecting the auditor. In our model, adverse selection
problems rule out direct disclosure of type. Rather, the audit
findings together with the choice of the auditor constitute an
(imperfect) signal of firm type. Further in MT, the auditor makes
no error while auditing good client firms. Thus, a `low' report
unmistakably identifies a bad client firm, leading to a separating
result in their scenario.
14This is different from and less general
than our model where the audit is imperfect for both types. We
have shown that, if the probability of an error in auditing good
client-firms is non-zero, though small, a separating equilibrium
cannot exist. Most importantly, in either of the two viable
equilibrium outcomes with a high quality audit pool in our study
(named DP, for Differentiated Pooling and DSS, for Differentiated
Semi-Separating), the firms of different types in the same pool
ex
ante
expect different market outcomes which are imperfectly
related to their true firm types. As a result, for sufficiently high
audit costs, though the same for both types of client firms, high
quality auditing may be economical for good client firms but
uneconomical for bad client firms. This feature of our framework
enables us to eliminate a number of un-intuitive equilibrium
outcomes. Unlike our model, there is no intuitive correspondence
between audit costs across different equilibria in MT.
2. THE NATURE AND SUPPLY OF AUDIT SERVICES
In this section we sketch a model of auditor's maximization
problem that is consistent with the observed reality. It is well
recognized that audit services are heterogeneous commodities
15and that some audit firms, perhaps the larger ones, provide
higher quality audit services than others, achieving lower error
rates (higher accuracy).
16Further, as Feltham, Hughes and
Simunic observe:
. . . the literature on audit quality . . . has only demonstrated the existence of two or perhaps three distinct auditor quality levels (1992, p. 377).
Thus, we model the difference in audit quality by assuming the
existence of two separate types of audit firms, who differ in their
operating efficiencies. We show that the economic incentives of
the audit firms leads to supply of two levels of audit qualities
corresponding to their respective efficiency levels. Thus, an audit
firm does not merely assert that it is offering a `high' or a `low'
quality service. Given their efficiency types, their incentives are
well understood and the quality of the services they supply are
rationally anticipated by the market participants. For the
signaling equilibrium analyzing the demand side, we focus on
the fact that two audit services of different qualities, which are
priced differentially, exist simultaneously in the market. Both
audit services are informative, one more so than the other in a
sense to be made precise below. Neither of them, however, is
perfectly informative.
17(i) The Audit Firm's Role
The audit firm in our scenario provides a third-party evaluation
of the financial position of the client firms.
18There are
n
firms
which are potential audit clients. The insiders of each firm know
its true financial position which, however, is not directly
observable to the outside market. The financial position for any
client firm, denoted by
a
, can be either type 1 (good) or type 2
(bad). The types
t
"f
1
;
2
g
T
denotes the two types such that
type 1 is the better of two types. Under full information, the
market would value
a
1and
a
2as
V
a
1
and
V
a
2
19such that
V
a
1
>
V
a
2
. Without full information, the market must use
Audit firms often perform more than their duties prescribed
under the Generally Accepted Auditing Standards (GAAS).
Cushing and Loebbecke (1986) observe that:
many improvements in auditing standards and techniques are reflected in the pronouncements of the AICPA . . . some time after such improvements are made in the practices of individual firms.20
Senior members of the big six firms have often asserted that their
firms observe a higher standard than that which is required by
law.
21Since the audit firms need only to conform to the
Generally Accepted Auditing Standards (GAAS) for their audit
methods, regulation in the audit market cannot be the source of
such product differentiation. To wit, consider the following
quote from Elliot (1993):
There is clear undeniable evidence of demand for audit services before (security) laws were passed, and there's enormous demand today for audits that aren't required by law. My own firm (KPMG Peat Marwick), for example, does more voluntary audits than statutory required audits ± by an order of magnitude.
Providing a higher quality audit must therefore be in the best
interests of the concerned audit firms. Our auditors are strategic
in that they anticipate demand from the particular firm types and
provide an optimal level of effort consistent with their best
interests.
22However, they are independent in that they report
their findings truthfully.
23Audit quality (defined more precisely
later) increases with effort.
(ii) Auditors' Cost Functions
Let A and B represent two different audit services or methods
m
;
m
"
fA
;
Bg
, where
A
is the higher quality method. The
suppliers of high quality audits are more cost-efficient than the
suppliers of low quality audits in that for a given level of effort,
higher quality auditors produce lower error rates. Since any audit
firm in our scenario supplies only one type of service, we shall use
increase with effort. The quality differential is characterized by
the difference function
e
C
Be ÿ
C
Ae
such that
0e
>
0
and
0
0. Define
C
0and
C
00to be the first and second
derivatives of the
C
:
function with respect to its argument. We
also assume that
C
0and
C
00functions are positive everywhere for
both auditors and both efforts types.
(iii) Audit Fees
We assume that the cost of audits to the client firm is fixed for
each method (
A
and
B
) and are equal to
k
Aand
k
B. Adequate
supply of both types of auditors and perfect competition among
them would ensure that both types of auditors break even across
firm types. The fees
k
Aand
k
B, thus, must allow each type of
auditor to recover their expected costs. The fee differential
between the high quality and low quality audits is a sum of the
operating cost differential, the expected liability cost differential
and the cost differential arising out of the loss of business due to
type I errors. Unless the higher quality audit commands higher
fees, incentives to supply higher quality audits would be
questionable. Therefore, we require that in equilibrium, the
high quality auditors are paid higher fees and the corresponding
fees
k
Amust be greater than
k
B. We are primarily concerned with
the differential in the fees for the two audit services, denoted by
K
where
K
k
Aÿ
k
B. To the extent high quality audits are
offered by the bigger firms as we mentioned earlier,
k
A>
k
Bis
consistent with the existing empirical evidence.
(iv) Audit Report and Audit Accuracy
After carrying out an audit program, the audit firm which provides
the higher quality audit service (
A
) comes up with a finding(
f
)
about the client firm which can be either `high' (HA) or `low'
(LA). Thus,
f
"f
HA, LA
g
F
. Similarly, the audit firm which
supplies the lower quality audit service (
B
) also issues its finding
(
g
) after its audit which again can be either `high' (HB) or `low'
(LB). Thus
g
"f
HB, LB
g
G
. The auditors' reports are seen by all
parties in this scenario. The probability of a `high' assessment by
an auditor of quality level
m;
m
2 fA;
Bg
, for a client firm of type
view
q1m
and
q2m
as indicators of the level of audit accuracy i.e., the
ex-ante probability that an auditor makes a correct report, for
audit method
m;
m
A;
B. Specifically, a high
q1m
indicate a low
type 1 error and a low
q2m
indicate a low level of type II error.
Given that the auditing standards ensure a basic level of auditor
efficiency, we view
q
tmas being determined by the auditors' efforts
e. Specifically,
q
1me
m
and
q
2me
m
are the two error levels chosen
by the auditors through their effort choices. We assume
qtm
em
to
be such that accuracy increases with effort but at a decreasing rate.
Formally,
q
1m0>
0;
q
1m00<
0 but
q
2m0<
0;
q
2m00>
0, where
0and
00respectively denote the first and second derivatives of the
respective functions. Further,
q
1m1
1 and
q
2m1
0. This
implies that perfect audit accuracy is impossible to achieve and
that audit failures will occur. For the sake of notational
convenience, henceforth we suppress the argument
em
from
q1m
and
q
2m. For the equilibrium analyzed in the demand section, weneed that the high and low quality audits are well defined and
require that
q1A
>
q1B
and
q2A
<
q2B. As we show below, this is
exactly what happens. Also for the low quality audit to have value,
for the equilibrium level of effort
e
B, we assume that
q1B
>
q2B
.
(v) Consequences of Audit Failure
An audit failure occurs when auditors report
H
L
findings for a
type 2 (1) client firm. If a type 1 firm is assigned
L, an auditor of
type
m
risks loss of business, denoted by
l
m. If, on the other hand,a type 2 firm is assigned
H, the auditor risks being sued by the
investors in the firm (and found guilty of audit failure by the
court) who claim to have been misled by the audit findings. The
expected cost of this (type II) error is represented by the penalty
function
P
m.24If
12
proportion of the type 1 (type 2) client
firms engage the type
A
auditors, the probability of a loss 1A
1B
must be
11
ÿ
q1A
(equivalently,
1
ÿ
1
1
ÿ
q1B
). Similarly,
the probability of a loss
P
AP
B
must be
2q
2A(equivalently,
probability that auditors with deeper pockets will face a higher
expected penalty for an audit failure than do other audit firms. If
the larger and more established audit firms are also the ones with
deep pockets, and since the size of an audit firm and audit quality
are correlated (DeAngelo, 1981), it follows that auditors of higher
quality may face higher expected costs due to type II errors.
Therefore, we assume that for non-zero
2in equilibrium,
independent of efforts, the higher quality auditor is exposed to
a higher level of expected penalty, i.e.,
2:
P
A1
ÿ
2
:
P
B.
25At
first we assume that
P
mis exogenous. An assumption of
exogenous
P
mseems to be consistent with a standard of strict
liability. Though the existing laws do not specify a standard of
strict legal liability, the existing rule of due-diligence defense is
becoming, in many respects, equivalent to that of a strict liability
(Nelson, Ronen and White, 1988). Minow (1984), Chatov (1987),
Gormley (1988) and Kripke (1988) among others have
documented the evolution of an increasingly popular strict
liability interpretation of auditors' liability by courts, especially
within the Securities Act of 1934. Later, we relax this assumption
and consider alternative regimes where
P
mis determined by some
audit outcome, and thus, is partially controllable by the auditor
through increased effort. Comparison with a fixed and exogenous
P
mthen serves as a benchmark.
(vi) The Auditors' Maximization Problem
The supply of audit quality is determined by the auditors'
maximization problem. In our model, both types of auditors are
assumed to be risk neutral. The type
A
auditors maximize their
income in the following way:
26Max
ek
Aÿ
C
Ae
A ÿ
11
ÿ
q
1A
l
Aÿ
2q
2A:
P
A:
A
Similarly, the type
B
auditors maximize their income in the
following way:
Max
ek
Bÿ
C
Be
B ÿ 1
ÿ
1 1
ÿ
q
1B
l
Bÿ 1
ÿ
2
q
2B:
P
B:
B
C
A0e
A
1q
1A0l
Aÿ
2q
2A0:P
A
01q
1Al
Aÿ
02q
2A:P
A;
C
B0eB
1
ÿ
1q
1B0lB
ÿ 1
ÿ
2q
2B0:PB
ÿ
01q
1BlB
02q
2B:PB
:
The above first order conditions can be interpreted as equating
the marginal operating costs with the marginal benefit from
reduction of penalty and business losses after considering the
impact on change in demand.
28Although
1,2,01and
02are decided in equilibrium, we know
that
1
2
1 is the maximum values of
1and
2and
therefore, in the equilibrium if
1
2
1, both
01and
02must
be equal to zero. For the sake of tractability, we assume that both:
@
21@e
A@e
Band:
@
22@e
A@e
Bare also zero.
We now note two important requirements with respect to the
supply of the audit services that must be satisfied in any proposed
equilibrium. The first is that the optimal solution of the auditors
effort choice problem must satisfy
e
A>
e
Bimplying that,
motivated by self interest, the more efficient auditors supply a
high quality audit service. If so, a meaningful inference about
audit quality can be derived from the knowledge of auditor
efficiency by the market participants. The second requirement is
that the fees of a type
A
auditor will be greater than the fees of
the type
B
auditor. Both these conditions form part of the
equilibrium conditions developed in the following section.
3. THE ECONOMIC ENVIRONMENT AND DEMAND FOR AUDIT SERVICES
(i) Basic Information Structure
To communicate the true financial position of their firm to the
rest of the market, the insiders may engage either of two kinds of
available audit services, high quality (
A
) (offered by the more
efficient auditors) and low quality (
B
) (offered by the less
efficient auditors), described in the previous section. This choice
transmits a message
m
"
fA
;
Bg
. An audit is mandatory;
therefore no audit is not a possible message. Since these two
types of audit services are provided by two distinct classes of audit
firms, this message (audit method) can be observed by the rest of
the market.
(ii) The Outsiders' Role
Besides the informed managers and the auditors, the other
players in this game are the outside investors who decide on the
market values of the client firms. The market beliefs and market
values in our model are conditioned by both what the client firms
convey through their choice of the audit method and what the
audit firms report through their findings. This feature makes our
structure somewhat different from a standard signaling game.
The market value of a client firm from a high quality audit
V
A
following the release of the audit report is a function of the
managers' message of choosing high quality auditing (
A
) and the
audit findings (
H
or
L
). Thus,
V
A"
fV
HA;
V
LAg. Similarly, the
market value of the firm from a low quality audit is
V
B"
fV
HB;
V
LBg
. Given that the audit services
A
and
B
are
imperfectly efficient, neither
V
HAnor
V
LAif the firm chooses
A
,
and neither
V
HBnor
V
LBif the firm chooses
B
, may exactly reflect
(iii) Optimization Program of Client Firms
We assume that all players in the game are risk neutral. The
objective function of the insider-manager of a client firm in this
scenario consists of a weighted average of the true value of the
firm and its market value.
29Let
V
t
;
m
denote the expected
market value of a client firm of type
t
at the start of the game
(prior to the release of the audit report). Then, for
m
A
;
B
:
V
t
;
A
q
tAV
HA 1
ÿ
q
tA
V
LAÿ
K
;
and:
V
t
;
B
q
tBV
HB 1
ÿ
q
tB
V
LB;
where
K
k
Aÿ
k
Brepresents the differential between the audit
fees required for the two audit services and
q
tm's are obtained
from the audit firm's optimization program (see Section 2
above).
If
is the weight attached to the market value and
1
ÿ
to
the true value of the firm by the insider-manager, the objective
function,
W
V
t
;
m
, is:
W
V
t
;
m
V
t
;
m
1
ÿ
V
a
t;
m
"f
A
;
B
g;
t
"f1;
2g:
The insider-manager chooses
m
or the type of audit to maximize
the above objective function.
Figure 1 presents the game tree described above. Although so
far we have presented the game in terms of pure strategies,
A
or
B
, our analysis of the game allows for mixed strategies as well.
Denote by
tthe probability with which client firms of type
t
send
a message
m
A
. Intuitively,
tcan be viewed as the proportion
of type
t
firms that choose a high quality audit.
(iv) Nash Sequential Equilibrium
(NSE)
We use the concept of sequential equilibrium developed by
(Kreps and Wilson, 1982) and adopt the definition in our setting
in the following way. A Nash-Sequential Equilibrium (NSE) , is an
ordered
set
t
; ;
e
A;
e
B;
k
A;
k
B
satisfying
the
following
conditions:
C1. Client firms choose
t2
argmax
f
W
V
t
;
tg;
t2 0;
1;
Figure 1
C2. Audit firms choose their efforts
e
such that a type
A
audit
firm chooses:
e
A2
argmax
fk
Aÿ
C
Ae
a ÿ
11
ÿ
q
1Al
Aÿ
2q
2A:
P
Ag
and a type
B
audit firm chooses:
e
B2
argmax
fkB
ÿ
CB
eB
ÿ
1
ÿ
1
1
ÿ
q
1BlB
ÿ
1
ÿ
2q
2B:
P
Bg
:
C3.
k
A;
k
Bare such that both types of audit firms break even
in equilibrium.
C4. Quality and prices of the two audit supplies are well
defined; i.e.,
e
A>
e
Band
k
A>
k
B.
C5. Outside investors, in a perfectly competitive market,
choose
V
t
;
t
so that they break even across the
unobserved types of the client firms, implying:
X
t
V
t
;
m
;
1;
2
V
a
1
1
ÿ
V
a
2
:
C6. If
m
;
m
A
;
B
, is an equilibrium action, market posterior
beliefs following an
H
finding
tjHm
;
1;
2
, and
market
posterior
beliefs
following
an
L
finding
tjLm
;
1;
2
, are determined by Baye's rule in the
following manner:
1
jHm
;
1;
2
q
1m 1mq
1m1m
q
2m2m1
ÿ
;
2
jHm
;
1;
2
q
2m 2m1
ÿ
q
1m1m
q
2m2m1
ÿ
:
1
jLm
;
1;
2
1
ÿ
q
1m
1m1
ÿ
q
1m
1m
1
ÿ
q
2m
2m1
ÿ
;
2
jLm
;
1;
2
1
ÿ
q
2m
2m1
ÿ
1
ÿ
q
2m
2m1
ÿ
1
ÿ
q
1m
1mwhere:
tm
tif
m
A
and
1
ÿ
tif
m
B
.
Further,
q
tm's follow properties outlined in Section 2(
iv
)
and are obtained from the audit firm's effort choice
outlined in C2.
4. DEMAND FOR AUDIT SERVICES: EQUILIBRIUM RESULTS
(i) Some Basic Issues
Before we present our equilibrium results, we intuitively discuss a
few technical issues and results that we need for developing our
equilibrium results.
For any well defined equilibrium, we require that
e
A>
e
Band
k
A>
k
B, which is guaranteed by the supply side results developed
in Section 2. For all
e
A>
e
Bthe properties of the
q
tm:
function
would assure that 1
>
q
1A>
q
1Band
q
2B>
q
2A>
0. Further, for a
type
B
audit to be informative, we require that
q
1B>
q
2B.
Combining the restrictions and allowing for weak inequality, we
have:
1
>
q
1Aq
1B>
q
2Bq
2A>
0
;
. . . .
(A1)
where at least one of the two inequalities holds strictly. The above
accuracy structure implies two conditions. First, both auditors
A,
and
B, are generally efficient but not perfect in that both auditors
are more likely to report a `high' finding for a type 1 firm than
for a type 2 firm. Second, type
A
auditors are more accurate than
type
B
auditors. We also denote
K
k
Aÿ
k
B, the fee differential
in an equilibrium, which plays a prominent role in our analysis.
In Section 3,
V
Hmand
V
Lmare defined as the market value of a
client firm conditional on the firm drawing
H
and
L
findings
respectively, given an audit method
m;
f
m
2
A;
B
g
. When
1proportion of type 1 firms and
2proportion of type 2 client
firms choose
m, define
V
Hm1
;
2
and
V
Lm1
;
2
as the
corresponding values. Risk-neutrality on the part of the capital
market participants implies that:
V
Hm1
;
2
V
a
1
1
j
Hm
;
1;
2
V
a
2
2
j
Hm
;
1;
2
and:
V
Lm1
;
2
V
a
1
1
j
Lm
;
1;
2
V
a
2
2
j
Lm
;
1;
2
:
Note that
V
a
1
>
V
Hm1
;
2
and
V
Lm1
;
2
>
V
a
2
.
Similarly, define
V
t;
m
;
1;
2
as the expected outcome for a
client firm of type
t;
t
1
;
2, if it opts for an audit method
m,
given that
1proportion of type 1 firms and
2proportion of type
2 firms choose
m. Further, define
W
V
t;
m
;
1;
2
as the
posterior probability functions embedded in the values indicate,
given
t
and
m
the values depend on three sets of parameters:
f
t;
t
1
;
2
g
,
f
t;
t
1
;
2
g
and
fq
tm;
m
A
;
Bg
. While the first
set is drawn by the nature so to speak, the other two are not.
Changes in the parameters may affect the type-specific values
differently. This raises the following questions: whether, and
under what conditions, are the expected outcomes from the
same
audit method different for the two different types of client firms?
Further, whether, and under what conditions, are the expected
outcomes of a client firm of a given type different under the two
different
audit methods? These questions are important and are
addressed through a series of technical observations that are
necessary for the proofs of the propositions and detailed in the
Appendix.
(ii) Equilibrium Results
Since firms of each type can choose either
A
or
B
or randomize
between the two strategies, there are nine type-strategy
combinations that are candidates for an equilibrium outcome.
At first we show that unlike Melumad and Thoman (1990), no
separation of types is possible in our model.
Proposition 0.
There is no separating NSE. Specifically, neither
of the following two situations is NSE.
(I) Type 1 chooses high quality auditing
1
1
and type 2
chooses low quality auditing
2
0.
(II) Type 2 chooses high quality auditing
2
1
and type 1
chooses low quality auditing
1
0.
Proof.
See Appendix.
(1) type 1 firms strictly prefer high quality auditing
1
1
if
type 2 firms prefer high quality auditing or randomize
between
high
quality
and
low
quality
auditing
0
<
2
1; and
(2) type 2 firms choose only high quality auditing
1
if
the differential between the fees of the two audit
methods, given by
K
where
K
k
Aÿ
k
B, is sufficiently
low; randomize between the two pure strategies
0
<
2
<
1
if
K
is higher; and avoid high quality auditing
altogether
2
0
if
K
is very high (as we shall see below,
K
will have to be unrealistically high to make this
happen).
The statements in (1) and (2) above together imply that, for
the most part, in equilibrium
1
1 and 0
<
21. In these two
cases the quality and prices of audit supply are well defined. In
equilibrium, given
q
1Aq
1Bq
2Bq
2Awith a strict inequality
holding in at least one case, type 1 firms benefit more from
higher efficiency associated with high quality auditing than from
low quality auditing. If type 2 firms expect the same outcome
from the two audit methods,
A
and
B, inducing them to
randomize between the two strategies, type 1 firms will strictly
prefer
A
with its higher expected outcome. Of course, if type 2
firms prefer
A
to
B, it follows that type 1 firms will prefer it even
more so. This explains the preference of a type 1 firm for high
quality auditing. As for a type 2 firm, it expects less from
A
than
from
B
when
1
2
1,
A
being more informative than
B
.
However, if it switches from
A
to
B
, it is liable to fetch
V
a
2
or its
true value; since type 1 firms do not choose
B
, any firm choosing
B
is deemed a type 2 firm and is valued accordingly. If the
differential in audit fees are sufficiently low, its expected
outcome from
A
may still exceed
V
a
2
. This results in the
equilibrium strategy combination
1
2
1
.
If audit fees are higher, the expected outcome net of costs for a
type 2 firm from high quality auditing may be no higher than
V
a
2
, making it indifferent, and consequently randomize,
between the two strategies at the margin. This results in the
equilibrium strategy combination
1
1
;
0
<
2<
1
. It is also
true that given the demand pattern, audit firms' supply decision
of audit quality and the break-even condition results in higher
fees for a type
A
audit. In this equilibrium the firms which choose
high quality auditing consist of all type 1 and some type 2 firms.
On the other hand, all firms which opt for low quality auditing
are type 2. This equilibrium is semi-separating in that sense.
How high should the differential audit fees be so that no type 2
firm chooses
A
, making
2
0? For this to happen,
V
a
2
must
exceed the expected outcome from high quality auditing net of
fees which in this case is
V
a
1 ÿ
K
(if no type 2 firm chooses
A
,
any firm which chooses
A
commands the valuation appropriate for
a type 1 firm). However, if
V
a
2
V
a
1 ÿ
K
, it is not worthwhile
even for type 1 firms to employ this method. In other words, audit
fees must be so high as to make all firms reject a high quality audit,
an implication which contradicts what is observed empirically (see
Table 1). Given no demand for high quality auditing, in this case,
it is also not clear whether the audit fees and quality will be
differentiated in our model. The only two possible equilibrium
outcomes in our model are (1) a differentiated pooling (DP)
equilibrium where all firms opt for
A
or high quality auditing if the
fee differential between the two services is below a certain level; (2)
a differentiated semi-separating (DSS) equilibrium, if the
differential is above that level but not too high. In the two
equilibria, the
A
-pool is differentiated in the sense that a type 1
Proposition 1.
(1) For 0
K
K
, where
q
2AV
HA1
2
1
1
ÿ
q
2A
V
LA1
2
1 ÿ
V
a
2
K
, the DP equilibrium
1
2
1
is an NSE and robust to Universal Divinity. (2)
For
K
>
K
>
K
, where
V
a
1 ÿ
V
a
2
K
, the DSS
equi-librium
1
1
;
0
<
2<
1
is viable and is unique.
Proof
: See Appendix.
In the second equilibrium, both actions,
A
and
B
, are observed.
Proposition 1 demonstrates that the first equilibrium is robust to
the strong Universal Divinity refinement due to Banks and Sobel
(1987) and, therefore, survives weaker refinements such as
Cho-Kreps Intuitive criterion. The proof of uniqueness is
straight-forward though tedious. Instead, we present below a graphical
representation of the equilibrium condition in Figure 2 that
brings out the intuition behind the uniqueness property.
The equilibrium pool valuation expected by a type 2 firm can
be expressed as a quadratic expression in
2and can be
graphically represented as the downward sloping curve. The
equilibrium would exist when a type 2 firm becomes indifferent
between this value and its true value
V
a
2
, which it will receive
upon correct identification by the market. That level of
2is
represented by the point of intersection of the horizontal line
(representing the level of
V
a
2) and the downward sloping (in
pool) valuation line. As the graph indicates, this value is always
unique. The formal proof of uniqueness is available from the
authors upon request.
30(iii) Other Equilibrium Candidates
criterion.
31This should be contrasted with the equilibrium result
in Proposition 1 which, as we have noted above, is robust to the
much stronger Universal Divinity refinement.
Finally, we note that whether (i)
2
1, or (ii)
2
0, or (iii)
0
<
2<
1, it is never optimal for a type 1 firm to employ a mixed
strategy 0
<
1<
1, ruling out (7), (8) and (9) in Table 2.
Further, if type 2 randomizes, type 1 strictly prefers
A
. This rules
out (6) in the table. To see the intuition behind cases (i) and (ii)
above, note that when type 2 firms choose a pure strategy, say
A
, a
type 1 firm can have itself identified as indeed a type 1 firm by
choosing the other strategy (
B
in this case) and command a
higher market value. It will, therefore, never be indifferent
between the two strategies. The intuition behind case (iii) above
is straightforward. Given
q
1Aq
1Bq
2Bq
2Awith a strict
inequality holding in at least one case, type 1 firms stand to
benefit more from higher efficiency associated with high quality
Figure 2
auditing than do type 2 firms. If type 2 firms expect the same
outcome from the two audit methods,
A
and
B
,
32inducing them
to randomize between the two strategies, type 1 firms will strictly
prefer
A
with its higher expected outcome Note that this rules
out both (6) and (9) in Table 2.
5. IMPACT OF ACCURACY AND FEES ON DEMAND FOR AUDITING
(i) Social Value of Auditing
In our setting, the social value of auditing exists because both
types of auditing have some discriminating power and a type 1
firm expects to gain from the valuation induced by the audit
report.
33Allocational distortions are totally eliminated if
q
1
1
and
q
2
0, i.e., if the auditing is perfect. In this case,
1
jH
2
jL
1. In this situation, similar to Melumad and
Thoman (1990) a separating equilibrium in which type 1 firms
choose a high quality audit while type 2 firms choose the
minimum cost alternative is feasible because, with perfect audit
efficiency, type 2 firms can no longer hope for mis-classification.
Thus, if auditing is mandatory, type 2 firms choose the low quality
audits. In this case type 1 firms will expect a market value of
V
a
1 ÿ
K
Awhile type 2 firms will expect
V
a
2
. As we show in the
following Proposition 2, the gain from auditing is an increasing
function of the accuracy
q
1and a decreasing function of the type
2 error
q
2. Since
q
1q
2
is increasing (decreasing) in auditors'
efforts
e
, it follows that the gains from auditing is increasing in
the auditors' efforts
e
.
34Proposition 2.
The gain from both types of auditing,
G
m, to a
type 1 firm in equilibrium is an increasing function of
q
1mand
a decreasing function of
q
2m:
@Gm
@q
1m>
0 and
@Gm
@q
2m<
0
;
where
G
mq
1mV
Hm1
1
;
2
1
ÿ
q
1mV
Lm1
1
;
2 ÿ
K
m ÿ V
a
1
1
ÿ
V
a
2:
Proof
. See Appendix.
(ii) Impact of Audit Fees on Audit Demand
The equilibrium results established in Proposition 1 above
indicate that, as the differential in fees between a high quality
audit and a low quality audit, represented by
K
, increases,
demand for high quality auditing decreases. For
K
in the low
range
0
K
K
, where
K
q
2AV
HA1
2
1
1
ÿ
q
2A
V
LA1
2
1
ÿ
V
a
2
from Proposition 1 above, all type 1
and type 2 firms choose a high quality audit. For
K
in a higher
range
K
>
K
>
K
, where
K
V
a
1 ÿ
V
a
2
, though all type 1
firms continue to use high quality audit services, not all type 2
firms do so. However, if
K
is in the very high range
K
>
K
, all
firms dispense with high quality auditing.
The fact that an increase in high quality audit fees reduces
demand for high quality auditing relative to low quality auditing
is not surprising. To that extent, high quality auditing behaves
just like any other economic good. Note, however, that this
demand is defined over a domain in
K
which is bounded from
both above and below and is downward sloping over this range.
Though all type 1 firms demand high quality auditing over this
range, the proportion of type 2 firms that demand high quality
auditing over this range is low for high audit fees. To see this
intuitively, the equilibrium condition for type 2 in Proposition 1,
given by
q
2AV
HA:;
2<
1
1
ÿ
q
2A
V
LA:;
2<
1
ÿ
V
a
2
K
,
is decreasing in
2, with the result that a higher
K
entails a lower
2(Figure 3).
(iii) Impact of Audit Accuracy on Audit Demand
underlying differential in audit fees,
K
, is such that demand for
high quality auditing may exist in the first place: in other words,
K
<
K
V
a
1 ÿ
V
a
2
.
Proposition 3.
When
K
<
K
V
a
1 ÿ
V
a
2
, as the accuracy
of high quality auditing increases, the demand for high quality
auditing decreases.
Proof
. See Appendix.
The above results have implications for the incentive of an audit
firm to provide high quality audit services. In our model, the level
of audit accuracy and the operating costs of providing accuracy
Figure 3
are both increasing in audit efforts, indicating that higher
accuracy requires higher investigation costs. This can also be
easily verified from the first-order condition for optimization by
the audit firm. On the other hand, we have just seen above that
the levels of fees as well as the level of accuracy are negatively
related to the demand for audit services. Surely, the negative
impact of audit fees and audit accuracy together on the demand
for their services could be a disincentive for the audit firms to
initiate costly efficiency-improving changes in their technology.
35This is a problem, because higher audit accuracy is socially
desirable. It improves the expected outcome for type 1 firms while
reducing the outcome for type 2 firms and, therefore, reduces
allocational distortions. Given this problem for the audit firms,
the role of legal liabilities, that is supposed to provide incentives
for increased auditor efforts becomes especially important.
6. THE ROLE OF LEGAL LIABILITIES IN MOTIVATING AUDITORS
Our result from the demand side that demand declines in
accuracy suggests that perhaps individual firms, left to
themselves, may lack sufficient incentives to provide socially
optimal levels of effort and accuracy. In this context, we discuss
the potential effectiveness of the legal system in correcting audit
failures and improving auditor performance. We are here
concerned with instances of audit failure and liabilities as under
the Securities Acts of 1933 and 1934 and common law of torts
rather than audit fraud.
36Recall that in our scenario the auditors
report their findings truthfully.
reality the bigger (and presumably better) auditors are subjected
to relatively higher penalty for the reason of `deep pocket' and
for the fact that they are often the only solvent party around
(Kothari et al., 1988). Thus it is interesting to examine the
impact of such a `deep pocket' award on auditors' effort choice
problem in a strategic auditing setting which we do next. We
then consider two alternative legal regimes; the negligent liability
regime with a specified `due care' standard where the liability is a
function of the level of care exercised during the auditing
process and a strict liability regime where the legal liability is
based on the loss suffered by the investors.
(i) Role of the Legal System: The Benchmark Case
What is the impact of raising the legal liability of an auditor? The
auditors
effort
choice
problem
in
the
semi-separation
equilibrium is characterized by the following two equations that
are obtained from the auditor's first order condition derived in
Section 2(
vi
):
C
A0e
A
q
1A0l
Aÿ
2q
2A0:
P
Aÿ
02Aq
2A:
P
A;
C
B0eB
ÿ 1
ÿ
2
q
2B0:
PB
02Bq2B
:
PB
:
From now on we introduce the notation
02Aand
02Bto
distinguish between the derivative of
2with respect to
e
Aand
e
B. Notice that demand for auditing from type 2 firm is denoted
2for a type
A
auditor and
1
ÿ
2
for a type
B
auditor. Our
result that increased accuracy, and hence effort, reduces the
demand for auditing implies that
02A<
0 whereas
02B>
0. The
effect of a change in legal liability on the auditor's marginal cost
can now be ascertained by differentiating the above two
equations, with respect to
PA
and
PB
, as follows:
@
C
A0e
A
@
P
A ÿ
2q
2A0ÿ
02q
2A;
@
C
B0e
B
@
P
B ÿ 1
ÿ
2
q
2B0
02q
2B:
efforts for a type
A
auditor. For a type
B
auditor,
02Bis positive but
q
20Bis negative and therefore, the effect of increased
PB
is to
increase the marginal cost of effort of type
B
auditors at the
optimum resulting in a higher level of effort.
(ii) Impact of `Deep Pocket' Awards
Though increase in legal liabilities can induce a higher level of
effort, in reality it may neither be institutionally feasible nor
desirable to increase the legal liability of all auditors. This will be
particularly so, if Type
B
auditors, are relatively smaller and have
less wealth. On the other hand, Type
A
auditors could also be the
bigger and more successful auditors with `deep pockets' who may
end up paying a disproportionate share of the total liability.
Thus, it may so happen that
PA
would increase keeping
PB
unchanged. First of all, increased liability will make the higher
quality (type
A) auditing more expensive which will drive its
demand down. That apart, the increased effort and thus,
increased accuracy by type
A
auditors will also drive the demand
for high quality auditing down. The question is, how will this
change in demand and change in liability for type
A
auditors
influence the action of a type
B
auditor? We focus on the
semi-separation equilibrium to check that, and take the derivative of
the type
B
auditor's marginal cost with respect to
PA
to find that:
@
C
B0eB
@
PA
ÿ
@
2@
PA
q
0
2B