You are on page 1of 21

J. Account.

Public Policy 28 (2009) 71–91

Contents lists available at ScienceDirect

J. Account. Public Policy


journal homepage: www.elsevier.com/locate/jaccpubpol

Mandatory audit firm rotation: Fresh look versus


poor knowledge
Tong Lu *, K. Sivaramakrishnan
Bauer College of Business, University of Houston, Houston, TX 77204-6021, United States

a r t i c l e i n f o a b s t r a c t

JEL classification: Our aim in this paper is to investigate the effects of mandatory
G31 audit firm rotation (MAR) on companies’ investment decision and
G34
auditor choice in a capital market setting. We compare a MAR
D82
regime with a non-MAR regime in a setting in which auditors’
Keywords: independence and companies’ opinion shopping are real concerns.
Auditor independence To capture auditor independence and opinion shopping, we model
Auditor conservatism auditor biases (a conservative bias or an aggressive bias) and client
Audit quality firms’ incentives to engage auditors with desired biases. We find
Investment efficiency that when firms engage in opinion shopping, MAR improves
Auditor rotation investment efficiency for some firms but impairs investment effi-
ciency for other firms. More generally, we contribute to the litera-
ture by demonstrating the real effects of auditing on corporate
resource allocation decisions.
Ó 2009 Elsevier Inc. All rights reserved.

1. Introduction

External auditors serve a valuable function in capital markets by lending credibility to financial
statements issued by public companies, thereby reducing information risk. The sanctity of this audit
role depends crucially on the nature of the auditor’s attestation. It has been argued that long-term
auditor–client relationships can result in auditors becoming complacent and lax in their attestation
roles. Lack of auditor independence from the client firm is also a matter of concern. One suggested
avenue to address these concerns is to place term limits by enforcing mandatory audit firm rotation
(MAR). Indeed, MAR has received much attention in legislative and regulatory circles recently, as
reflected in a General Accounting Office study (GAO, 2003) required by Section 207 of the

* Corresponding author. Tel.: +1 7137430448; fax: +1 7137434828.


E-mail address: tlu4@uh.edu (T. Lu).

0278-4254/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jaccpubpol.2009.01.006
72 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

Sarbanes-Oxley Act to look into its costs and benefits. However, there is little by way of systematic
analysis in the literature as to whether MAR is at all desirable in the US corporate environment. In
the paper, we present a theoretical model that evaluates the desirability of MAR relative to the current
regime in which the decision to change an auditor is entirely voluntary.
Opponents of MAR argue that a new auditor will not have the benefit of client-specific knowledge
of a previous auditor (GAO, 2003). This poor knowledge of the new auditor hampers the effectiveness of
the audit process and can result in a deadweight loss to society. Sunder (2003) goes even a step further
to suggest removing the mandatory audit requirement in the first place, ‘‘leaving each firm to decide if
it wishes to have an audit certificate accompany its reports.” On the other hand, proponents of MAR
claim that a new auditor would bring a fresh look to the auditing task (Conference Board, 2003), and
that it is an effective way of ensuring auditor independence and preventing ‘‘opinion shopping.” The
focus of this paper is to model this classic trade-off between fresh look and poor knowledge to address
the desirability of MAR in a setting in which auditor independence and opinion shopping are real
concerns.
To do so, however, we must first specify a criterion for evaluating the desirability of a regime, MAR
or otherwise. We use investment efficiency as the criterion by analyzing the impact of an audit regime
on corporate investment decisions, whose cash flow implications are ultimate interest to sharehold-
ers. A regime in which investment decisions come closest to first best investment levels (optimal
investment levels in the absence of any informational asymmetries and agency problems) would be
considered desirable.
In any model of MAR, lack of auditor independence must potentially be an issue in the absence of
MAR. The auditor independence standard requires that the auditor ‘‘must be without bias with respect
to the client since otherwise he would lack that impartiality necessary for the dependability of his
findings, however excellent his technical proficiency may be” (AU Section 220—Independence). We
capture this aspect in our model by introducing auditors who may be biased in viewpoint (a conser-
vative bias or an aggressive bias). There are two ways to introduce such a bias. One approach is to
examine a setting in which the bias arises as a result of auditors’ incentives. A second approach is
to assume that auditors are predisposed to biases in judgment, leading them to process information
in a way that may or may not favor the client (Bazerman et al., 1997). By choosing between a conser-
vative auditor and an aggressive auditor, client firms can potentially align auditor bias with their own
reporting incentives.
Both approaches to modeling auditor attestation constitute interesting avenues to pursue. In the
text, we pursue the latter approach because it allows us to incorporate auditor bias in a tractable man-
ner and focus squarely on our primary objective of examining the desirability of MAR in terms of its
impact on corporate investment decisions. That is, our focus is not so much on the source of auditor
bias as it is to examine the consequences of auditor bias. In Appendix 2, we introduce auditor incentives
and show how auditor biases are endogenized.
Specifically, we assume that auditors can be one of two types: conservative or aggressive (Krishnan,
1994). When faced with uncertainty, conservative auditors tend to err on the side of understatement,
while aggressive auditors tend to err on the side of overstatement. If a firm desires, say, an aggressive
auditor and the incumbent auditor is aggressive, then the firm can retain the incumbent; otherwise,
the firm can shift to a new auditor who is aggressive. In other words, we are essentially examining the
demand for auditor bias by clients.
We assume that a firm, which has private knowledge of its prospect (i.e., the probability of its fu-
ture financial condition), chooses an observable investment level to maximize its expected stock price
conditional on the firm’s audited accounting report, net of the investment cost. Under MAR, the firm
must pick a new auditor, but has the discretion to choose between a conservative and an aggressive
new auditor. In the absence of MAR (hereafter, the non-MAR regime), the firm has an additional option
of retaining the incumbent auditor.
Firms are often known to engage in opinion shopping (Lennox, 2000). Opinion shopping refers to a
firm’s incentive to switch auditors to secure a more favorable audit opinion ex post (i.e., after the
incumbent auditor proposes an audit opinion). In general, opinion shopping compromises the quality
of external audits. Clearly, opinion shopping is a potential issue in the non-MAR regime because audi-
tor switches are entirely voluntary in this regime. However, even under MAR, the firm can potentially
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 73

wait till it privately learns its financial condition and then exercise its discretion in choosing between a
conservative and an aggressive new auditor. Thus, our model presents us with a unique opportunity to
examine how opinion shopping incentives might affect the desirability of MAR—an issue that has not
received much attention in the debate over MAR.
In our model, the incumbent auditor has a knowledge advantage over the new auditor in the sense
that the incumbent auditor can detect the true financial condition with a higher probability; equiva-
lently, the new auditor suffers from poor knowledge. However, to the extent that the chosen new audi-
tor is of a different type from the incumbent auditor, the new auditor brings a fresh look to the audit
task. Using this structure, we examine the equilibrium effects of MAR on companies’ investment deci-
sions and auditor choices.
We find that, when the incumbent auditor is aggressive, MAR results in a switch to a conservative
auditor more frequently than in the non-MAR regime, and when the incumbent auditor is conserva-
tive, MAR results in a switch to an aggressive auditor more frequently than in the non-MAR regime.
Thus, if the purpose of imposing MAR is to bring in fresh look, it would clearly be achieved.
More importantly, to the extent that opinion shopping is a real concern, we find that it has impor-
tant implications for the desirability of MAR from an investment efficiency standpoint. To establish
this result, we first ignore opinion shopping and assume that firms switch auditors ex ante, i.e., before
privately observing their financial condition. In this ex ante auditor choice setting in which opinion
shopping is not an issue, we show that MAR always impairs investment efficiency.
Next, we introduce opinion shopping by allowing firms to switch auditors ex post, i.e., after pri-
vately learning their financial condition. In this ex post auditor choice setting, firms that have received
an unfavorable report from the incumbent auditor have a natural incentive to seek a favorably biased
opinion from a new auditor. Such an opinion shopping incentive drastically alters the implications of
MAR for investment efficiency. In particular, it is no longer the case that MAR always impairs invest-
ment efficiency. Given an aggressive incumbent auditor, MAR impairs corporate investment efficiency
for firms with good prospects but improves corporate investment efficiency for firms with bad pros-
pects. In contrast, given a conservative incumbent auditor, MAR impairs corporate investment effi-
ciency for firms with bad prospects but improves corporate investment efficiency for firms with
good prospects. To our knowledge, our study is the first to highlight the impact of opinion shopping
on the desirability of MAR.
The intuition for the above results is as follows. In the ex ante auditor choice setting, the capital
market knows that the firm has not yet learned its financial condition, so auditor switching is not
viewed negatively by the market. Therefore, a firm that desires a switch to improve its investment effi-
ciency will do so whether in a MAR regime or in a non-MAR regime; however, a firm that does not
wish to switch is forced by MAR to switch, which results in a deterioration of its investment efficiency.
In brief, in an ex ante auditor choice setting, MAR impairs investment efficiency for some firms without
improving investment efficiency for others.
However, in the ex post auditor choice setting, the market knows that the firm privately knows its
financial condition at the time of switching, and is therefore cognizant of its motives to switch. The
consequent negative market reaction would prevent some firms—those firms that would otherwise
switch auditors in the ex ante setting to improve investment efficiency—from switching auditors.
On the other hand, MAR forces all firms to switch, and so the market cannot tell which firms are opin-
ion shopping and which are not.1 Consequently, MAR would enable some firms to improve their invest-
ment efficiency by making ‘‘desirable” switches, but in the meantime, MAR would impair other firms’
investment efficiency by forcing them to make ‘‘undesirable” switches.
In related research, Elitzur and Falk (1996) (hereafter, EF) focus on the effect of MAR on the audi-
tor’s planned audit quality and employ a multi-period framework. EF find that the level of planned
audit quality will diminish over time under MAR. In contrast, we focus on auditor biases and employ
a one-period framework.
Gietzmann and Sen (2002) (hereafter, GS) find that MAR should only be imposed in thin markets
where a few clients dominate the auditor’s client portfolio, whereas we find that the desirability of

1
We thank an anonymous referee for pointing out this argument on the market’s beliefs under MAR.
74 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

MAR depends on the client prospect and the auditor bias. Our study differs from theirs in important
ways. First, GS focus exclusively on auditor aggressiveness, whereas we study both aggressive and
conservative auditor biases. Second, GS concentrate on the auditor’s work/shirk decision while we
examine the client’s investment decision and auditor choice, and the capital market price.
Kornish and Levine (2004) (hereafter, KL) model a contractual setting where the auditor serves two
clients, an audit committee and a manager; we study a market setting, where the client firm chooses
its auditor in light of its expected capital market price. KL argue that the audit committee can use dis-
missal threats to discipline the auditor and therefore MAR is detrimental because it reduces the effec-
tiveness of such a threat. We find that whether MAR is desirable depends on the client prospect and
the auditor bias.
Dopuch et al. (2001) study MAR in an experimental setting. While they assume exogenous benefits
to the firm from financial reports, we derive them endogenously by incorporating a capital market.
They focus on auditor aggressiveness; we study both aggressiveness and conservatism. Their experi-
mental results show that MAR decreases investment but we show that MAR may either increase or
decrease investment, depending on the client prospect and the auditor bias.
The paper proceeds as follows: Section 2 describes our model. In Section 3, we investigate the cap-
ital market’s pricing of the firm and the corporate investment decision given a particular auditor type.
We then analyze the client firm’s auditor choice under both MAR and non-MAR regimes and identify
implications for the investment decision. We do so first in an ex ante case in Section 4 and then in an
ex post case in Section 5. Finally, Section 6 concludes.

2. Model

2.1. Technology and sequence of events

Given our purpose of evaluating the desirability of MAR using investment efficiency as the crite-
rion, our model centers on a firm’s investment decision. The firm chooses an investment level repre-
2
sented by k. The cost of the investment is 12 k . The investment yields a return gk at some future point in
time if the financial condition (i.e., the state of nature) is good ðgÞ, which occurs with probability k; it
yields a return of zero if the financial condition is bad ðbÞ, with probability 1  k. The parameter g > 0
captures the productivity of the investment in monetary terms in the good state.
The firm has private knowledge of its financial prospect, k, and its financial condition (g or b) once it
is realized. Once the financial condition is realized, the firm prepares a report subject to an audit. The
auditor discovers privately the true financial condition with probability q and approves or disapproves
the client-proposed report. With probability q, the auditor accumulates conclusive evidence about the
client’s financial condition, g or b; with probability 1  q, the auditor accumulates inconclusive evi-
dence, denoted by i. We call q audit quality. The audited accounting report is either G or B, where G
signifies ‘‘good” and B signifies ‘‘bad.” Conditional on the audited accounting report and other available
information, the market prices the firm at m. Finally, the return to investment, gk or 0, is realized.
As a benchmark, we first consider the case in which the firm can credibly commit to choose its
auditor before it learns the realized financial condition (g or b). We refer to this benchmark as the
ex ante auditor choice setting. The following time line presents the sequence of events in this setting:
Time line 1:

 Nature determines the realization of firm prospect, k, which is privately learned by the firm.
 The firm makes an investment of k units and chooses its auditor.
 The firm observes the realized financial condition (state), g or b.
 The audited accounting report, G or B, is issued.
 The capital market prices the firm at m.
 The return to investment, gk or 0, is realized.

In practice, however, there is no mechanism by which a firm can credibly commit to choose its
auditor ex ante. A more realistic model is one which allows for a tactical auditor switch by the firm,
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 75

i.e., when the firm has the discretion to choose its auditor ex post, that is, when the firm can potentially
engage in opinion shopping. However, the ex ante setting allows us to first examine the desirability of
MAR in the absence of opinion shopping, which we then use to evaluate the impact of opinion shop-
ping on the desirability of MAR. The following time line presents the sequence of events when opinion
shopping is present:
Time line 2:

 Nature determines the realization of firm prospect, k, which is privately learned by the firm.
 The firm makes an investment of k units.
 The firm observes the realized financial condition (state), g or b.
 The audited accounting report, G or B, is proposed by the incumbent auditor.
 The firm either retains its incumbent auditor or switches to a new auditor.
 The capital market prices the firm at m.
 The return to investment, gk or 0, is realized.

2.2. Preferences

We focus exclusively on the conflict between two generations of shareholders and assume that the
firm is run by a manager who wishes to boost the price of the firm in the capital market. Such a man-
agerial incentive may come from stock-based compensation contracts. This assumption is also in
keeping with other models in accounting and finance that focus on inter-generational shareholder
conflicts (see, for example, Dye, 1985; Fishman and Hagerty, 1989).2
The firm chooses an investment of k and an auditor to maximize the price prospective shareholders
will be willing to pay, less the investment cost. The prospective shareholders in a competitive capital
market will receive the eventual return to investment, less the price they pay for the firm. Thus, the
social welfare is maximized when the return on investment (the eventual return to investment less
the investment cost) is maximized.3

2.3. Audit function

Given the information asymmetry between the current and prospective shareholders regarding the
realized state of nature, auditing plays a crucial role of reducing information risk in support of a trade.
The auditor can be one of two types: aggressive ðAÞ or conservative ðCÞ.4 Auditor conservatism is a
well-understood term in the auditing and accounting literatures. An auditor is said to be conservative
when he or she is more inclined to err toward approving a bad report, B, when faced with uncertainty
regarding the true state (that is, when the auditor accumulated inconclusive evidence, i). On the other
hand, an aggressive auditor is more inclined to err toward approving a good report, G, when faced with
the same uncertainty.5

2
We also relax this assumption to introduce an element of internal agency conflict (‘‘empire building”). The results are
qualitatively similar and available upon request.
3
That is, we assume here that it is the firm that incurs the investment cost. Alternatively, we can assume that the new
shareholders incur the investment cost. It does not change the investment decision, k, because in either case, the investment is
chosen to maximize the expected eventual return to investment less the investment cost. When the investment cost is incurred by
the firm, the firm maximizes its expected price (which is the expected eventual return to investment) less the investment cost;
when the investment cost is incurred by the new shareholders, the firm maximizes its expected price, which equals the expected
eventual return to investment less the investment cost.
4
Auditor decisions can be modeled here (see the next footnote); that is, we can model auditors as strategic players. We choose
to do so in Appendix 2 to streamline the analysis in the text.
5
We show in Appendix 2 how auditor conservatism/aggressiveness can be endogenized. In brief, we show that auditor
conservatism/aggressiveness is determined by a trade-off faced by the auditor. That trade-off is summarized by a ‘‘fee/liability
ratio.” Therefore, outside investors can use this ratio to infer whether the auditor is conservative or aggressive. In other words, even
though we assume in the text that auditor conservatism/aggressiveness is directly observable, this assumption is innocuous in that
auditor conservatism/aggressiveness can be inferred.
76 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

audited accounting reports


given given
auditor auditor
audit
state conservatism aggressiveness
evidence

q g G G
g
λ
1−q i B G
1−q i B G
1−λ b
q b B B

Fig. 1. Auditor conservatism and auditor aggressiveness.

Fig. 1 illustrates auditor conservatism/aggressiveness and the resultant audited accounting reports.
Referring to Fig. 1, if the report G is approved by a conservative auditor, it is fully informative because
it is revealing of the true state g. However, when there is uncertainty, the conservative auditor only
allows the report B and therefore B is noisy. Similarly, if the report B is issued by a firm whose auditor
is aggressive, it is fully informative because it is revealing of the true state b. However, when there is
uncertainty, the aggressive auditor approves the report G and therefore G is noisy.
From the above discussion, we can see that auditor conservatism or auditor aggressiveness is not
always bad: auditor conservatism may produce a type I error while auditor aggressiveness may pro-
duce a type II error. Under certain conditions, one is better; under other conditions, the other is better.
(We are going to identify those conditions in Sections 4 and 5.) As shown in Sections 4 and 5, because
different firms in different conditions demand for different types of auditors, both conservatism and
aggressiveness will stay in equilibrium and neither will disappear or be regulated out.
With the introduction of two types of auditors, now we define ‘‘fresh look:”
Definition. An auditor switch results in a fresh look when the new auditor type is not the same as the
old auditor type.
We assume that the incumbent auditor has the advantage of client-specific knowledge over the
new auditor, or equivalently, the new auditor has ‘‘poor knowledge.” We capture this knowledge dif-
ference through the audit quality, q. On the other hand, an incumbent auditor may not be able to see
‘‘correctly” in her audit process because of her ‘‘coziness” with the client due to her long tenure. A new
auditor instead is not plagued with this problem and comes with a fresh look which may increase the
probability of detecting the client’s financial condition (state). We can represent these two effects on
the audit quality through qða; bÞ, where a represents poor knowledge and b represents fresh look. We
assume that ddqa < 0 and dqdb
> 0; that is, a poorer knowledge decreases audit quality but a fresher look
increases audit quality. Empirical evidence indicates that the effect of poor knowledge dominates
the effect of fresh
 look (see,
 for example, Myers et al., 2003; Stanley and DeZoort, 2007). Therefore,
   6
we assume that ddqa > dq
db. Thus, we have the following (the subscript I refers to the incumbent audi-
tor and the subscript N refers to the new auditor):

Definition. Poor knowledge of the new auditor relative to the incumbent auditor implies qI > qN .7

6
We thank an anonymous reviewer for this way of modeling the poor knowledge and fresh look effects and for the empirical
reference.
7
Under a MAR regime, the incumbent auditor may boost up her audit quality towards the end of her tunure in order to ‘‘clean
up skeletons in her closet.” This is consistent with our assumption that qI > qN . We thank an anonymous reviewer for rasing up this
issue.
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 77

We can characterize an auditor fully by the tuple fT; qj g, where T 2 fC; Ag and j 2 fI; Ng. The follow-
ing table presents the various auditing scenarios along the poor knowledge and fresh look dimensions:
(MAR denotes mandatory audit firm rotation.)

The client’s auditor choice Incumbent fC; qI g Incumbent fA; qI g


Retain the incumbent auditor (infeasible under MAR) fC; qI g fA; qI g
Switch to a conservative new auditor fC; qN g fC; qN g
Switch to an aggressive new auditor fA; qN g fA; qN g

Clearly, in the non-MAR setting, the firm can choose to retain the incumbent auditor, whether con-
servative or aggressive, or hire a new auditor, either conservative or aggressive. In the MAR setting, the
firm does not have the option of retaining the incumbent auditor, but it can still choose either a con-
servative or an aggressive auditor to be the new auditor.

3. Investment

We begin our analysis by examining the complete information case in which the firm’s financial
prospect, k, is public knowledge. Let us first consider the case of an aggressive auditor. As Fig. 1 im-
plies, by Bayes’ theorem, the market price conditional on the audited report is given by
k
mAG ¼ gk and mAB ¼ 0;
1  ð1  kÞq
where the superscript A denotes the aggressive auditor and the subscripts G and B denote the audited
reports. Referring to the structure in Fig. 1, we can calculate the expected market price given an
aggressive auditor as
 
k
PðGjAÞmAG þ PðBjAÞmAB ¼ ½1  ð1  kÞq gk ¼ kgk:
1  ð1  kÞq
Similarly, as Fig. 1 implies, with a conservative auditor, the market price conditional on the audited
report is given by
kð1  qÞ
mCG ¼ gk and mCB ¼ gk;
1  kq
where the superscript C denotes the conservative auditor. Therefore, the expected market price given
a conservative auditor is
 
kð1  qÞ
PðGjCÞmCG þ PðBjCÞmCB ¼ kq½gk þ ð1  kqÞ gk ¼ kgk:
1  kq
Thus, the expected market price is the same under either auditor type, and the firm will choose k to
2
maximize the expected price less the investment cost, kgk  12 k . From the corresponding first-order
condition, the first best investment level is
kFB ðkÞ ¼ kg: ð1Þ
Thus, when information is symmetric, neither the auditor biases (C or A) nor the differential knowl-
edge between the incumbent auditor and the new auditor (qI or qN ) matters.
We next analyze the case in which the firm prospect (i.e., the probability of the state), k, is private
information to the firm. We analyze this case in three parts. In the remaining portion of this section,
we examine the firm’s optimal investment given the auditor’s type. We then investigate the firm’s
optimal auditor choice in the next two sections.
Because the firm makes its investment decision based on its private information ðkÞ, the choice of k
is informative to the market in that the market can potentially infer k from k. Let b
k denote the market’s
inference. For a particular level of k, the market infers that the firm prospect is bk, and we use kðbkÞ to
78 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

refer to that level of investment that triggers the market’s inference, b


k. Denote the firm’s expected
payoff by Vðbk; kÞ when a type k firm pretends its type to be b k by investing kðb
kÞ. We are interested
in equilibria in which the market correctly infers the true k from the firm’s investment choice k:8
Let VðkÞ  Vðk; kÞ represent the firm’s expected payoff when the firm type is revealed fully through
the investment policy kðkÞ. The firm would implement any feasible fully-revealing schedule kðkÞ only
when it is incentive compatible.
Definition. A fully-revealing investment schedule kðkÞ is incentive compatible if

VðkÞ  Vðk; kÞ P Vðb


k; kÞ 8k; b
k: ð2Þ
For now, we will not specify whether the auditor is incumbent or new, but will address this issue at
the end of this section.

3.1. The optimal investment given an aggressive auditor

Let us consider the case in which the firm is with an aggressive auditor. As Fig. 1 implies, the mar-
ket prices conditional on the audited reports of G and B are given by, respectively
b
k
mAG ¼ gk and mAB ¼ 0; ð3Þ
1  ð1  b
kÞq
where bk represents the market’s inference about k. With these market prices, we can calculate the
firm’s expected payoff at the time of the investment decision as
b
k 1 2 b
Vðb
k; kÞ ¼ ½1  ð1  kÞq gkðb kÞ  k ð kÞ : ð4Þ
|fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} 1  ð1  kÞq b 2 ffl{zfflfflfflffl}
probability of report G |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} cost |fflfflffl
of investment
market price given report G

We now state the main result on the equilibrium investment schedule with an aggressive auditor.
Proposition 1. For a firm with an aggressive auditor:

(A) The equilibrium investment schedule kðkÞ solves the differential equation:
0 1q
½kðkÞ  kgk ðkÞ ¼ gkðkÞ ð5Þ
1  ð1  kÞq
with the boundary condition kð0Þ ¼ 0,
0
(B) k ðkÞ > 0, and
(C) kðkÞ P kFB ðkÞ.

3.2. The optimal investment given a conservative auditor

Refer to Fig. 1. By Bayes’ theorem, with a conservative auditor, the market prices conditional on the
audited reports of G and B are given by, respectively
b
kð1  qÞ
mCG ¼ gk and mCB ¼ gk: ð6Þ
1b kq
With these market prices, we can calculate the firm’s expected payoff at the time of the investment
decision as
b
kð1  qÞ 1 2 b
Vðb
k; kÞ ¼ kq gkðb kÞ þ ð1  kqÞ gkðb kÞ  k ð kÞ :
|{z} |fflffl{zfflffl} |fflfflfflfflffl{zfflfflfflfflffl} 1b kq 2
|fflfflffl{zfflfflffl}
probability of report G market price given report G probability of report B |fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl}
cost of investment
market price given report B

ð7Þ

8
We show at the end of Appendix 1 that a pooling equilibrium does not exist. We thank an anonymous reviewer for checking
out whether a pooling equilibrium exists.
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 79

As before, we restrict our attention to the set of investment schedules that are fully-revealing and
state the main result with respect to properties of the equilibrium investment schedule with a conser-
vative auditor.
Proposition 2. For a firm with a conservative auditor:

(A) The equilibrium investment schedule kðkÞ solves the differential equation:

0 1q
½kðkÞ  kgk ðkÞ ¼ gkðkÞ ð8Þ
1  kq
with the boundary condition kð0Þ ¼ 0,
0
(B) k ðkÞ > 0, and
(C) kðkÞ P kFB ðkÞ.

Propositions 1(C) and 2(C) together establish that, irrespective of whether the auditor in place is
conservative or aggressive, overinvestment results. To understand the reason for overinvestment,
consider a simple case in which the private information k can take one of two possible values—
low and high. To maximize its expected price, the low k firm has an incentive to mimic the high
k firm. To do so, the low type firm must make a large investment because a larger investment im-
plies a more favorable firm prospect in the eyes of the market (that is, kðbkÞ is increasing in b
k). How-
ever, the low k firm also realizes that it has a lower chance of producing a favorable audited report
than the high k firm (see Fig. 1). Hence, the low k firm might not reap the full benefits from mim-
icking but surely would bear the full cost of mimicking. Thus, by overinvesting enough, the high k
firm effectively deters the low k firm from mimicking and so reveals its type through k in
equilibrium.
Thus far, we have not distinguished between an incumbent auditor and a new auditor. Propo-
sitions 1 and 2 apply equally to incumbent and new auditors with the use of appropriate audit
quality ðqÞ to express (5) and (8)—an incumbent auditor discovers the true realized state with
probability qI and a new auditor does so with probability qN , with qI > qN . Specifically, the invest-
ment schedules for a firm with an incumbent auditor and for a firm with a new auditor are as
follows:

Investment schedule for a firm With an aggressive auditor With a conservative auditor
0 1qI 0 1qI
With an incumbent auditor ½kðkÞ  kgk ðkÞ ¼ 1ð1kÞqI gkðkÞ ½kðkÞ  kgk ðkÞ ¼ 1kq gkðkÞ
I

0 1qN 0 1qN
With a new auditor ½kðkÞ  kgk ðkÞ ¼ 1ð1kÞq gkðkÞ ½kðkÞ  kgk ðkÞ ¼ 1kq gkðkÞ
N N

4. Ex ante auditor choice

In this section, we analyze the benchmark case in which the firm chooses its auditor before the
realization of its financial condition (state), g or b (refer to time line 1 in Section 2). Recall that in
the non-MAR regime, the firm has the discretion to retain the incumbent auditor, whether conserva-
tive or aggressive, or switch to a new auditor, whether conservative or aggressive. In contrast, the firm
must switch auditors under MAR, but has the option to hire either a conservative or aggressive auditor
to replace the incumbent auditor. Thus, in comparing the two regimes, we have a number of different
combinations to consider.

4.1. Auditor choice given an aggressive incumbent auditor

In a non-MAR regime, given an aggressive incumbent auditor, the firm can choose to keep the
aggressive incumbent auditor, or hire either an aggressive or conservative auditor. The following prop-
osition characterizes the firm’s auditor choice.
80 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

Proposition 3 (Ex ante auditor choice). In a non-MAR regime, a firm with an aggressive incumbent
auditor retains its auditor if k P k and switches to a conservative new auditor if k 6 k , where
1 1
k  qI 1qN
< .
1 þ 1q 2
qI N

To understand this result, recall from Proposition 1 that there is always overinvestment in equilib-
rium. Overinvestment is a costly way of credibly revealing firm type; but a more informative audited
report reduces the signaling burden on investment. Now the question is: Which type of auditor can
help improve the informativeness of audited reports? As Fig. 1 implies, conservative attestation
may produce understatements (with probability kð1  qÞ), whereas aggressive attestation may pro-
duce overstatements (with probability ð1  kÞð1  qÞ). Thus, the likelihood of misstatement is lower
with a conservative auditor for a firm with poorer prospects (i.e., a firm with lower k) but for a firm
with better prospects (higher k), the likelihood of misstatement is lower with an aggressive auditor.
In addition, a new auditor’s poor knowledge ðqN < qI Þ increases the likelihood of misstatement and
so reduces the informativeness of financial reporting, and therefore, other things being equal, poorer
knowledge increases overinvestment. Therefore, we have the following:
Remark 1

(1) Firms with poorer prospects prefer conservative auditors, whereas firms with better prospects
prefer aggressive auditors.
(2) Among the same type of auditors, firms prefer incumbent to new auditors.

Conventional wisdom might indicate that good firms seek more reputable auditors and ‘‘more rep-
utable” means ‘‘more conservative,” i.e., Proposition 3 and Remark 1 might appear counterintuitive.
We interpret reputable auditors as those who minimize the likelihood of misstatements. A conserva-
tive auditor may send false alarms (the type I error) while an aggressive auditor may be unduly opti-
mistic (the type II error). Therefore, when choosing between a conservative auditor and an aggressive
auditor, the firm must trade off these two errors. For a good firm, undue optimism is not a concern
because that firm’s fundamental is truly good, but a false alarm is damaging. For that reason, a good
firm prefers an aggressive auditor. So, indeed, as one might think, good firms seek more reputable
auditors, but not in the sense that the chosen auditor is conservative; rather, the auditor is ‘‘reputable”
in the sense that she minimizes the likelihood of misstatements.
Thus, as Proposition 3 demonstrates, with an aggressive incumbent auditor, a firm with a high k
will retain its incumbent auditor because she is aggressive but a firm with a low k will switch to a con-
servative auditor.

firm investment
k k (λ; A, qI ) optimal investment given an aggressive incumbent auditor
k ( λ; A, q N ) optimal investment given an aggressive new auditor
k ( λ ; C, q N ) optimal investment given a conservative new auditor
k FB (λ ) first best investment

k (λ ; C , q N ) g

k (λ; A, qI )

k (λ; A, q N )

k FB (λ )

financial
0 λ prospect
0 λ* 1 1
2

Fig. 2. Equilibrium investment given alternative auditor choices (the case of an aggressive incumbent auditor).
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 81

This result is depicted in Fig. 2. This figure shows investment schedules with firm prospect k along
the x-axis, and the investment k on the y-axis. The first best investment schedule is denoted by kFB ðkÞ,
which is our benchmark to evaluate investment efficiency. In general, an investment policy that exhib-
its less amount of overinvestment and comes closer to the first best investment schedule would be
preferable.
Specifically, in the region corresponding to k 6 k , the investment schedule kðk; C; qN Þ—the schedule
given a conservative new auditor—comes closest to the first best investment schedule and is therefore
the most preferred. In contrast, in the region corresponding to k > k , the investment schedule
kðk; A; qI Þ—the schedule given an aggressive incumbent auditor—comes closest to the first best invest-
ment schedule. In brief, as the firm chooses an auditor, it behaves as if selecting the ‘‘lower envelope”
of the alternative investment policies in this figure.
We examine next what would transpire under the MAR regime. Notice that in the region k 6 k ,
there would be an auditor switch anyway even under the non-MAR regime. Therefore, we restrict
our attention to the region k P k to offer a comparison between MAR and non-MAR regimes in the
following proposition:
Proposition 4 (Ex ante auditor choice). For firms with an aggressive incumbent auditor and with k P k :

(A) MAR increases overinvestment.


(B) Under MAR, the firm switches to a conservative new auditor if k 6 12 and to an aggressive new auditor
if k P 12.

Simply put, MAR results in poor knowledge and consequently reduces the informativeness of the
audited report, thereby increasing the level of overinvestment necessary for effective communication
of the firm’s private information k. To minimize the potential damage caused by a new auditor’s poor
 
knowledge, the firm would switch to a new conservative (aggressive) auditor when k 6 12 k P 12 .
These choices mitigate the likelihood of misstatement in the respective regions, thereby dampening
the adverse effect of poor knowledge on the informativeness of the audited report. As the following
table and Fig. 2 demonstrate, differences between MAR and non-MAR regimes arise in the region
k P k :

Given fA; qI g k < k k 2 k ; 12 k > 12
Non-MAR Switch to fC; qN g Retain fA; qI g Retain fA; qI g
MAR Switch to fC; qN g Switch to fC; qN g Switch to fA; qN g

Thus, MAR increases the incidence of fresh look: more firms switch from aggressive to conservative
  
auditors. In a non-MAR regime, for firms with poorer prospects k 2 k ; 12 , the benefit from the incum-
bent auditor’s good knowledge exceeds the benefit from a fresh look, and so the firm would like to re-
tain its incumbent auditor. However, when MAR is imposed, the incumbent’s good knowledge is lost,
and now only the ‘‘fresh look” matters. Therefore, such firms with poorer prospects switch to a conser-
vative auditor (Remark 1). On the other hand, in the region k P 12, MAR essentially results in a switch
from an incumbent aggressive auditor to a new aggressive auditor (Remark 1). Thus, in this region,
there is loss in investment efficiency due to poor knowledge without even the benefit of the fresh look.

4.2. Auditor choice given a conservative incumbent auditor

Next we turn to the case where the firm has a conservative incumbent auditor. Proceeding along
similar lines as in the case of the aggressive incumbent auditor above, we state the following
proposition:
Proposition 5. In a non-MAR regime, a firm with a conservative incumbent auditor retains its auditor if
1 1
k 6 k and switches to an aggressive new auditor if k P k , where k  qN 1qI
> .
1 þ 1q 2
q N I
82 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

firm investment
k optimal investment given a
conservative incumbent auditor
optimal investment given a
conservative new auditor
optimal investment given an k (λ ; C , q N ) g
aggressive new auditor
first best investment

k (λ; A, q N )

k (λ ; C , q I )

k FB (λ )

financial
0 λ prospect
0 1 λ ** 1
2

Fig. 3. Equilibrium investment given alternative auditor choices (the case of a conservative incumbent auditor).

The driving forces in this case are the same ones identified in Remark 1. Thus, with a conservative
incumbent auditor, a firm with a low k will retain its incumbent auditor because she is conservative
but a firm with a high k will switch to an aggressive new auditor.
The auditor choice and its impact on investment schedules, when the incumbent auditor is conser-
vative, is depicted in Fig. 3. In the region corresponding to k P k , the investment schedule
kðk; A; qN Þ—the schedule given an aggressive new auditor—comes closest to the first best investment
schedule and is therefore the most preferred. In contrast, in the region corresponding to k 6 k , the
investment schedule kðk; C; qI Þ—the schedule given a conservative incumbent auditor—comes closest
to the first best investment schedule. In brief, as the firm chooses an auditor, it behaves as if selecting
the ‘‘lower envelope” of the alternative investment policies in this figure.
Since there would be auditor switch in the region k > k even in a non-MAR regime, MAR would
not make a difference there. The following proposition characterizes auditor choice under MAR in the
region k 6 k .
Proposition 6. For firms with a conservative incumbent auditor and with k 6 k :

(A) MAR increases overinvestment.


(B) Under MAR, a firm switches to an aggressive new auditor if k > 12 and to a conservative new auditor if
k < 12.
Under MAR, to minimize the potential damage caused by a new auditor’s poor knowledge, the firm
would switch to a new aggressive (conservative) auditor when k P 12 k < 12 . These choices mitigate
the likelihood of misstatement in the respective regions, thereby dampening the adverse effect of poor
knowledge on the informativeness of the audited report and reducing the signaling burden on invest-
ment. As the following table highlights (see also Fig. 3), differences between MAR and non-MAR re-
gimes arise in the region k 6 k .
1 

Given fC; qI g k < 12 k2 2;k k > k
Non-MAR Retain fC; qI g Retain fC; qI g Switch to fA; qN g
MAR Switch to fC; qN g Switch to fA; qN g Switch to fA; qN g

Thus, MAR increases the incidence of fresh look: more firms switch from conservative to aggressive

auditors. In a non-MAR regime, for firms with k 2 12 ; k , the benefit from the incumbent auditor’s
good knowledge exceeds the benefit from a fresh look, and so the firm would like to retain its incum-
bent auditor. However, when MAR is imposed, the incumbent’s good knowledge is lost, and now only
the ‘‘fresh look” matters. Therefore, such firms with better prospects switch to an aggressive auditor
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 83

(Remark 1). On the other hand, in the region k < 12, MAR essentially results in a switch from an incum-
bent conservative auditor to a new conservative auditor. Thus, in this region, there is loss in invest-
ment efficiency due to poor knowledge without even the benefit of the fresh look.
In sum, in an ex ante setting in which the firm is assumed to be able to commit to switching audi-
tors before the realization of the underlying state, our results indicate that MAR always damages
investment efficiency (Propositions 4(A) and 6(A)). Thus, in an ex ante auditor choice setting, regula-
tory intervention is unnecessary. However, in practice, it is not clear that there exist mechanisms by
which firms can credibly commit to switch auditors ex ante. Opinion shopping, or ex post auditor
switches, however undesirable, cannot always be prevented. In the next section, we allow for the pos-
sibility of opinion shopping and examine the impact of MAR on investment efficiency. The top half of
Fig. 4 summarizes the results for the ex ante auditor choice.
1 1
We know from Proposition 3 that k  qI 1qN
and from Proposition 5 that k  qN 1qI
. It
1 þ 1q qN
1 þ 1qN qI
I
is obvious then that when the quality difference between the incumbent auditor and the new auditor
shrinks (that is, when qI and qN get closer), both the above thresholds converge to 12 (refer to Fig. 4).
Intuitively, when the poor knowledge effect does not exist, only the fresh look matters.

5. Ex post auditor choice (opinion shopping)

In this section, we analyze the firm’s ex post auditor choice under both MAR and non-MAR regimes.
By ex post auditor choice or opinion shopping, we refer to the situation in which the firm chooses its
auditor after the realization of the state (g or b) and conditional on the incumbent auditor’s proposed
report, G or B (refer to time line 2 in Section 2).

5.1. Auditor choice given an aggressive incumbent auditor

In a non-MAR regime, given an aggressive incumbent auditor, the firm can choose to keep the
aggressive incumbent auditor, or hire either an aggressive or conservative auditor. The following prop-
osition characterizes the firm’s auditor choice:

Ex ante auditor choice


Given an aggressive
incumbent auditor:
non-MAR switch to conservative retain

MAR switch to conservative switch to aggressive


Given a conservative
incumbent auditor:
retain switch to aggressive
non-MAR

MAR switch to conservative switch to aggressive

0 1 1
λ* 2 λ** λ
Ex post auditor choice
Given an aggressive
incumbent auditor:
retain
non-MAR

MAR switch to conservative switch to aggressive

Given a conservative
incumbent auditor:
retain
non-MAR

MAR switch to conservative switch to aggressive

Fig. 4. Auditor choices in ex ante and ex post cases.


84 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

Proposition 7. In a non-MAR regime, a firm with an aggressive incumbent auditor does not switch
auditors, even though switching would improve its investment efficiency if it switched to a conservative
1 1
new auditor when k 6 k , where k  < .
1 þ qI 1qN 2 1qI qN

Notice that the client firm’s preference over auditors in terms of investment efficiency is the same as
that in the benchmark ex ante auditor choice scenario (Proposition 3). However, the firm will never
switch auditors ex post.9 Obviously, only the firm receiving an audit report B has an incentive to switch
auditors; otherwise, such a report B by an aggressive auditor would perfectly reveal the firm’s financial con-
dition to be b (see Fig. 1) and the capital market would rationally price the firm at mAB ¼ 0 (see Eq. (3)). How-
ever, auditor switching does not help such a firm: the market would be rationally suspicious of a firm that
switches auditors and would infer that such a firm must have a state b and have not had the report G ap-
proved by the incumbent auditor. Therefore, the market price would be mAB ¼ 0 for switching firms. Thus,
switching does not result in any incremental benefit to the client firm, and any auditor switching costs
(unmodeled here) would result in the firm strictly preferring retaining the incumbent auditor.
We examine next what would transpire under the MAR regime:
Proposition 8. For firms with an aggressive incumbent auditor:

(A) MAR decreases overinvestment for k < k and increases overinvestment for k > k .
(B) Under MAR, the firm switches to a conservative new auditor if k 6 12 and to an aggressive new auditor
if k P 12.

The following table demonstrates the differences between MAR and non-MAR regimes:

Given fA; qI g k < k k 2 k ; 12 k > 12
Non-MAR Retain fA; qI g Retain fA; qI g Retain fA; qI g
MAR Switch to fC; qN g Switch to fC; qN g Switch to fA; qN g

Thus, given an aggressive incumbent auditor, MAR improves investment efficiency for firms with
poorer financial prospects ðk 6 k Þ. Under a non-MAR regime, these firms would desire a switch to
a new conservative auditor but would not do so, in fear of market skepticism. But under a MAR regime,
every firm must switch auditors, so the capital market cannot distinguish these firms from other firms
making similar switches. So firms with poorer financial prospects are able to switch to their preferred
auditors under the guise of mandatory auditor rotation. However, for firms with better financial pros-
pects ðk > k Þ, investment inefficiency is minimized with their (aggressive) incumbent auditors, but
MAR forces them to switch to a new auditor with poor client-specific knowledge, thereby impairing
those firms’ investment efficiency.

5.2. Auditor choice given a conservative incumbent auditor

Next we turn to the case where the firm has a conservative incumbent auditor. Proceeding along
similar lines as in the case of the aggressive incumbent auditor above, we state the following
proposition:
Proposition 9. In a non-MAR regime, the firm with a conservative incumbent auditor does not switch
auditors, even though switching would improve its investment efficiency if it switched to an aggressive new
1 1
auditor when k P k , where k  > .
1 þ qN 1qI 21qN qI

9
Of course, in reality, client firms may switch auditors because they have grown to a size that the incumbent auditor lacks a
capacity to handle. We do not introduce this factor into our model.
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 85

Given a conservative auditor, only the firm receiving an audit report B has an incentive to switch
auditors and such a firm can have either a financial condition g or b (see Fig. 1). Therefore, the capital
market will rationally price the firm at mCB ¼ kð1qÞ
1kq
gk (see Eq. (6)). As before, switching does not result
in any incremental benefit to the client firm, and any auditor switching costs (unmodeled here) would
result in the firm strictly preferring retaining the incumbent auditor.
The following proposition characterizes auditor choice under MAR:
Proposition 10. For firms with a conservative incumbent auditor:

(A) MAR increases overinvestment for k < k and decreases overinvestment for k > k .
(B) Under MAR, a firm switches to a conservative new auditor if k 6 12 and to an aggressive new auditor if
k P 12.

The following table demonstrates the differences between MAR and non-MAR regimes:
1 

Given fC; qI g k < 12 k2 2;k k > k
Non-MAR Retain fC; qI g Retain fC; qI g Retain fC; qI g
MAR Switch to fC; qN g Switch to fA; qN g switch to fA; qN g

Thus, given a conservative incumbent auditor, MAR improves investment efficiency for firms with
better financial prospects ðk > k Þ. Under a non-MAR regime, these firms would desire a switch to a
new aggressive auditor but would not do so, in fear of market skepticism. But under a MAR regime,
every firm must switch auditors, so the capital market cannot distinguish these firms from other firms
making similar switches. So firms with better financial prospects are able to switch to their preferred
auditors under the guise of mandatory auditor rotation. However, for firms with poorer financial pros-
pects ðk < k Þ, investment inefficiency is minimized with their (conservative) incumbent auditors, but
MAR forces them to switch to a new auditor with poor client-specific knowledge, thereby impairing
those firms’ investment efficiency. The bottom half of Fig. 4 summarizes the results for the ex post
auditor choice.

6. Conclusions

Our aim in this paper is to investigate the effects of MAR on corporate investment, corporate audi-
tor choice, and capital market price. We compare a MAR regime with a non-MAR regime. We find that
when opinion shopping is absent, MAR always impairs investment efficiency. However, when the
firm’s opinion shopping incentive is present, whether MAR improves or impairs corporate investment
efficiency depends on both client prospect and auditor bias.
Our model can be extended to introduce economic incentives of the auditor so as to study the issue
of auditor decisions to the fullest extent. To do so, we need to incorporate auditor fees and auditor lia-
bilities (refer to Appendix 2 for such an extension).
In our model, we have set aside internal agency conflicts and focused instead on the conflict be-
tween two generations of shareholders. To examine the impact of internal agency conflict on our re-
sults in an analytically tractable and yet practical manner, we introduce the well known ‘‘empire
building” incentives (Stulz, 1990; Hart, 1995) into the model by assuming that the manager derives
a private benefit from the investment decision that is increasing in the level of the investment (Kumar
and Sivaramakrishnan, 2008). All else equal, such empire building incentives result in a managerial ten-
dency to overinvest. Our inferences with respect to the desirability of MAR remain qualitatively the
same with the introduction of this additional feature. However, there are other internal agency conflicts
that we have not considered, such as those arising from moral hazard concerns. An avenue for future
research is to extend the basic model here to include other sources of internal agency conflicts.
If an audit firm would know that it will be replaced soon it would make an extra effort to clean up
its auditing and so the incumbent’s audit effort, qI , would be higher. However, this benefit has to be
86 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

weighed against the learning curve of the new auditor. To address this issue, one must study a multi-
period extension of the model analyzed in this paper.
More generally, using the example of auditor switches, we demonstrate that the role of auditing in
reducing information risk has a dual effect: (1) auditing serves as a corporate governance mechanism
to affect price efficiency and (2) auditing affects corporate resource allocation decisions, thereby influ-
encing expected net cash flows. Thus, auditing alters the very cash flows it helps to predict.

Acknowledgement

We thank the helpful suggestions and comments from the two anonymous reviewers, Martin Loeb
and Larry Gordon (the editors), and Linda Myers.

Appendix 1. Proofs

Proof of Proposition 1. We employ the mechanism design methodology of Kanodia and Lee (1998) to
solve the complicated structure of the economy involving interactions between firms and capital
markets. A logical first step is to identify conditions under which a fully-revealing investment
schedule is also incentive compatible. h

Lemma 1. For a firm with an aggressive auditor, the investment schedule kðkÞ is incentive compatible (IC)
iff
(a) V 0 ðkÞ ¼ q 1ð1kÞq
k
gkðkÞ; 8k, and
k
(b) q 1ð1kÞq gkðkÞ is increasing in k.

Proof of Lemma 1. Both parts of Lemma 1 are results of standard adverse selection techniques; there-
fore, only sketches of proofs are provided here. Part (a) results from taking the derivative of Vðbk; kÞ in
(4) with respect to k and then evaluating the derivative at b
k ¼ k. Part (b) requires that the above deriv-
ative be increasing in k.
Lemma 1 implies that if the firm chooses an investment schedule that meets the conditions
specified therein, then that investment schedule will be incentive compatible. The firm will choose an
investment level to solve the following program, the last constraint of which is the firm’s individual
rationality (IR) constraint:
1 2
Max VðkÞ () Max kgkðkÞ  k ðkÞ; ð9Þ
kðÞ kðÞ 2

subject to
k
V 0 ðkÞ ¼ q gkðkÞ; ½IC conditionðaÞin Lemma 1
1  ð1  kÞq
k
q gkðkÞ is increasing in k; ½IC conditionðbÞ Lemma 1
1  ð1  kÞq
VðkÞ P 0: ½IR condition
(A) Notice that from (9)
0 0
V 0 ðkÞ ¼ gkðkÞ þ kgk ðkÞ  kðkÞk ðkÞ:
Substituting for V 0 ðkÞ into condition (a) of Lemma 1, we have

k 0 0
q gkðkÞ ¼ gkðkÞ þ kgk ðkÞ  kðkÞk ðkÞ;
1  ð1  kÞq

which leads to the desired result, (5).


Individual rationality (IR) requires that VðkÞ P 08k. Because V 0 ðkÞ P 0 from condition (a) of Lemma
1, IR condition can be expressed as Vð0Þ P 0, which implies that kð0Þ ¼ 0.
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 87

(B) From condition (b) of Lemma 1


0 1q
kgk ðkÞ þ gkðkÞ > 0:
1  ð1  kÞq
The above expression and (5) together yield
0 0 1q
kðkÞk ðkÞ ¼ kgk ðkÞ þ gkðkÞ > 0;
1  ð1  kÞq
0
) k ðkÞ > 0:
Because the investment schedule is monotonically increasing, the market can infer the firm’s type on
the basis of the observed investment level.
0 0
(C) Notice that ½kðkÞ  kgk ðkÞ P 0 from (5). Because k ðkÞ > 0 (proved in part (B)), kðkÞ  kg P 0, or,
equivalently
kðkÞ P kg ¼ kFB ðkÞ and
where the above equality is given in (1). h

Proof of Proposition 2. Similar to Lemma 1, the following lemma presents conditions under which a
fully-revealing investment schedule is incentive compatible. h

Lemma 2. For a firm with a conservative auditor, the investment schedule kðkÞ is incentive compatible iff

(a) V 0 ðkÞ ¼ q 1kq


1k
gkðkÞ; 8k, and
1k
(b) q 1kq gkðkÞ is increasing in k.

Proof of Lemma 2. Both parts of Lemma 2 are results of standard adverse selection techniques. Part
(a) results from taking the derivative of Vðbk; kÞ in (7) with respect to k and then evaluating the deriv-
ative at b
k ¼ k. Part (b) requires that the above derivative be increasing in k.
The rest of the proof is very similar to that of Proposition 1 and is therefore omitted. h

Proof of Proposition 3. Recall from Proposition 1 that there is always overinvestment in equilibrium
and overinvestment is costly to the firm. Therefore, the firm will select an investment policy that
exhibits the least amount of overinvestment and comes closest to the first best investment schedule.
Let kðk; A; qI Þ denote the optimal investment given an aggressive incumbent auditor, kðk; A; qN Þ denote
the optimal investment given an aggressive new auditor, kðk; C; qN Þ denote the optimal investment
given a conservative new auditor, and kFB ðkÞ denote the first best investment.
In the following, we demonstrate that (i) kðk; A; qI Þ dominates kðk; A; qN Þ because the overinvest-
ment in the former case is (weakly) less than that in the latter case and (ii) kðk; C; qN Þ is the closest
schedule to the first best schedule for k 6 k and kðk; A; qI Þ is the closest schedule to the first best
schedule for k P k . In the following analysis, we refer to Fig. 2 and the investment schedules in the
table at the end of Section 3.
It can be easily shown that the investment schedules kðk; A; qN Þ and kðk; C; qN Þ intersect at k ¼ 12, and
that kðk; A; qI Þ and kðk; C; qN Þ intersect at
1
k¼ qI 1qN
 k :
1 þ 1q qN
I

Because qI > qN by assumption, k < 12.

(i) An interior intersection of kðk; A; qI Þ and kðk; A; qN Þ would require that qI ¼ qN , which contradicts
the assumption that qI > qN . Therefore, kðk; A; qI Þ and kðk; A; qN Þ do not intersect at any interior
value of k. So one must be everywhere (weakly) above or below the other. From (5), it turns out
that kðk; A; qI Þ is below kðk; A; qN Þ because qI > qN by assumption. So the firm will not choose
kðk; A; qN Þ because kðk; A; qI Þ is closer to the first best schedule.
88 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

(ii) At k ¼ 0, using L’Hopital’s Rule, we get


0 0
k ð0; A; qI Þ ¼ 2g and k ð0; C; qN Þ ¼ ð2  qN Þg:
Therefore, at k ¼ 0, the slope of kðk; A; qI Þ is greater than the slope of kðk; C; qN Þ. This implies that for
small values of k (i.e., k 6 k ), kðk; C; qN Þ is closer to kFB ðkÞ than kðk; A; qI Þ is and vice versa. h

Proof of Proposition 4. (A) It is clear from Fig. 2 that, for k P k , among all the three investment
schedules, kðk; A; qI Þ is the closest schedule to kFB ðkÞ. However, under MAR, kðk; A; qI Þ is infeasible
because the incumbent auditor must be replaced. Then the firm must choose other less efficient
schedules, thereby increasing overinvestment.
(B) Under MAR, a comparison of two schedules, kðk; C; qN Þ and kðk; A; qN Þ, determines which type of
auditor will be chosen. At k ¼ 12, these two schedules coincide, so the firm is indifferent. For k < 12,
kðk; C; qN Þ is closer to kFB ðkÞ than kðk; A; qN Þ is, so a conservative new auditor is preferred. For k > 12,
kðk; A; qN Þ is closer to kFB ðkÞ than kðk; C; qN Þ is, so an aggressive new auditor is preferred. h

Proof of Proposition 5. The proof is similar to that of Proposition 3 and therefore is omitted. h

Proof of Proposition 6. The proof is similar to that of Proposition 4 and therefore is omitted. h

Proof of Proposition 7. The result on the firm’s preference over auditors in terms of investment effi-
ciency is the same as that in Proposition 3. The result on the actual auditor switching is different. Obvi-
ously, given an aggressive auditor, only the firm receiving an audit report B has an incentive to switch
auditors and such a firm must have a financial condition b (see Fig. 1). Therefore, an auditor switch
does not produce any incremental benefit to the client because the capital market will rationally price
a switching firm at mAB ¼ 0. Anticipating such a pricing, the firm would not make an auditor switch. h
Proofs of Propositions 8, 9, and 10: The proofs of these three propositions are similar to that of
Proposition 7 and therefore are omitted. h

1.1. Nonexistence of a pooling equilibrium

In a pooling equilibrium, firms with different values of k choose the same level of investment, so
that outside investors cannot perfectly infer k from k. Let k denote the market’s conjecture about k.
Now the firm’s expected payoff at the time of the investment given an aggressive auditor is
k 1 2
V A ¼ ½1  ð1  kÞq gk  k ; ð10Þ
|fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} 1  ð1  kÞq 2
|{z}
probability of report G |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl}
cost of investment
market price given report G

which implies that the optimal investment given an aggressive auditor is


k
kA ¼ ½1  ð1  kÞq g: ð11Þ
1  ð1  kÞq
Substituting the above value of k into (10) yields
1 2
VA ¼ k : ð12Þ
2 A
Similarly, the firm’s expected payoff at the time of the investment given a conservative auditor is
kð1  qÞ 1 2
VC ¼ kq gk þ ð1  kqÞ gk  k ;
|{z} |{z} |fflfflfflfflffl{zfflfflfflfflffl} 1  kq 2
|{z}
probability of report G market price given report G probability of report B |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl}
market price given report B cost of investment

ð13Þ
which implies that the optimal investment given a conservative auditor is
kð1  qÞ
kC ¼ kqg þ ð1  kqÞ g: ð14Þ
1  kq
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 89

Substituting the above value of k into (13) yields


1 2
VC ¼ k : ð15Þ
2 C
Thus, from (15) and (12), V C P V A if and only if kC P kA if and only if kð1  2kÞ P kð1  2kÞ. This
implies that firms with different values of k may pick different values of k (kC or kA ), depending on
whether kð1  2kÞ P kð1  2kÞ. This contradicts the definition of a pooling equilibrium, which states
that firms with different values of k choose the same level of investment.

Appendix 2. The auditor’s decisions

In this appendix, we endogenize the auditor’s two decisions: audit quality (audit effort) and attes-
tation. We first discuss the auditor’s attestation decision (i.e., whether to approve or disapprove the
client preferred report, G) and then discuss the auditor’s audit quality decision (i.e., what level of q
the auditor chooses).

2.1. Auditor attestation: conservatism/aggressiveness

We assume that, in making her attestation decision, the auditor is torn between the benefits ex-
pected from her client’s potential business opportunities and the expected legal liability for overstate-
ment in the client report. We denote by / the benefit to the auditor from the client’s potential business
opportunities. / can be nonaudit fees, fees from repeat audit engagements, etc. Thus, by granting or
withholding /, the client firm can effectively exert pressure on the auditor for a favorable audit opin-
ion. Further, we assume that if an overstatement is exposed—that is, if the return to investment is 0
but the audited accounting report was G—the auditor is legally liable to investors.10 We denote the
auditor’s legal liability by L.
The client firm uses / to pressure its auditor to approve its favorable report, G. It grants / if its audi-
tor approves G and withholds / if its auditor disapproves G. When making her attestation decision, the
auditor trades off her expected fee against her expected liability. The auditor’s expected payoff and her
attestation decision rule are as follows:

Audit evidence Payoff from approving G Payoff from disapproving G Approving G if and only if
/
g [Conclusive] / 0 L P0
/
i [Inconclusive] /  ð1  kÞL 0 L P1k
/
b [Conclusive] /L 0 L P1

Disapproving G means forfeiting /, but the auditor risks no legal liability for overstatement. Thus,
the auditor’s expected payoff from disapproving G is 0. In contrast, approving G leads to the receipt of
/, but the auditor may bear a litigation risk for potential overstatement. If the audit evidence is g, the
auditor is safe; if the audit evidence is b, the auditor will be definitely liable for L. If the audit evidence
is inconclusive, the auditor assesses her client’s business risk at 1  k (the probability of financial con-
dition b); thus, her expected liability is ð1  kÞL. The auditor’s decision rule, stated in the last column of
the table, is derived by comparing her expected payoffs from approving and disapproving G. From the
decision rule, we derive the publicly issued accounting report,11 which depends on the tension between
fee and liability, /L , and the audit evidence, g or b or i. The audited accounting reports resulting from the
auditor–client interactions are summarized in the following and are illustrated in Fig. 1 in the text.

10
We exclude auditor liability for understatement because of its rarity in the real world. The reason for this phenomenon may be
twofold: first, the client firm can penalize its auditor by withholding business opportunities, /, and second, as Kornish and Levine
(2004) observe, it is extremely difficult for capital market traders to make a case in court for foregone investment opportunities
resulting from understated financial statements. In contrast, it is relatively easy to make a case for the actual investment loss
incurred because of overstatement.
11
Recall that if the auditor approves G, the client firm issues report G and if the auditor disapproves G, the client firm must issue
report B.
90 T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91

Claim 1

(i) [Auditor conservatism] When /L 6 1  k, the audited accounting report issued is B unless the auditor’s
evidence is g, in which case report G is issued.
(ii) [Auditor aggressiveness] When /L 2 ½1  k; 1, the audited accounting report issued is G unless the
auditor’s evidence is b, in which case report B is issued.
(iii) [Extreme auditor aggressiveness] When /L P 1, the audited accounting report issued is always G.

Therefore, outside investors can infer auditor conservatism/aggressiveness by comparing the fee–
liability ratio, /L , with the client business risk, 1  k.12 So there is no loss of generality to assume that
auditor conservatism/aggressiveness is known to outsiders, as we did in the text.

2.2. The auditor’s audit quality (audit effort) decision

For a given fee–liability ratio, /L , when choosing her audit quality q, the auditor can anticipate her
subsequent attestation strategy (conservative, aggressive, or extremely aggressive). The auditor’s ex-
pected payoff at the time of her audit quality choice is her expected payoff from her subsequent attes-
tation strategy less cðqÞ, the cost of performing her audit procedures with quality q. We assume that
cðqÞ is increasing and convex.
 
The case of auditor conservatism /L 6 1  k :
/
According to Claim 1 and Fig. 1, when L 6 1  k, the auditor approves G only when her audit evi-
dence is g. If her evidence is g (the probability of which is kq), she will receive / (see the table in
the preceding subsection). Therefore, the auditor’s expected payoff is
kq/  cðqÞ: ð16Þ
/ 
The case of auditor aggressiveness L 2 ½1  k; 1 :
According to Claim 1 and Fig. 1, when /L 2 ½1  k; 1, the auditor approves G when her audit evidence
is either g or i. If her evidence is g (the probability of which is kq), she will receive /; if her evidence is i
(the probability of which is 1  q), she will receive /  ð1  kÞL (see the table in the preceding subsec-
tion). Therefore, the auditor’s expected payoff is
kq/ þ ð1  qÞ½/  ð1  kÞL  cðqÞ: ð17Þ
 /
The case of extremely aggressive attestation L P 1 :
According to Claim 1 and Fig. 1, when /L P 1, the auditor approves G regardless of her audit evi-
dence. Thus, the auditor always receives /. Because the probability of overstatement (approving G
when the firm’s financial condition is b) is 1  k, the auditor’s expected liability is ð1  kÞL. Therefore,
the auditor’s expected payoff is
/  ð1  kÞL  cðqÞ: ð18Þ
The auditor’s choice of audit quality is derived from the first-order conditions of (16)–(18) and is
described in the following.
Claim 2. The auditor’s choice of audit quality, q, is characterized as follows:
/
c0 ðqÞ ¼ k/ for 6 1  k ðthe case of auditor conser v atismÞ; ð19Þ
L
/
c0 ðqÞ ¼ ð1  kÞðL  /Þ for 2 ½1  k; 1 ðthe case of auditor aggressiv enessÞ; ð20Þ
L
/
q ¼ 0 for P 1 ðthe case of extremely aggressiv e attestationÞ: ð21Þ
L

12
It can be shown that Claim 1(iii), the extreme auditor aggressiveness, would not be sustained in equilibrium. Note that the
extreme auditor aggressiveness always produces report G, therefore such a report has no incremental informational content and
thus is ignored by the capital market. Anticipating the market’s non-response, the client firm has no incentive to induce the
extreme auditor aggressiveness in the first place.
T. Lu, K. Sivaramakrishnan / J. Account. Public Policy 28 (2009) 71–91 91

Therefore, using (19)–(21), outside investors can infer audit quality, q, on the basis of the fee–lia-
bility ratio, /L , and the client business risk, 1  k. So there is no loss of generality to assume that audit
quality is known to outsiders, as we did in the text.

References

Bazerman, M., Morgan, K.P., Lowenstein, G.F., 1997. The impossibility of auditor independence. Sloan Management Review 38,
89–94.
Conference Board, 2003. Findings and Recommendations – Part 3: Auditing and Accounting. Conference Board SR 03-04.
Conference Board, New York, NY.
Dopuch, N., King, R.R., Schwartz, R., 2001. An experimental investigation of retention and rotation requirements. Journal of
Accounting Research 39, 93–117.
Dye, R.A., 1985. Disclosure of nonproprietary information. Journal of Accounting Research 23, 123–145.
Elitzur, R., Falk, H., 1996. Planned audit quality. Journal of Accounting and Public Policy 15, 246–269.
Fishman, M., Hagerty, K., 1989. Disclosure decisions by firms and the competition for price efficiency. The Journal of Finance 44,
633–646.
General Accounting Office (GAO), 2003. Public Accounting Firms: Required Study on the Potential Effects of Mandatory Audit
Firm Rotation. GAO 04-216. General Accounting Office, Washington, DC.
Gietzmann, M.B., Sen, P.K., 2002. Improving auditor independence through selective mandatory rotation. International Journal
of Auditing 6, 183–210.
Hart, O., 1995. Firm, Contracts and Financial Structure. Oxford University Press, London.
Kanodia, C., Lee, D., 1998. Investment and disclosure: the disciplinary role of periodic performance reports. Journal of
Accounting Research 36, 33–55.
Kornish, L., Levine, C., 2004. Discipline with common agency: the case of audit and nonaudit services. The Accounting Review
79, 173–200.
Krishnan, J., 1994. Auditor switching and conservatism. The Accounting Review 69, 200–215.
Kumar, P., Sivaramakrishnan, K., 2008. Who monitors the monitor? The effect of board independence on executive
compensation and firm value. The Review of Financial Studies 21, 1371–1401.
Lennox, C., 2000. Do companies successfully engage in opinion shopping? The UK experience. Journal of Accounting and
Economics 29, 321–337.
Myers, J., Myers, L., Palmrose, Z., Scholz. S., 2003. Mandatory auditor rotation: evidence from restatements. Working Paper,
Texas A&M University.
Stanley, J., DeZoort, T., 2007. Audit firm tenure and financial restatements: an analysis of industry specialization and fee effects.
Journal of Accounting and Public Policy 26, 131–159.
Stulz, R., 1990. Managerial discretion and optimal financing policies. Journal of Financial Economics 26, 3–27.
Sunder, S., 2003. Rethinking the structure of accounting and auditing. Indian Accounting Review 7, 1–15.

You might also like