You are on page 1of 46

A Model of Aggregate Reporting Quality

Venky Nagar Paolo Petacchi


University of Michigan George Mason University
venky@umich.edu ppetacch@gmu.edu

September 27, 2016


We are deeply grateful to the Editor (Richard Lambert) for his extended insights on our model
and its connection to empirics, and to two reviewers for their detailed comments. We also thank Mary
Barth, Phil Berger, Carlos Corona, Paul Fischer, Pingyang Gao, Clement Har, Ra Indjejikian, Stefan
Reichelstein, Florin Sabac, and seminar participants at the Junior Theory Conference, University of
Chicago, Penn State, University of Pittsburgh, and Suolk University.

Electronic copy available at: https://ssrn.com/abstract=2473699


A Model of Aggregate Reporting Quality

Abstract
ABSTRACT: We characterize rms aggregate reporting quality in an economy where a
rational capital market as well as a regulator discipline rms reporting choices. When the
regulator is resource constrained, multiple aggregate reporting choices are possible in equilib-
rium. This multiplicity is driven not just by the regulatory constraint, but also by how this
constraint interacts with managerial incentives and the level of reporting discretion available to
rms. The model oers concrete explanations for empirical ndings linking aggregate report-
ing quality to underlying institutions, and suggests new empirical approaches that can exploit
multiplicity to create more powerful tests.
Keywords: reporting quality, aggregate earnings, institutions.

Electronic copy available at: https://ssrn.com/abstract=2473699


A Model of Aggregate Reporting Quality

I. INTRODUCTION

Accounting principles dene rules to map the economic substance of transactions into
accounting reports. However, managers who have stakes in the accounting report can
distort this process. Accordingly, empirical research nds that a countrys nancial
reporting standards by themselves do not explain its reporting quality; the countrys
institutions that discipline and curb management misbehavior also play an integral role
(Holthausen 2009). But establishing the manner in which a countrys institutions im-
pact its reporting quality has not been straightforward.1 Consequently, reviewing the
empirical literature, Holthausen (2009, 453) concludes that we do not have very specic
understanding of the main factors that shape nancial reporting.
This study therefore develops a theoretical model of aggregate reporting choices by
moving away from the traditional single-rm setting and introducing an economy of
multiple rms. In this economy, each individual managers reporting choices depends
both on rational pricing by the capital market and regulatory enforcement. The models
equilibrium highlights the role of multiple rms: a given rms stock price and the
chances of regulatory penalty depend not only on that managers reporting choice, but
also on other rms reporting choices. This feature is a natural consequence of having a
model with multiple rms. We then show that for some (reasonable) parameter values
of the model, multiple equilibria exist. That is, for the same set of parameter values,
there is an equilibrium where many rms manipulate their earnings and an alternative
equilibrium where few rms do so. We then discuss the implications of multiple equilibria
for empirical analysis. Specically, we compare the issue of multiple equilibria to the
more commonly discussed econometric issue of measurement error. While both can be
addressed by having researchers toss out problematic observations, the reasons are
1
For example, the Leuz, Nanda, and Wysocki (2003) study of international earnings quality nds
that enforcement is unrelated to reporting quality in the univariate setting (Table 2, Panel C), but
becomes signicant when countries are atheoretically grouped into clusters based on their institutions
empirical distributions (Table 3), and then is again signicant or insignicant in the multivariate setting
depending on the linear regression specication (Table 5).

Electronic copy available at: https://ssrn.com/abstract=2473699


dierent, as well as the basis upon which problematic is determined. Briey, the
standard approach is to drop observations that are poorly measured or are outliers,
whereas our approach is to drop observations whose values fall into some theoretically
pre-specied ranges.
Our model envisions an economy of good and bad rms. Each bad rms man-
ager has to decide whether to report honestly or misreport as being good. Her report
is used both by rational investors to price her rm, and by the regulator, who has to
decide whether to investigate the rm or not. Both these uses are aected by the other
rms reporting choices. Investors price a good report based on their beliefs of how
many bad rms are misreporting. The regulator, on the other hand, has limited re-
sources and cannot investigate all rms reporting choices (Becker 1968).2 If many bad
rms misreport, there is a good chance that the regulators budget may be exhausted
before he gets to a given rm. Thus, for individual managers, both the share price ben-
et and the regulatory penalty costs of misreporting vary with the prevalence of other
rms misreporting in the economy. This feature leads to multiple reporting equilibria
in our model.3
We show that for some values of the exogenous regulatory and managerial incentive
parameters, the economy as a whole may land in an equilibrium where many bad rms
misreport or in an equilibrium where few bad rms misreport. Both equilibria are
stable in that once all rms land in an given reporting equilibrium, no rm will nd
it benecial to deviate.4 For other values of the regulatory and managerial incentive
2
Beckers model appears to be especially relevant in the recent deregulatory era where regulators
had been curtailed substantially (e.g., our Appendix A; Rajan 2005). For example, Labaton (2008)
documents that the SECs oce set up to oversee the use of complex securities in the $4 trillion banking
industry had not had a director and, as of September 2008, had not completed a single inspection in
the previous year and a half.
3
An important consideration for our model is the well-known economic fact that congestion eects
in regulatory deterrence can by themselves drive multiplicity: i.e., a low crime level is a deterrence
in itself because the lone criminal knows that the police have ample spare resources to catch him.
However, this line of reasoning has not been extensively used to understand aggregate nancial reporting
patterns in the empirical accounting literature, in part because, in addition to regulation, one has to
account for additional disciplining mechanisms such as rational endogenous pricing in capital markets
and managerial incentives. The value of our model is that it shows how these additional disciplining
mechanisms work with deterrence congestion.
4
We explain at the end of the second section how in each equilibrium the bad rms decide among
themselves which ones will misreport and which ones will not. But from an aggregate perspective, the
point to note is that in one equilibrium many rms misreport, and in the other equilibrium few rms

2
parameters, there is a unique reporting equilibrium where only a given number of bad
rms misreport.
An economy in the multiple equilibria case will sometimes be at one equilibrium, and
sometimes at the other. The researcher has no way to predict which equilibrium point
the economy may pick.5 Furthermore, when the exogenous parameter values change a
bit, the economy may arbitrarily and discretely jump from one equilibrium to another,
leaving the researcher with no clear prediction on how the exogenous parameters aect
the aggregate reporting outcome. On the other hand, the economy in the second case
has only one equilibrium that moves predictably as exogenous parameters change, thus
allowing the researcher to clearly predict the relationship between exogenous parame-
ters and reporting outcomes. In the third section, we explain in detail how an empirical
researcher can exploit this feature to create more powerful tests of institutional fac-
tors on reporting outcomes in an attempt to resolve the mixed results (see footnote
1). Specically, the standard argument views the mixed ndings as arising from noisy
empirical measures of institutions, and suggests that better measurement is the way to
go (Holthausen 2009, 454). As an alternative, our model shows that selecting and com-
paring appropriate subsets of the sample can also be an eective path to more powerful
tests.
Although multiple equilibria are supported by mounting empirical evidence (Angele-
tos and Lian 2016), they have a love-hate standing in economic theory (Angeletos and
Werning 2006). Morris and Shin (1998) and Weinstein and Yildiz (2007) argue that
multiple equilibria are not robust to assumptions about the agents private information,
or represent knife-edge cases. In response, Atkeson (2001), Rey (2001), and Angeletos
and Werning (2006) argue that multiplicity re-emerges when privately informed agents
receive public disclosures, while Angeletos and Lian (2016, Section 4.8) argue that the
Weinstein-Yildiz renement still eectively retains the multiplicity problem. Summa-
do.
5
A common explanation for this indeterminacy in economies with no information asymmetry is
that the agents in the economy coordinate and pick an equilibrium based on some extrinsic non-
fundamental parameter (e.g., Cass and Shell 1983). In this regard, multiple equilibria are dierent than
an equilibrium arising from randomized strategies. In the latter case, individual players randomize
across many actions, but the overall equilibrium is uniquely specied.

3
rizing the literature, Angeletos and Lian (2016, Section 4.8) conclude that the global
games literature cannot be used to eliminate multiplicity.
The next section develops and solves the model. The third section discusses empirical
implications and theoretical contributions. The nal section concludes.

II. MODEL

Setup

The Economy: We start with an economy of size F > 0. There are exogenous
N F > 0 independent rms run by independent managers. These rms operate in a
capital market with rational risk-neutral price-taking investors and a regulator. Of the
N F rms, exogenous NG F rms are good rms that have a true valuation of
1, and exogenous NB F = N F NG F are bad rms have a true valuation of
0. In all other respects, the rms and their respective managers are indistinguishable.6
To remove small error terms, we follow Angeletos and Werning (2006) and consider the
limit economy F . In this case, the law of large numbers implies that we can also
think of the model not in terms of the number of 1 (i.e., good) rms but in terms of
NG
the probability of being a good rm (= N
). We will therefore state the model in terms
of per capita, by which we will think of the economy as having NB bad rms and NG
good rms.
In the beginning, no one knows the true valuation of any specic rm. At Time
0, each manager learns the true value of her rm. Investors and the regulator are still
uninformed about the true values of specic rms: they only know that NB rms per
capita are bad or 0-valued and NG rms per capita are good or 1-valued.7 At Time
1, all managers simultaneously report the value of their rms. They have discretion to
report either 1 or 0.8 At Time 2, the regulator investigates the rms for misreporting.
6
See Aghion and Bolton (1992) for a model of a rm whose value has binary support. We have
also developed a model (available upon request) where rm values have continuous support. Our main
results hold in that model as well.
7
All the other exogenous variables are common knowledge to all actors in the model. We therefore
do not have global-games factors at play.
8
We assume simultaneous instead of sequential reporting to avoid strategic Stackelberg issues be-
tween early and late reporting rms and the ensuing implications for regulatory and investor behavior

4
He penalizes any manager who is caught misreporting, and forces her to publicly issue
a restatement, which investors learn about. At Time 3, the investors price all the
disclosures, and the managers are paid o. Specically, managers get compensated on
the stock price less any levied penalties for misreporting. Figure 1 provides a time-line.9

Figure 1 about here

Managers incentives and actions: Each manager is risk-neutral, and her stock price
interest in her rm is a long ownership share > 0, which is publicly known through
compensation disclosures. A manager of a 0 (bad) rm who reports 1 without being
caught therefore makes times the price of the rm in Time 3. A manager of a bad rm
who reports 1 and subsequently is caught makes c in Time 3 because the regulator
publicly levies the penalty c, and also forces an earnings restatement that drives the
stock price to zero. The penalty is paid by the manager, not the rm. Each manager
chooses her report to maximize her net expected payos.10
Regulators existence and budget: We next discuss the regulatory action at
Time 2. We take the regulator such as the SEC as exogenously given.11 This regulator
has an exogenous per capita budget K > 0 dollars to investigate NG + NB rms (or
a total economy-wide budget of K F, F ) in resources, with the budget being
determined by the Legislature. Note that the SEC is not funded by its nes all nes
go to the U.S. Treasury General Fund or the Fair Fund (Sarbanes-Oxley Act, 15 U.S.C.
7246(a), i.e., the Fair Fund Provision).12 We therefore do not link the regulators
budget K to the nes c. We also do not model the complex negotiations process by
(Foster 1981).
9
Our model is thus a one-shot game. Rey (2001, Section 2) also indicates that for tractability
reasons, many global game models are also one-shot or repeated versions of one-shot games.
10
For simplicity, we assume that and c are the same across all rms. Allowing heterogeneity in
these parameters will induce heterogeneity in rms incentives to misreport. However, the qualitative
nature of our results, in particular the existence of a region where multiple equilibria occur, should
continue to hold.
11
See Rajan and Zingales (2004, Ch. 7) and Shleifer (2005) for theories on the existence and impor-
tance of a public regulator. Most economies also supplement the public regulator with private auditors.
The force of these auditors is zero presently, but we later incorporate a private cost of misreporting,
which we model an additional negative constant. This additional constant does not materially add to
the intuition, so we keep it zero for now.
12
http://www.yalelawjournal.org/comment/sec-monetary-penalties-speak-very-loudly-
but-what-do-they-say-a-critical-analysis-of-the-secs-new-enforcement-approach.

5
which the various factions of the Legislature agree to the budget, except to note that
this process makes the value of K common knowledge (i.e., rms have no residual private
information about the budget). See Appendix A for an illustration of these complexities.
Regulators incentives: Following Kothari, Ramanna, and Skinner (2010, 252),
the regulators incentives are modeled on the SEC, whose purpose is to protect in-
vestors, maintain fair, orderly, and ecient markets, and facilitate capital formation.13
The benets of accurate reporting extend far beyond our model. For example, capital
formation is beneted by accurate reporting because the price system becomes more in-
formative and is thus more ecient at directing resource allocation (Hayek 1945; Bond,
Edmans, and Goldstein 2012; Gao and Liang 2013). Consequently, we do not measure
the benets of SECs actions based on the penalties and transfers it collects from rms.
Instead, we model the SECs goal as maximizing the number of truth-telling rms in
the economy.14
Nonetheless, the SEC has to be strategic in how it fulls its social objective function.
The SEC Enforcement Manual Section 2.3.1 indicates that the criteria for opening an
investigation not only include factors such as the potential magnitude of the violation
and the potential losses involved or harm to an investor or investors, but also whether
the conduct can be investigated eciently and within the statute of limitations period.15
Keeping both the objectives and the eciency requirements in mind, we model the
regulators overall strategic objective as: Lexicographic (maximize number of rms in the
economy that report truthfully; minimize spending). That is, the regulator maximizes
the number of truth-telling rms in the economy in an ecient manner by spending the
budget only when there is a positive probability of detecting misreporters (because this
13
See http://www.sec.gov/about/whatwedo.shtml. While we rely on the SEC for institutional
support for our model, we note that the corresponding regulators in other countries are quite similar to
the SEC in their objectives. For example, see the website of the UK public regulator, the Financial Re-
porting Council, at https://frc.org.uk/Our-Work/Conduct/Corporate-Reporting-Review.aspx.
14
In this regard, the SEC is dierent from the IRS whose mandate is to explicitly collect taxes from
each taxpayer. IRSs objective function can therefore be modeled in a pure transfer-like manner as
maximizing the expected taxes collected less investigation costs (Graetz, Reinganum, and Wilde 1986).
The SEC has no direct tax-like revenue transfer from its constituent rms as a part of its objective
function: in fact, the SEC is expressly forbidden from acting on behalf an individual investor (who
must instead sue privately to recover damages), and all nes it collects must be handed to the U.S.
Treasury General Fund or the Fair Fund. See http://www.sec.gov/divisions/enforce/about.htm#
.VDlyrvn3OG8.
15
See http://www.sec.gov/divisions/enforce/enforcementmanual.pdf.

6
spending will in expectation improve the quality of the nancial markets), and stops
spending budget if the probability of detecting misreporters falls to zero (because this
spending will no longer improve the quality of the nancial markets). As noted before,
we assume that the government publicly reveals these features of the regulator with
complete precision, and so it is common knowledge.
Regulators investigation process: Investigating each rm is assumed to cost the
regulator one dollar; so at the per capita level of NG good rms and NB bad rms, the
regulator can investigate at most K rms. We assume that the regulator makes no errors
in the investigation process (Schwartz 1997; Dechow, Ge, and Schrand 2010, Section
3.3.1.1). Each investigation will therefore reveal the true value of the rm. Therefore,
the only way for a misreporting rm to avoid getting caught is if the regulator never got
around to investigating it in the rst place.
In the class of equilibria we consider, the good rm will always truthfully report 1; it
is only the bad rm that may choose to misreport up to a report of 1. The regulator
therefore rst checks if the number of per capita 1 reporters is NG . If so, all rms are
telling the truth, and the regulator does not conduct any investigation, for he has reached
his social goal at minimum cost. Otherwise, the regulator must scan the universe of 1
reporters to detect misreports. We assume that the regulator starts inspecting the 1
reporters one after the other till he has run out of budget, or he believes there are no
more misreporters (at which point further spending is a waste). Since all 1 reporters
are otherwise indistinguishable, the regulator randomly lines up the 1 reporters into his
investigation list.
The main advantage of the investigative process is that every dollar the regulator
spends has a strictly positive probability of netting a misreporter. The regulator there-
fore does not waste any portion of his budget, and thus fulls his objective.
The empirical literature on SECs Accounting and auditing enforcement releases
(AAERs) supports our modeling assumptions (Dechow et al. 2010, Section 3.3.1). This
literature recognizes that many manipulating rms may go undetected by the SEC (high
Type II error rate), but that the rate of false accusations (Type I error rate) is lower; this
is what happens in our model as well. This research also shows that the likely triggers of

7
AAERs are managerial compensation and capital market incentives (key aspects of our
model) as well as debt covenants and corporate governance issues (Dechow et al. 2010,
Section 3.3.1.1). The SEC thus appears to have access to information that enables it
assess the probability of misreporting, which is what we assume in our model. In any
event, should the regulator be less than ecient, i.e., he wastes some of his resources,
we can easily incorporate that in the model by assuming that the regulators eective
budget is an amount K  < K.
Regulators disclosure and penalty process: The regulator does not disclose
the identity of the rms that were inspected but did not misreport. This is consistent
with the SEC enforcement manual Section 2.3.3 which states that all investigations are
condential unless otherwise ordered by the commission.16 Because the regulator does
not disclose any investigations, we assume that there is no credible way for a good rm
to signal that it was indeed investigated and found to be a truthful reporter.17
The regulator may infer at some point in the investigation process, based on de-
duction by elimination, that all the remaining unexamined rms reporting 1 are misre-
porters. But as Hermalin and Katz (1991) make it clear, this is observable not veriable
information. Hermalin and Katz (1991) argue that only veriable information in
our case information collected through investigations is admissible in court. So we
assume that the penalties for a misreporter can only be levied when the regulator has
accumulated clear evidence based on his investigations and thus can defend himself in
court. At that point, the misreporting manager is forced to pay a ne and mark the
rms valuation to 0. This completes all the actions at Time 2.18
Assumption on the regulators budget: The above regulatory incentive also
implies that no rm will lie if K > NB + NG , because if any bad rm lies and reports 1,
16
See http://www.sec.gov/divisions/enforce/enforcementmanual.pdf.
17
We have developed an extension of the model (available upon request) where the good rms try to
signal their type through costly activities such as dividends. Multiplicities obtain even in the extended
model.
18
One can question why the regulator does not simply publicly disclose his beliefs that all the
unexamined rms reporting 1 are misreporters. Our answer is that the objective of the regulator is
to correct actual misreporting in the economy, not make unveriable disclosures. Our assumption is
consistent with the SEC enforcement manual Section 2.3.3, which privileges the condentiality of its
investigations. Our models investors, by contrast, will use observable non-veriable information in
their pricing process.

8
the regulator will see that the number of rms reporting 1 is not NG . He will therefore
spend his budget till he has successfully found all misreporters. Since his per capita
budget exceeds the per capita number of rms, he will always nd all the misreporters.
We therefore assume, consistent with our Appendix A, that the regulator is resource
constrained by K < NB + NG .
Investors pricing process: Investors are rational and risk-neutral, and observe
both the reports and the assessed penalties. Note that unlike the court which can
only penalize veried information, investors price observable information that is not yet
veried. They rationally compute the expected number of uncaught misreporters and
discount the nal reports accordingly.19

Equilibrium

Because managers in our economy have incentives to maximize share prices, we consider
equilibria where the good rms always report 1, and the bad rms may report 0 or 1;
i.e., the direction of misreporting is always up (and good rms cannot report up).20
Our rst lemma relates to the regulators behavior in Time 2 (proof in Appendix B):

Lemma 1 1. If the equilibrium is such that there are only NG per capita 1 reporters,
all rms are telling the truth, and the regulator does not conduct any investigations.

2. If the equilibrium is such that there are NG + M > NG per capita 1 reporters (i.e.,
there are M bad rms who are misreporting), every misreporting rm at Time 1
rationally anticipating this equilibrium assesses its ex ante probability of getting
K
caught as NG +M
. Specically, the only feasible equilibrium values of M are such
that K < NG + M.
19
We therefore assume that the nes levied on the manager are not disgorged to investors, but instead
transferred to an exogenous entity such as the Treasury General Fund.
20
In these equilibria, investors will always value a report of 0 as 0. We will show in equation (1) that
the payo to a good rm for reporting 1 is always > 0. So the only rms reporting 0 in equilibrium will
be the bad rms, justifying the investor valuation. We will further show that in equilibria where no rm
reports 0, investor valuation of 0 as 0 is justied through the intuitive criterion. In other models, some
managers may want to lower (and not raise) the price of their rms due to upcoming stock grants or
MBO or other personal considerations. Our assumption is that such managers do not form a signicant
portion of our economy in any given time-period (see the corroborating evidence in Kothari, Shu, and
Wysocki (2009)). Furthermore, because all agents are price takers in our economy, our results should
be robust to a few managers misreporting downwards.

9
Lemma 1 is important because, in equilibrium, there may be no misreporters, and
the regulator will consequently conduct no investigations. But this outcome obtains
because every bad rm knows that if it chooses to misreport, it faces a strictly positive
probability of getting caught. That is, the regulators ex post behavior is irrelevant to
truth-telling rms. The only rms who care about the regulators ex post behavior are
the misreporting rms, and that is where Lemma 1 applies. In equilibrium, there may
be no misreporters, but as is standard in game theory, this equilibrium emerges from
Lemma 1 being applied to the o-equilibrium path.
Our goal is to further characterize the properties of M, the number of bad rms that
misreport 1. First, note that the total number of per capita rms reporting 1 at Time 1
is NG + M. The regulator investigates at most K of these rms. Lemma 1 states that if
K > NG + M in equilibrium, all M rms will be caught for sure. So the only possibility
is M = 0. Therefore, if M > 0 in equilibrium, it has to be also true that the endogenous
M satises K < NG + M. In that case, in Time 3, the nal report of a misreporting rm
is either 0 or 1 depending on whether the regulator caught it and forced it to correct its
misreporting, or not. In this case (proof in Appendix B):

Lemma 2 If M > 0, all rms know at Time 1 that a standing report of 1 in Time 3
will be priced in Time 3 as:
NG
KM
(1)
NG + M NG +M

Therefore, at Time 1 in the equilibrium case M > 0, the manager of a bad rm who
reports 1 computes the following as her expected payo in Time 3:

   
K NG K
1 KM
+ (c) (2)
NG + M NG + M NG +M
NG + M

We can divide the managers net payo by . We denote the scaled penalty as
C c . So the expected scaled payo for a manager of a bad rm to reporting 1 is (the
payo to a bad rms manager from reporting 0 is 0):

   
K NG K
1 KM
+ (C) (3)
NG + M NG + M NG +M
NG + M

10
Both the costs and benets of misreporting vary with the number of misreporters.
This can lead to multiple xed points or equilibrium number of misreporters, as the
following proposition makes explicit (proof in Appendix B):

Proposition 1 1. In all equilibria, good rms always report 1.

2. If NG KC then there is a unique equilibrium where no bad rm reports its value


as 1. All rms thus report truthfully. The regulator conducts no investigations.

3. If NG > KC then there are three cases to consider. Dene:



K K 4NG C
M + 1+ NG (4)
2 2 NG KC

(a) If M NB then no bad rm reports its value as 1. All rms thus report
truthfully. The regulator conducts no investigations.21

(b) If 0 M < NB there are three alternating stable and unstable equilibria:

i. No bad rm reports its value as 1. That is, all rms report truthfully.
The regulator conducts no investigations. This equilibrium is stable to
small perturbations.

ii. Exactly M bad rms report their values as 1 (and the other bad rms
report 0). The regulator conducts investigations till he exhausts his budget
or nds all the misreporters (note that M + NG > K as required). This
equilibrium is unstable to small perturbations.

iii. All bad rms report their values as 1. The regulator conducts his inves-
tigations till he exhausts his budget or nds all the misreporters. This
equilibrium is stable to small perturbations.

(c) If M < 0 then there is a unique equilibrium where all bad rms report their
value as 1. The regulator conducts his investigations till he exhausts his budget
or nds all the misreporters.
21
We separate the NG KC case explicitly from M NB case because M could be complex or
undened when NG KC.

11
We make several remarks about this equilibrium. First, note that NG is common
knowledge. Therefore, by examining the number of rms reporting 1, everyone can
ascertain the realized equilibrium in the multiple equilibrium case. Second, recall that
c
K is the number of rms that the regulator can scan, and C
is the penalty that the
regulator imposes on a manager (scaled by his ownership stake). KC can thus be viewed
as a measure of regulatory strength relative to the managers ownership incentives driven
by endogenous prices. Intuitively, one would expect that all bad rms will misreport
when regulatory costs are low or managerial incentives are high, no rms will misreport
when regulatory costs are high or managerial incentives are low, and multiplicity may
arise when regulatory forces (scaled by incentives) are moderate. What Proposition
1 does is to quantify these thresholds precisely. Note, by contrast, that a pure crime
deterrence model without capital markets provides no role for managerial ownership
incentives driven by endogenous prices, whereas in our model it is KC K c , the
regulatory strength relative to ownership incentives, that drives the equilibrium: the
regulatory strength aects the managers misreporting penalty, and her ownership level
determines her wealth eect from the endogenous pricing eect of uncaught misreporting
(which itself depends on the regulatory strength).
The rst criterion that prevents misreporting is if KC > NG (see Figure 2). However,
as Proposition 1 shows, this is not necessarily the tightest threshold. That is, if KC =
(NG ) where  is a very small positive number, M in equation (4) can be very large
and uniqueness can still obtain. Only when M (0, NB ) does multiplicity obtain.
On the opposite side, when C becomes very small, M K NG in equation (4). If
K > NG , then M > 0 and multiplicity is still possible. To see this, consider the case
where no bad rm misreports. If one bad rm misreports, the regulator will catch that
rm (K > NG ) and assess the penalty C. This is sucient to make no misreporting an
equilibrium. The multiplicity thus depends on M in equation (4) in the manner that
Proposition 1 makes precise.
Figure 2 provides an illustration of all three possibilities. When C is low (C =
0.02), all bad rms misreport. When C is high (C = 2), no bad rm misreports. The
multiplicity happens when C is at a moderate level (C = 1 in the example). This

12
three-solution case has two corner solutions: no bad rm misreports, and all bad rms
misreport. The results follow the well-known properties of multiple equilibrium models
(see Section 2.12 and 6.14 of Romer (1996)). In particular, the two corner or end
solutions are stable. When every bad rm misreports, misreporting pays more than
zero, ensuring stability. When no bad rm misreports, misreporting pays less than zero,
ensuring stability. The interior or middle solution by contrast is not stable. If a few
bad rms reporting 0 were to deviate to misreporting 1, the payos from misreporting
rise for all bad rms, providing an incentive for every bad rm to misreport. One could
therefore conjecture that when C is moderate, the most likely equilibria are all bad rms
reporting 0, or all bad rms reporting 1.
The model of multiplicity itself cannot specify which equilibrium the economy will
actually pick. This point has been extensively examined in the literature to which we
have nothing to add; we refer the interested reader to the macroeconomics textbook
Romer (1996, Sections 2.12 and 6.14), Rey (2001, Section 4). For example, one belief
structure commonly used to get agents to choose among multiple equilibria is to posit
that all agents believe that the economy will be at one equilibrium when an economically
irrelevant external variable (sunspot) takes a high value, and at another when it takes
a low value (Cass and Shell 1983).
We next show that our multiple equilibria in Proposition 1 are robust to a well-known
renement, the intuitive criterion. In Proposition 1, investors equilibrium belief is that
the value of a report of 0 is 0. This belief is justied in equilibria where the only 0
reporters are bad rms. In equilibria where no rm reports 0, we show that this belief
satises the intuitive criterion, a commonly-used criterion to assess the reasonableness of
o-equilibrium beliefs (Kreps 1990, Section 12.7.4). The intuitive criterion requires that
if for all o-equilibrium beliefs of investors upon receiving the report 0, a good rm (or
a bad rm) does strictly worse than its current reporting of 1, then the only allowable
o-equilibrium investor belief upon receiving 0 is that the rm is a bad rm (or a good
rm). However, a bad rm cannot do strictly worse: consider the case when investors
value a report of 0 at 1; then a bad rm pays no penalty and earns 1 for telling the
truth, which is more than it earns now. If a similar dominance criterion fails to apply

13
to the good rm as well (e.g., when investors believe a report of 0 means 1, and a good
rm makes more by reporting 0 (net of regulatory penalties) than it is making now),
our investors o-equilibrium beliefs trivially satisfy the intuitive criterion. Otherwise,
the only admissible possibility is that investors value a report of 0 at 0. Our investors
o-equilibrium beliefs in this case also satisfy the intuitive criterion.

Figure 2 about here

The All-or-Nothing Outcome

A somewhat unappealing feature of Proposition 1 is that the only equilibria are all bad
rms misreporting or no bad rms misreporting. The reason for this is not economic,
but geometric. The payo curve intersects the x-axis only once (see Figure 2). More
intersection points can be created if the horizontal x-axis were to be lifted. In that
case, the intersection points will occur in the interior, and we can get the more appealing
case of only some bad rms misreporting.
Lifting the x-axis has a very simple economic rationale. Till now, we assumed that the
private cost of misreporting was 0. Following Fischer and Verrecchias (2000) reasoning,
we assume that there is an exogenous private cost m > 0 to the manager to misreport.
This cost may represent expected reputation eects, or the eort to convince the private
auditor and other parties, etc.
In this case, the threshold equilibrium payo to misreporting 0 as 1 is not 0, but
m, i.e., expected misreporting benets should at least cover the private misreporting
costs. Accordingly, the horizontal x-axis must be lifted from 0 to m. The constant
lift reects the economic feature that the private misreporting cost m is not subject to
enforcement thinning; each misreporter has to pay m, irrespective of how many other
rms have misreported. By contrast, the regulatory budget K is in a curve that bends
in part due to enforcement thinning. Following equation (3), we get:

   
K NG K
1 KM
+ (C) m (5)
NG + M NG + M NG +M
NG + M

We show in the proof to Proposition 1 that the denominator is strictly positive in

14
equilibrium. We can therefore rewrite the equation as (with x NG + M):

x2 (NG KC) xK(NG KC) KNG KC m(x2 Kx + KNG )x 0 (6)

This is a cubic equation in the endogenous variable M, and therefore has one or three
real roots. The only relevant roots are those in [0, NB ], i.e., the number of misreport-
ing bad rms has to be non-negative and less than the number of bad rms. We also
know from our discussion of Figure 2 that a stable crossing has to be downward sloping,
for if the crossing were upward sloping, even if one extra truthfully reporting bad rm
were to change its report to 1, it would raise the benets of reporting 1, causing other
truthfully reporting bad rms to misreport as well. On the other hand, if the crossing
were downward sloping, then having one extra misreporting rm would reduce the ben-
ets of misreporting, and so some bad rm would want to revert to truthful reporting.
(Likewise if a misreporting rm were to report truthfully, it would raise the benets of
misreporting, causing some other truthful bad rm to switch to misreporting.)
The roots of a cubic polynomial are available in closed form.22 Consequently, by
checking the positioning of these roots in the range [0, NB ], one can trivially write
the precise conditions for interior multiplicity in a manner analogous to Proposition 1.
However, because the closed-form representation of cube roots is tedious, we do not
attempt a lengthy rewrite of Proposition 1. Instead, we show visually how the notion
of multiplicity is robust to lifting the x-axis from 0 to m. In other words, the all-or-
nothing or corner results in Proposition 1 are not necessary for multiplicity. In fact, by
lifting the x-axis, the economic rationale for which is the private cost of misreporting, we
can move the intersection points and get both multiplicity and uniqueness with interior
points.
22
https://en.wikipedia.org/wiki/Cubic_function.

15
Figures 3 and 4 about here

For example, Figure 3 is a setting where there is a unique stable interior solution (at
M = 565.28), analogous to the unique corner results with C = 0.02 in Proposition 1 and
Figure 2. On the other hand, Figure 4 is a case where there are two stable equilibria: i)
no bad rm misreports, and ii) some (i.e., 139.96 rms) but not all bad rms misreport.
(There is a unstable interior equilibrium in between at 31.96.) Stability can be easily
assessed in a manner analogous to Figure 2. At 139.36 misreporting rms, no additional
bad rm has incentives to misreport, because it is not worth paying the private cost
m = 0.4 to do so. Furthermore, should a misreporting bad rm back o, another bad
rm will nd it expedient to misreport because the payos to misreporting are now
greater than the private cost m. A similar argument ensures the stability at M = 0 in
Figure 4 and M = 565.28 in Figure 3. (By contrast, the logic is exactly the opposite at
the unstable equilibrium M = 31.96 in Figure 4.) These results are thus analogous to
the multiplicity results in Proposition 1 and Figure 2 with C = 1 where the two stable
equilibria are no bad rm misreporting and all bad rms misreporting.
More important, the intuition with m is the same as with C, holding K constant. If
m is very large, (say m = 1), the horizontal m line will be too high in Figure 3, and no
bad rm will misreport. When m is very small (m = 0.2), there is uniqueness with a
large number (but not all) bad rms misreporting. When m is moderate (m = 0.4), as in
Figure 4, there are multiple equilibria. In sum, multiplicity is not related to the corner
solution concept, but can also obtain with interior intersection points. The economic
intuition of multiplicity can therefore be divorced from the corner solution aspect.
The interior equilibrium with private misreporting costs raises the question on how
the bad rms decide which ones will misreport. The process by which agents in a game
reach a Nash equilibrium is a complex subject (Kreps 1990, Ch. 12). One possibility
in Figure 3 is that the bad rms could order themselves according to some focal-point
ordering, and the rst 568.25 of them misreport. Another possibility is that each bad rm
568.25
independently decides to misreport with a probability of 850
. Because the economy is
a limiting case, this will yield the desired equilibrium. Note that under either scenario,
no bad rm has a strict incentive to deviate.

16
Finally, the distinction between Figures 3 and 4, both of which arise from the polyno-
mial equation (6), can be viewed as a global one, in that changes in m can cause global
changes in equilibria: the number of stable equilibria jumps from 1 to 2. A dierent
approach to viewing these gures is a local one. Consider Figure 3 rst. Because the
slope of the cubic payo polynomial (6) with respect to M, the number of misreporting
rms, is not zero at the intersection point M = 565.28, we can use the implicit function
theorem to argue that small changes in the exogenous parameters will lead to small
changes in the equilibrium value of the number of misreporting rms.23 We can make a
similar local argument for Figure 4 when M = 139.46.
By contrast, consider the all-or-nothing corner stable solutions in Figures 2 and 4.
The payo function at these points is nonzero and continuous. Small changes in the
exogenous parameters will therefore still leave the payo function at a nearby nonzero
value without any change in sign (each nonzero real has an open neighborhood that does
not contain zero). As a result, the corner points stay put as stable equilibrium points
(and the number of equilibria will also not change). We summarize all these observations
as:

Observation 1 Changes in exogenous parameters can have both global and local eects
on the equilibrium number of misreporting rms. Globally, the number of stable equilibria
can jump from 1 to 2. Locally, the interior equilibrium points (only some bad rms
misreporting) will move with small changes in exogenous parameters, but the corner
equilibria (all or no bad rms misreporting) will stay put.

We next discussion the empirical implications of Proposition 1 and Observation 1.


23
In almost all cases, the number of equilibria will not change. This is a global argument, which we
can make by noting that we can restrict our attention to a compact set, and continuous functions in
a compact set are uniformly continuous. So the payo function will not change much anywhere in our
compact region of interest.

17
III. EMPIRICAL IMPLICATIONS AND THEORETICAL
CONTRIBUTIONS

Empirical Aggregate Reporting Literature

We rst use our model to explain some puzzling results in Leuz et al.s (2003) seminal
country-level study of international earnings quality. That study, as noted in footnote 1,
nds that enforcement is unrelated to reporting quality in the univariate setting (Table
2, Panel C), but becomes signicant when countries are atheoretically grouped into
clusters based on their institutions empirical distributions (Table 3), and then is again
signicant or insignicant in the multivariate setting depending on the linear regression
specication (Table 5). Leuz et al. (2003) acknowledge in Section 4.4 that their choice of
institutions such as enforcement and ownership is motivated more by data and empirical
considerations than theory. First and foremost, Proposition 1 of this study shows how
institutions such as ownership and enforcement are theoretically linked to accounting
quality.24 Such a clustering of the sample is not immediately evident from a pure crime
deterrence model that does not account for the endogenous price eects of managerial
ownership on the managers reporting choices.
Our model not only identies institutions that drive reporting quality, but also shows
how these institutions interact to create a reporting regime. This feature of the model
is especially useful for empirical design. For example, Leuz et al. (2003) use a simple
multivariate linear regression model in their Section 4.4 to measure the impact of various
institutions on reporting quality. However, our Proposition 1 says that uniqueness of
outcomes obtains only when the regulatory force is strong and managerial incentives are
weak, or when the regulatory force is weak and managerial incentives are strong. Given
that empirical measures of regulatory strength and ownership incentives are already
available (e.g., Dechow et al. 2010), we rst suggest that the strongest empirical results
on aggregate earnings quality will obtain when researchers compare a subset of countries
with strong public and private regulatory forces and weak managerial incentive forces
24
Leuz et al. (2003) measure ownership patterns over all shareholders. Our model is more specic
in that it pertains to managerial shareholdings or incentives.

18
(where earnings quality is high) with countries with weak public and private regula-
tory forces and strong managerial incentives (where earnings quality is low), and ignore
countries in the middle (where earnings quality can be high or low due to multiplicity).25
The above across-subsample regression approach which drops the middle subsample
(where earnings quality could be high or low due to multiplicity), can be powerful in
many instances, but has some limitations. Conceptually, it does not capture any within-
group or local variation, and the researcher also loses sample size. The loss of sample
size can be sometimes crucial. What if the remaining sample is lopsided and contains
mostly only one of the subgroups, i.e., just the subgroup where all bad rms misreport,
or just the subgroup where no bad rms misreport? Observation 1 has full force here,
because it shows that the corner solutions have no variation when the exogenous variables
vary. Such lack of variation might make it dicult to detect any role in institutions in
reporting quality. In this case, bringing back the middle subsample can be advantageous,
for reasons described next.
The problem with the multiple-equilibria middle subsample is that within-group,
there is no 1:1 mapping between the exogenous parameters and aggregate reporting
quality: two economies with the same exogenous parameters could have all bad rms
misreporting and no bad rms misreporting, respectively (Proposition 1, part (3b)).
But despite this noise, the middle subgroup average of misreporting could be lower
than, say, the remaining subsample where M is very small.26 In such cases, the empirical
researcher may gain power on net by comparing the middle subsample with the remaining
subsample. In sum, the middle subsamples indeterminacy can sometimes add noise
that overwhelms any relation between institutions and reporting quality, in which case
removing the middle subsample is the preferred choice. In other cases, while noisy,
25
While Proposition 1 combines regulatory forces and managerial incentives into a single variable M
and gives precise threshold values for multiplicity and uniqueness, the empirical benets of estimating
these threshold values to create subsamples may be limited, given our models various specialized
assumptions. A more eective empirical approach would be to use cluster analysis and/or informal
cutos of the sample based on regulatory and incentive metrics to create the subsamples. See, e.g., the
empirical approach of Nagar and Yu (2014).
26
In the middle group, our model cannot predict the probability with which an economy lands in any
given equilibrium; the model only predicts that there are two stable equilibria. But for any probability
in (0, 1), the aggregate misreporting in the middle group will be lower than the misreporting in the
subsample whereM is very small.

19
the middle subsamples average is likely to be between the other two subgroups, and
including the middle subsample can add power on net by increasing sample size and
facilitating across-group comparisons of reporting quality.
This insight extends when we include a private cost of misreporting. As Observa-
tion 1 shows, the interior equilibrium is not like the corner solution; it moves when the
exogenous parameters move. As a result, there is both within-group and across-group
variation which the researcher can exploit to create more powerful tests. While the mid-
dle subsample with multiplicities still does not have a 1:1 mapping from the exogenous
parameters to the reporting outcomes, on average the misreporting level could still be
lower than, say, the subsample with weak institutions and a high unique level of mis-
reporting (e.g., Figure 4 versus Figure 3). As a result, including the middle subsample
can increase noise due to multiplicity, but can also facilitate across-group comparisons
by increasing sample size. Nagar and Yu (2014) is an example of a recent study that
illustrates the empirical challenges in subsample selection issues for testing models of
multiplicity.
In sum, our models emphasis on sample oers a counterpoint to the standard empir-
ical emphasis on measures, i.e., the view that mixed ndings arise from noisy empirical
measures of institutions and better measurement techniques are the solution (Holthausen
2009, 454). But such calls may not be eective suggestions if the underlying phenomenon
exhibits multiplicities. In that case, the empirical researcher is better o adjusting the
sample based on where multiplicities are likely. In fact, Leuz et al. (2003) take this
approach implicitly in Table 3, when they atheoretically cluster subsets of countries,
and compare across the clusters of subsets. Our model not only provides theoretical
foundations for such groupings of countries, but also suggests that Leuz et al. (2003)
could have gained more power by retaining this approach in other regressions such as
their Table 5, instead of incorporating all countries as they chose to do. For example,
their Table 5, which includes all countries, shows that enforcement is not associated with
enforcement if ownership (an important factor in our model) is controlled for. Our model
suggests that this non-result could have arisen not from a poor measure of enforcement,
but from a poor sample choice.

20
Another empirical study to which our model is relevant is Zakolyukina (2014). That
study applies structural methods to realized GAAP violation data to estimate unobserv-
able fundamentals such the probability of misstatement and detection for a given rm.
As with this model, Zakolyukinas structural model rests on managerial incentives and
a regulator. However, Zakolyukinas regulator has a constant probability of detection
irrespective of the prevalence of misreporting, an assumption that our model suggests
is not reasonable in the multiple equilibria case. Our model suggests that one empir-
ical way to justify Zakolyukinas assumption is to re-estimate her structural model in
periods where managers have lower stock-based incentives (our model suggests a unique
equilibrium in this case), and check if the estimated parameters are similar to the full
sample case. Such an empirical test, which would increase the robustness of Zakolyuk-
inas estimates, is not immediately apparent without our model (or even with a pure
regulatory deterrence model where endogenous rational capital market prices and the
resulting ownership incentive considerations play no additional disciplining role).

Theoretical contributions

Our model also adds to the theoretical accounting literature on misreporting. The
traditional focus of misreporting models has been on single-rm misreporting (Fischer
and Verrecchia 2000; Strobl 2013, 452). An important step towards multi-rm reporting
was taken by Lambert, Leuz, and Verrecchia (2007), who unpacked the CAPM by
rewriting returns in terms of future cash ows, and introduced accounting reports as
signals of these future cash ows. Subsequent studies such as Strobl (2013) built on
Lambert et al. (2007) by allowing managers to manipulate these accounting signals.
The key feature tying all the rms together is CAPMs exogenous correlation among
the rms cash ows, which drives both misreporting and investor pricing. Likewise,
in Povel, Singh, and Winton (2007), investors ex ante level of bullishness about the
economy drives the aggregate level of misreporting.
A common feature of all the above models is that the cost of detecting misreporting
is exogenously held constant (e.g., Povel et al. 2007, 1229). We relax this assumption,
and, following Becker (1968), assume that public enforcers have nite xed resources

21
and cannot evaluate every potential violation. Consequently, as more actors engage
in violations, enforcers will run out of resources to detect, convict, and punish these
malefactors. This enforcement thinning feature of Becker has not received sucient
attention, even in studies that apply Beckers idea to nancial markets (e.g., Shleifer
and Wolfenzon 2002). Our model takes a step in that direction.
There are several models that obtain coordination multiplicities by imposing various
exogenous restrictions on individual trading behavior: individual traders in Angeletos
and Werning (2006) do not have enough resources to attack the asset, while individual
arbitrageurs in Barlevy and Veronesi (2003) face nancial constraints that limit their
stock holding and maximum short positions. Traders limited resources force them
to coordinate, and this collective action externality, in conjunction with patterns of
each agents higher-order beliefs about other agents, leads to multiplicities and excess
volatility in prices (Angeletos and Lian 2016). The externality in our model occurs in
a dierent part of the economy, namely the legal institutions overseeing management
disclosures. In fact, traders in our model are nancially unconstrained and rational, a
feature that precludes trading-related externalities.
An important feature of the above trading models is that they need to assume in-
formation asymmetry among the agents in order to generate informationally strategic
trades. However, Morris and Shin (1998) show that multiplicity vanishes in settings
where each individual trader has even an arbitrarily small amount of private informa-
tion. Weinstein and Yildiz (2007) likewise show that multiple equilibria represent a
knife-edge case that can be collapsed to a unique equilibrium with an arbitrarily small
perturbation. These arguments were countered by Atkeson (2001) and Rey (2001) who
showed that multiplicity could be robustly recovered when agents with private informa-
tion also received a more precise public signal.27 Angeletos and Lian (2016, Section 4.8)
further note that the Weinstein-Yildiz renement does not privilege a specic equilib-
rium out of the multiple possible ones, but can be used to pick any one of them. They
27
Reys (2001) and Angeletos and Wernings (2006) Figures 1 make this point starkly, where no
matter what the level of the agents private information precision, there is a threshold precision of the
public signal beyond which multiplicity obtains all the way to common knowledge. From an economic
perspective, public signals emerge naturally and endogenously from public disclosures and market prices
arising from the trades of privately informed agents.

22
argue that in eect, multiplicity has not vanished at all. Summarizing the state of the
literature, Angeletos and Lian (2016, Section 4.8) conclude that the global games liter-
ature is not a panacea for getting rid of multiple equilibria, or for giving policy makers
the satisfaction of sharp policy advice. In our setting, therefore, the only information
that is private to each rm is its knowledge whether it is good or bad. All other in-
formation is common knowledge. We acknowledge that this information structure is a
signicant simplifying assumption.
Finally, Bertomeu and Magee (2011, 2015) develop models of reporting quality and
reporting cycles, where the regulators preference itself changes with economic perfor-
mance. This change occurs because dierent kinds of rms gain political power over the
regulator in dierent epochs. Our model suggests that such shifts in political power over
the regulator are not always necessary to generate reporting cycles: our model generates
multiple outcomes (and thus reporting cycles) by keeping the regulators budget and
strategy xed.

Policy Implications

In addition to the above empirical implications, our model also suggests a policy implica-
tion with respect to multiplicity. When an economic system exhibits multiplicity, small
perturbations can cause systemic waves that move the economy from one equilibrium
to the other, as noted in the previous section. Multiplicity in our model occurs when
regulator are constrained and managers have strong stock incentives as well as much
reporting discretion. These conditions appeared to be satised in the period leading up
to the U.S. mortgage crisis: the SEC was highly constrained (Labaton 2008), managers
had wide discretion for reporting the value of nancial instruments under reporting rules
such as SFAS 157 (Norris 2008), and stock price-based compensation was the industry
norm. An economy-wide misreporting wave was thus feasible, according to our model.
But more important, the same kind of temporary shock that can shift an economy to
the bad equilibrium can be deliberately applied in reverse as a policy choice to shift
the economy to the good equilibrium (Matsuyama 2005). This point is relevant to the
current policy debate on preventing future crises. Specically, this debate has centered

23
around the idea of strengthening regulation and curbing managerial incentives (e.g.,
Kashyap and Stein 2008; Blinder 2009; Department of Treasury 2009). However, the
political support for such permanent regulatory expansion is typically far from unied
(e.g., Wyatt 2010). A temporary expansion of regulatory oversight is more feasible
politically, and according to our model, likely to be eective in shifting the reporting
equilibrium from a high misreporting regime to a low misreporting regime.

IV. CONCLUSION

This study builds a model where management incentives, rational traders, and regula-
tory constraints contribute to create an economy capable of multiplicities in aggregate
reporting behavior. From an empirical perspective, this is an important theoretical re-
sult for two reasons. First, even after holding key institutional features constant, the
reporting quality across similar countries shows high variation (e.g., Ball, Robin, and
Kothari 2000). A model of reporting multiplicities is a useful way of thinking about these
cross-sectional phenomena. Second, our model also suggests that there could be time-
series variation in reporting quality in the same country. For example, the widespread
belief in the pre-crisis period in the US was that aggregate misreporting of balance sheets
could not occur in rational capital markets, and any departure from this axiomatic belief
(e.g., Rajan 2005) was met with considerable skepticism by mainstream economists (e.g.,
Knight 2005).28 Finally, we show that multiplicities are not an empirical deterrent, but
instead an important feature that empiricists can employ to construct more powerful
tests of the role of institutions in determining earnings quality. In particular, our model
shows how choosing the correct subsets of the sample can lead to more powerful tests.
Our emphasis on sample contrasts the empirical literature which focuses on improving
measures of institutions to gain power (e.g., Holthausen 2009, 454).
Our approach to modeling reporting phenomena at the economy-wide level is a con-
siderable departure from traditional accounting analytical models that explore the re-
porting choices of a single rm. The key tradeo we make is to simplify the analytical
28
By contrast, in a recent speech, Larry Summers has argued that the study of multiplicities, regula-
tion, capital markets, and accounting are central to understanding current macroeconomic phenomena.
See http://larrysummers.com/commentary/speeches/brenton-woods-speech/.

24
treatment of each individual rm in order to aggregate the model at the economy-wide
level. Such tradeos are necessary because single-rm models have limited relevance to
the exploding empirical literature on reporting phenomena at the country-level (see Sec-
tion 4 in Dechow et al.s (2010) review). Simplifying individual-rm phenomena to speak
to the aggregate is thus an important and increasingly necessary analytical endeavor.
For example, a model where the good rms can opt out of our reporting game by using
costly signals such as dividends can help us understand the empirically observed con-
nections between aggregate reporting and aggregate dividends and repurchases (Skinner
2008).

REFERENCES

Aghion, P., and P. Bolton. 1992. An Incomplete Contracts Approach to Financial


Contracting. Review of Economic Studies 59: 473-494.

Angeletos, G., and C. Lian. 2016. Incomplete Information in Macroeconomics: Ac-


commodating Frictions in Coordination. Working Paper 22297, NBER.

Angeletos, G., and I. Werning. 2006. Crises and Prices: Information Aggregation,
Multiplicity and Volatility. American Economic Review 96: 1721-1737.

Atkeson, A. 2001. Comments on Rethinking Multiple Equilibria in Macroeconomic


Modeling by Morris and Shin. NBER Macroeconomics Annual 15: 162-171.

Ball, R., S. Kothari, and A. Robin. 2000. The Eect of International Institutional
Factors on Properties of Earnings. Journal of Accounting and Economics 29: 1-
51.

Barlevy, G., and P. Veronesi. 2003. Rational Panics and Stock Market Crashes. Journal
of Economic Theory 110: 234-263.

Becker, G. 1968. Crime and punishment: An Economic Approach. Journal of Political


Economy 76: 169-217.

25
Bertomeu, J., and R. Magee. 2011. From Low-quality Reporting to Financial Crises:
Politics of Disclosure Regulation Along the Economic Cycle. Journal of Accounting
and Economics 52: 209-227.

Bertomeu, J., and R. Magee. 2015. Mandatory Disclosure and Asymmetry in Financial
Reporting. Journal of Accounting and Economics 59: 284-299.

Blinder, A. 2009. Economic View: The Wait for Financial Reform. The New York
Times, September 6.

Bond, P., A. Edmans, and I. Goldstein. 2012. The Real Eects of Financial Markets.
Annual Review of Financial Economics 4: 339-360.

Cass, D., and K. Shell. 1983. Do Sunspots Matter? Journal of Political Economy 91:
193-227.

Dechow, P., W. Ge, C. Schrand. 2010. Understanding Earnings Quality: A Review of


the Proxies, Their Determinants and Their Consequences. Journal of Accounting
and Economics 50: 344-401.

Department of the Treasury. 2009. Financial Regulatory Reform. A New Foundation:


Rebuilding Financial Regulation and Supervision. Washington, DC.

Fischer, P., and R. Verrecchia. 2000. Reporting Bias. Accounting Review 75: 229-245.

Foster, G. 1981. Intra-industry Information Transfers Associated With Earnings Re-


leases. Journal of Accounting and Economics 3: 201232.

Gao, P., and P. Liang. 2013. Informational Feedback Eect, Adverse Selection, and
the Optimal Disclosure Policy. Journal of Accounting Research 51: 11331158.

Graetz, M., J. Reinganum, and L. Wilde. The Tax Compliance Game: Toward an Inter-
active Theory of Law Enforcement. Journal of Law, Economics, and Organization
2: 1-32.

Grossman, S., and J. Stiglitz. 1980. On the Impossibility of Informationally Ecient


Markets. The American Economic Review 70: 393-408.

26
Hayek, F. 1945. The use of knowledge in society. American Economic Review 35:
519-535.

Hermalin, B., and M. Katz. 1991. Moral Hazard and Veriability: The Eects of
Renegotiation in Agency. Econometrica 59: 1735-1753.

Holthausen, R. 2009. Accounting standards, nancial reporting outcomes, and En-


forcement. Journal of Accounting Research 47: 447-458.

Kashyap, A., and J. Stein. 2008. The $700 Billion Question. The New York Times
(September 22).

Knight, M. 2005. General Discussion: Has Financial Development made the World
Riskier? Available at: http://www.kc.frb.org/publicat/sympos/2005/PDF/
GD5_2005.pdf.

Kothari, S., S. Shu, P. and Wysocki. 2009. Do Managers Withhold Bad News? Journal
of Accounting Research 47: 241-276.

Kothari, S., K. Ramanna, and D. Skinner. 2010. Implications for GAAP from an
Analysis of Positive Research in Accounting. Journal of Accounting and Economics
50: 246-286.

Kouwe, Z. 2009. In Harsh Reports on S.E.C.s Fraud Failures, a Watchdog Urges


Sweeping Changes. The New York Times (September 30).

Kreps, D. 1990. A Course in Microeconomic Theory Princeton, NJ: Princeton Univer-


sity Press.

Labaton, S. 2008. The Reckoning: Agencys 04 Rule Let Banks Pile Up New Debt,
and Risk. The New York Times (October 1).

Lambert, R., C. Leuz, and R. Verrecchia. 2007. Accounting Information, Disclosure,


and Cost of Capital. Journal of Accounting Research 45: 385-420.

27
Leuz, C., D. Nanda, and P. Wysocki. 2003. Earnings Management and Investor
Protection: An International Comparison. Journal of Financial Economics 69:
505-527.

Matsuyama, K. 2005. Poverty Traps. Available at: http://faculty.wcas.northwestern.


edu/~kmatsu/PovertyTraps.pdf.

Mollenkamp, C., S. Craig, J. McCrackern, and J. Hilsenrath. 2008. The Two Faces
of Lehmans Fall Private Talks of Raising Capital Belied Firms Public Optimism.
The Wall Street Journal (October 6).

Morris, S., and H. Shin. 1998. Unique Equilibrium in a Model of Self-Fullling Cur-
rency Attacks. The American Economic Review 88: 587-597.

Nagar, V., and G, Yu. 2014. Accounting for Crises. AEJ: Macroeconomics 6: 184-213.

Norris, F. 2008. S.E.C. rule may relax asset rule. The New York Times (October 1).

Povel, P., R. Singh, and A. Winton. 2007. Booms, Busts, and Fraud. Review of
Financial Studies 20: 1219-1254.

Rajan, R. 2005. Has Financial Development Made the World Riskier? Available at:
http://www.kc.frb.org/publicat/sympos/2005/PDF/Rajan2005.pdf.

Rajan, R., and L. Zingales. 2004. Saving Capitalism from the Capitalists: Unleash-
ing the Power of Financial Markets to Create Wealth and Spread Opportunity,
Princeton, NJ: Princeton University Press.

Rey, H. 2001. Comments on Rethinking Multiple Equilibria in Macroeconomic Mod-


elling by Morris and Shin. NBER Macroeconomics Annual 15: 171-178.

Romer, D. 1996. Advanced Macroeconomics New York, NY: McGraw-Hill.

Schwartz, R. 1997. Legal Regimes, Audit Quality, and Investment. The Accounting
Review 72: 385-406.

28
Shleifer, A. 2005. Understanding Regulation. European Financial Management 11:
439-451.

Shleifer, A., and D. Wolfenzon. 2002. Investor Protection and Equity Markets. Journal
of Financial Economics 66: 3-27.

Skinner, D. 2008. The evolving relation between earnings, dividends, and stock repur-
chases. Journal of Financial Economics 87: 582-609.

Strobl, G. 2013. Earnings Manipulation and the Cost of Capital. Journal of Accounting
Research 51: 449-473.

Weinstein, J., and M. Yildiz. 2007. A Structure Theorem for Rationalizability with
Application to Robust Predictions of Renements. Econometrica 75: 365400.

Wyatt, E. 2010. A Scale-back is Possible in Financial Overhaul Law. The New York
Times (November 3).

Zakolyukina, A. 2014. Measuring Intentional GAAP Violations: A Structural Ap-


proach. Working paper, University of Chicago.

29
APPENDIX A

Excerpts from the S.E.Cs Testimony Before The Subcommittee On


Financial Services And General Government
Chairman Mary Schapiro, March 11, 2009
http://www.sec.gov/news/testimony/2009/ts031109mls.htm

The last year has been a wrenching time for the investors whom the SEC is charged
with protecting ...
One of my very rst actions as Chairman was to end the two-year penalty pilot
program, which had required the Enforcement sta to obtain a special set of approvals
from the Commission in cases involving civil monetary penalties against public compa-
nies as punishment for securities fraud ... Another change I implemented to bolster the
SECs Enforcement program was to provide for more rapid approval of formal orders of
investigation.
It is clear that, regardless of the ultimate ndings of the Inspector General, the
agency needs to improve its ability to process and pursue appropriately the more than
700,000 tips and referrals it receives annually ... In addition, the examination stas of
the SEC and FINRA are working together to identify better ways that incipient frauds
might be detected at an early stage ...

SEC Resources

Few of the initiatives I have identied can be implemented with the SECs existing
resources. Most of this agenda will require additional funding, particularly to rebuild
the agencys workforce and invest in new technologies ...
The agency has suered a signicant decline in stang levels, due to several years
of at or declining budgets. Between 2005 and 2007, the agency lost 10 percent of its
employees, a decline that inevitably aected all of the SECs major programs ...
Yet as the SEC sta has declined, the securities markets grew dramatically. For
example, since 2005 the number of investment advisers registered with the Commission
has increased by 32 percent and their assets under management have jumped by over

30
70 percent (to now more than $40 trillion) ... The SEC oversees more than 30,000
registrants including 12,000 public companies, 4,600 mutual funds, 11,300 investment
advisers, 600 transfer agencies, and 5,500 broker dealers. We do this with a total sta
of 3,600 people.
In the context of such rapidly expanding markets, I believe the recent reductions
in the SECs sta seriously undermined the agencys ability to eectively oversee the
markets and eectively pursue violations of the securities laws ...
Then, if we hope to restore the SEC as a vigorous and eective regulator, I believe
we must go even further. The President is requesting a total of $1.026 billion for the
agency in FY 2010, a 9 percent increase over the FY 2009 appropriation ... It will fund
an additional 50 sta for the SEC, enhance our ability to uncover and prosecute fraud,
and begin to build desperately needed technology.

31
APPENDIX B

Proof of Lemma 1. For brevity, we will not put the multiplication sign in front of
F . We assume that F is large enough that the numbers MF , etc., are close to integers
within some small error (this is possible because rationals are dense in the reals). The
number of all possible orderings of 1 reporters is (MF + NG F )!. Because all 1 reporters
look identical to the regulator, he randomly picks one of the orderings.
Consider a given misreporter. In (MF + NG F 1)! of these orderings, it occupies
the rst place; in (MF + NG F 1)! of these orderings, it occupies the second place; in
(MF + NG F 1)! of these orderings, it occupies the third place, and so on, till the nal
(MF + NG F 1)! of these orderings, where it occupies the last or MF + NG F place.29
So, in KF (M(MFF+N G F 1)!
+NG F )!
= K
M +NG
fraction of the orderings, the given misreporter is in
the rst KF places.
If the given misreporter has not been caught yet, the regulator has incentives to
spend the budget because every dollar he spends has a strictly positive probability of
netting this rm. Consequently, the rm is sure to get caught if occupies anywhere
within the rst KF spaces of the regulators list, and will be sure to go undetected if
it occupies KF + 1 space or higher. Since all orderings are ex ante equally likely, the
K
ex ante probability of getting caught is M +NG
. This result holds as F , restricting
to a subsequence that makes the numbers MF , etc., close to integers within some small
error.


Proof of Lemma 2. We will start with a nite economy of size F rms. For brevity,
we will not put the multiplication sign in front of F . We also start with an F large
enough that the numbers MF , etc., are close to integers within some small error (this
is possible because rationals are dense in the reals). We will take the limit F ,
restricting to a subsequence that makes the numbers MF , etc., close to integers within
some small error.
29
It is easy to see that (M F + NG F 1)!(M F + NG F ) = (M F + NG F )!.

32
At the beginning of the model in period 1, the nal Time 3 report of a misreporting
rm can be thought of as independent random variable xi , (i = 1..MF ), which takes
K K
the value 1 with probability 1 NG +M
; and 0 with probability NG +M
. By contrast, the
reports of the NG F good rms will always stay at 1 in Time 3. Consequently, at Time
1, the total number of the nal 1 reports in Time 3 can be represented by the random
variable: NG F + x1 + x2 + ... + xM F . Every managers Time 1 estimate of the price of
a nal Time 3 report of 1 will be the random variable:

NG F NG
= x1 +x2 +...+xM F
(7)
NG F + x1 + x2 + ... + xM F NG + M MF

The above nonlinear function of random variables xi can complicate the pricing
analysis considerably. The standard approach in the global games literature is to send
the size of the economy, F , to innity, and use the law of large numbers and the Slutsky
formula to eliminate this randomness (thus eectively making each rm a price-taker).
In particular, recall that the regulator scans the rms randomly, so the random variables
xi are independent, and the law of large numbers and the Slutsky formula apply:30

NG NG NG
lim x1 +x2 +...+xM F
= = K
(8)
F NG + M MF
NG + MExi NG + M(1 NG +M
)

In this case, managers know in advance (in Time 1) that a nal report of 1 in Time
3 will be priced in Time 3 as:31

NG
KM
(9)
NG + M NG +M


30
For example, consider the global games model in the well-cited study Angeletos and Werning
(2006). That study envisions an economy where each agent i receives a random signal xi on the
fundamental (the study drops the subscript i for notational clarity). On p. 1726, the study derives
the standard Grossman and Stiglitz (1980) price-taking realized demand for each trader as k(x, p) =
x
(x p). Then the study sums the individual realized demands over the innite number of traders
to get the aggregate realized demand of the unit continuum of traders as K(, p) = x ( p). Now,
this would not be possible if the economy only had a nite number of traders, because then the sum
(or average) of the individual random xi s would not precisely yield .
31
Because the equilibrium M > 0 must also satisfy K < NG + M , the price in equation (1) will be
the range (0, 1).

33
Proof of Proposition 1. The fundamental intuition behind the proof is that both
the benets and the costs of misreporting varies with the number of misreporters. This
can lead to multiple xed points in the number of misreporters. However, as we will see,
in order to identify these xed points, one has to ennumerate a fair number of cases.
Lemma 1, Part 2 is used throughout the proof to compute the probability of a
misreporter getting caught. Note that a truth-telling bad (0 value) rm always makes 0,
and a good (1 value) rm always reports 1. So the subsequent action of the regulator is
irrelevant to these rms. The only rm that cares about the regulators ex post behavior
is the misreporting rm, and that is where Lemma 1, Part 2 applies.
We compare the net benets to misreporters to 0 because in our equilibrium investors
will always believe (and value) a report of 0 at 0. This is justied as follows: the payo
to reporting 1 in equation (1) > 0; so no good rm will report 0 under such investor
belief. In equilibria where only bad rms report 0, this investor belief is justied. In
equilibria where no rm reports 0, we show in the body of the paper that this belief
satises the intuitive criterion. Comparing the net benets to misreporters to 0 will turn
to be comparing a quadratic equation to zero. Several cases arise because a quadratic
equation can be convex or concave, and may not have real roots; we have to check for
all possibilities.
Recall that all exogenous constants NB , NG , N NB + NG , K, C are strictly positive
and K < N NB + NG . Building on the proofs of the two lemmas, we drop F without
loss of generality, because it is a scaling variable.
The endogenous variable is M, where M is the number of the NB 0 value rms that
report 1 in equilibrium. That is, a total of M + NG rms report 1. Our goal is to
characterize the equilibrium value of M. Specically, we want to fully specify when M
takes a corner solution and when M takes an interior solution.
Note that we also have to make sure that any solution M > 0 satises K < NG + M.
This condition ensures that all probabilities and prices are positive in equation (10)
KM
below. In that case, M 0 value rms report 1, of which NG +M
get caught, and investors

34
KM
lump the remaining M NG +M
with the NG rms that are truly worth 1, resulting in
NG
a price of NG +M NKM
for rms that report 1. The ex ante payo to the manager of a
G +M

0 value rm to reporting 1 is:

  
K NG K
1 KM
+ (0 C) (10)
NG + M NG + M NG +M
NG + M

Note on Notation: It is easier to couch the equation (10) in terms of x where


x = NG + M is the total number of rms reporting 1. Our proof will largely use x, but
occasionally for ease of exposition, we will use M. Note, however, that there is only one
endogenous variable, and x and M are its two representations.
The payo to the manager of a 0 value rm from reporting 1 is:

  
K NG K
1 + (C) (11)
xx K(xN G) x
x
(x K)NG KC
= 2 (12)
x Kx + KNG x
x (NG KC) xK(NG KC) KNG KC
2
= (13)
(x2 Kx + KNG )x

In equilibrium, this payo has to be 0. We therefore have to examine both the


numerator and the denominator of equation (13).
We rst note the following observations:

Observation 2 The denominator of the fraction in equation (13) is positive when x >
K > 0 because x2 Kx + KNG is positive when x > K > 0.

We next consider the case when x < K.

Observation 3 If x < K in equilibrium, the regulator can inspect all rms reporting 1.
No 0 value rm will report 1 in that case and the equilibrium M = 0. Also note that
equation (10) is no longer valid.

We rst attempt to ascertain the conditions under which an interior equilibrium


solution M obtains. Such an equilibrium should satisfy:

35
1. 0 < M < NB

2. K < NG + M
  
K NG K
3. 1 NG +M N KM + NG +M
(C) =0
G +M N +M
G

The rst condition is interiority. The second condition is that the regulator does
not have enough resources to catch all misreporting rms (otherwise the equilibrium is
M = 0). The third condition is that misreporting 1 should have the same ex ante payo
for a 0 rm as truthfully reporting 0.
Rewriting the equilibrium in terms of the endogenous variable x (the total number
of rms reporting 1) the feasibility conditions become NG < x < N NB + NG and
K < x. The equilibrium condition becomes:

x2 (NG KC) xK(NG KC) KNG KC


=0 (14)
(x2 Kx + KNG )x

We have not yet proved the existence of an interior equilibrium point, but should
such an equilibrium exist, Observation 3 applies, and we can only consider solutions for
x > K. This implies that Observation 2 applies and the denominator in equation (14)
is strictly greater than zero in an interior equilibrium (if any); we can thus clear the
denominator.
We can therefore simply solve the numerator quadratic equation below to get the
interior solution (if any):

x2 (NG KC) xK(NG KC) KNG KC = 0 (15)

We next investigate whether the numerator quadratic equation (15) has a solution
x > K. The roots are:


K K 4NG C
1+ (16)
2 2 NG KC
The roots of the above equation determine where there is an interior solution. How-
ever, note that interior solution is not the only possible solution. There are two corner

36
solutions as well: all 0 value rms misreport (x = N) or no 0 value rms misreport
(x = NG ). We have to check the feasibility of all these possibilities. We rst begin with
the following observation:

Observation 4 When NG KC > 0 equation (15) is a convex upward facing parabola


in x (the second derivative with respect to x is positive). It also has two real roots because
it is negative at x = 0. The parabola is negative between the these roots, and positive
outside. By the same token, when NG KC < 0, the parabola is downward facing and
concave. The parabola has either two real roots or no real roots. If it has two real roots,
it is positive between these two roots and negative outside. If it has no real roots, it is
negative always (checked by evaluating the quadratic at zero). Finally, if NG KC = 0,
the quadratic is always negative with value KNG KC.

We now analyze all the possible solutions (interior and corner) for all the cases.
1. Case when NG KC:
If NG KC then one can see from equation (16) that an interior solution greater
than K is not possible: either the real roots are less than K or there are no roots. So
we have excluded the interior solution. We now check for possible corner solutions. For
this we need to examine the following cases:
1a. Case NG KC with real roots:
We show that the corner equilibrium x = N > K cannot happen. Such an x is greater
than both roots. Observation 4 tells us that the numerator quadratic of equation (14)
is negative. Observation 2 tells us that the denominator of equation (14) is positive. It
does not pay for any 0 value rm to report 1; it is better to report 0 and gain 0. The
optimal is therefore the corner solution M = 0; all rms report truthfully.
1b. Case NG KC with no real roots:
If there are no real roots, Observation 4 tells us the numerator quadratic in equation
(14) is always less than zero for all x, and thus for all x > K, including the corner
value x = N > K. This negativity condition also holds if NG = KC. Furthermore,
Observation 2 yields a positive denominator. A 0 value rm is therefore always better
o reporting 0 and M = 0.

37
The only equilibrium again is all rms reporting truthfully. This proves the rst part
of Proposition 1.
2. Case when NG > KC:
If NG > KC, we have two real roots:


K K 4NG C K K
x + 1+ > + =K (17)
2 2 NG KC 2 2

K K 4NG C K K
x 1+ < =0 (18)
2 2 NG KC 2 2

The rst root thus satises x > K, and by extension Observation 2. However, we
cannot state that the interior equilibrium M = x NG because we still do not know if
0 < x NG < NB . We must check this by hand for the given values of the exogenous
variables. But we rst establish a simple lemma on the numerator quadratic (equation
15) based on Observation 4:

Lemma 3 If NG > KC, the quadratic equation (15), x2 (NG KC) xK(NG KC)
KNG KC, in x crosses zero with a positive slope at x , and remains above zero for all
x > x .

Proof: Dierentiating x2 (NG KC) xK(NG KC) KNG KC with respect to


x and evaluated at x gives us:

2x (NG KC) K(NG KC) = (2x K)(NG KC) > 0

The positive sign obtains because x > K (this, as proved above, is a consequence
of NG > KC). Since x is the larger root, and the parabola is convex (the coecient
(NG KC) on x2 is greater than zero), it remains above zero for x > x (Observation
4). 

Eectively, Lemma 3 tells us the behavior of a convex quadratic at its right or


larger root x .
We now examine all possibilities for x NG :

38
2a. Case when NG > KC and 0 x NG < NB :
There are three equilibria.
The rst equilibrium is all rms telling the truth. If no 0 value rm is reporting 1,
the equilibrium x = NG , i.e., M = 0. Note that x < 0 < x = NG x . That is,
x = NG is between the two real roots. Therefore, when evaluated at x = NG , the left
hand side of equation (14) is negative because the numerator is zero (Observation 4)
and the denominator is positive at x = NG (i.e., (NG )2 KNG + KNG > 0). There is
thus no benet for a 0 value rm to reporting 1 when no other 0 value rm is. In the
corner case when x = NG , the 0 value rms are indierent, and we assume they tell the
truth. Therefore all rms tell the truth.
The second equilibrium is all 0 value rms reporting 1. Now x = N > x , and the left
hand side of equation (14) is positive (or zero in the corner case) because the numerator
is above zero (or zero in the corner case) due to Lemma 3 and the denominator is positive
(Observation 2 applies because x = N x > K). There is a positive benet for a 0
value rm to reporting 1 when all other 0 value rms are. Therefore, all rms in the
economy report 1.
The third equilibrium is the knife-edge where the equilibrium M = x NG . That
is, M = x NG of the 0 value rms report 1 and the remaining 0 value rms report
0. All managers of 0 value rms earn 0 in expectation. Equation (14) as well as all the
feasibility and interiority conditions are thus satised.
Finally, note that the number of rms reporting 1 is dierent in each possible equi-
librium; therefore, everyone can ascertain which equilibrium the economy is in.
2b. Case when NG > KC and x NG < 0:
In this case, the minimum feasible value of x is greater than x , i.e., x > x > K.
Therefore the left hand side of equation (14) is positive at the feasible values of x because
the numerator is positive (Lemma 3) and the denominator is positive (Observation 2
applies because x > K). The numerator will never come back to zero as x increases all
the way to N because the quadratic equation (15) has no zeroes larger than x (the only
other zero x is smaller than x ). Thus, reporting 1 is always better than reporting 0.
The only equilibrium is therefore all rms reporting 1.

39
2c. Case when NG > KC and x NG NB :
In this case, because x N, the feasible range of x is x . If x is in the range
x K, the regulator can inspect all rms and therefore M = 0. All rms report the
truth.
Next consider the feasible range N x > K. Because x x in this range, the left
hand side of equation (14) is negative because Lemma 3 indicates that the numerator
is below zero (or zero in the corner case). The numerator will stay below zero until x
goes all the way down to the other root x . That root is smaller than zero and is thus
below the positive feasible range of x. In addition, the denominator of equation (14) is
positive because x > K and Observation 2 applies. Therefore, the payo for a 0 value
rm for reporting 1 is strictly negative. Therefore no 0 value rm will report 1 and all
rms in the economy report the truth. In the corner case when x = N, the 0 value
rms are indierent; we assume that they report the truth. 

40
0 1 2 2 2 3 3

Managers learn Managers issue a Regulator Regulator Some rms Investors price Managers
the true value of report of rm inspects some discloses some remain the rms collect payos
their rms value rms misreporting undetected
rms and
imposes
penalties on
managers

Figure 1: Timeline of Events in the Reporting Game

41
Expected Benefits to Misreporting
1.0

C = 0.02
0.5

C=1
M
200 400 600 800

0.5 C=2

1.0

Figure 2: Expected payos of equation (3) to misreporting 0 as 1 as a function of M, the


number of misreporting rms. NG = 150, NB = 850, K = 100. There are thus 150 rms with
value 1 and 850 rms with value 0. The equilibrium number of misreporting rms M [0, 850].
If C = 0.02, all 0 value rms misreport. If C = 2, no 0 value rms misreport. If C = 1, there
are two stable equilibria: all 0 value rms misreport, or no 0 value rms misreport; and one
unstable equilibrium where 80.27 of the 850 rms misreport.

42
Figure 3: Uniqueness: Expected payos of equation (5) to misreporting 0 as 1 as a func-
tion of M, the number of misreporting rms, with private misreporting costs. The private
cost of misreporting is drawn as the horizonal line m = 0.2. In addition, C = 0.2, NG =
150, NB = 850, K = 100. There are thus 150 rms with value 1 and 850 rms with value 0.
The equilibrium number of misreporting rms M [0, 850]. The unique stable equilibrium is
the intersection point M = 565.28 (the other two intersection points are negative).

43
Figure 4: Multiplicity: Expected payos of equation (5) to misreporting 0 as 1 as a function of
M, the number of misreporting rms, with private misreporting. The private cost of misreport-
ing is drawn as the horizonal line m = 0.4. In addition, C = 0.2, NG = 150, NB = 850, K =
100. There are thus 150 rms with value 1 and 850 rms with value 0. The equilibrium number
of misreporting rms M [0, 850]. The multiple stable equilibria are M = 0 and the intersec-
tion point M = 139.46. The unstable equilibrium is the second intersection point M = 31.96.
(The third intersection point is negative.)

44

You might also like