Earnings Manipulation and the Cost of Capital
G¨
unter Strobl∗
Kenan-Flagler Business School
University of North Carolina at Chapel Hill
August 2008
Abstract
The widespread use of accounting information by investors and financial analysts
to help value stocks creates an incentive for managers to manipulate earnings in an
attempt to influence short-term stock price performance. This paper examines the role
of earnings management in affecting a firm’s cost of capital. Using an agency model with
multiple firms whose cash flows are correlated, we demonstrate that the extent of earnings
manipulation varies across the business cycle. Managers are more inclined to engage in
manipulation during periods of economic expansion. Because of this dependence on the
state of the economy, earnings manipulation influences a firm’s cost of capital despite the
forces of diversification. In particular, we find that manipulation reduces the correlation
between the firms’ cash flows and thus lowers the risk premium required by investors.
JEL classification: G12, M41
Keywords: Earnings management; Information manipulation; Cost of capital
∗ Please address correspondence to G¨
unter Strobl, Kenan-Flagler Business School, University of North
Carolina at Chapel Hill, McColl Building, C.B. 3490, Chapel Hill, NC 27599-3490. Tel: +1 919 962 8399;
Fax: +1 919 962 2068; E-mail address: strobl@unc.edu

---

1
Introduction
The last few years have witnessed a number of corporate scandals that have created the public
perception that accounting information provided in a corporate culture fixated on stock price
performance cannot be trusted. While media attention has focused on a few high-profile
cases of fraudulent accounting schemes (e.g., at Enron and WorldCom), recent empirical
studies suggest that the practice of earnings management is prevalent among publicly traded
companies.1 The findings indicate that firms manage earnings to influence stock market
perceptions, to increase management’s compensation, to reduce the likelihood of violating
lending agreements, and to avoid regulatory intervention.2
In this article, we investigate the role of earnings management in affecting a firm’s cost of
capital. Given the importance of a firm’s cost of capital for a variety of corporate decisions
(from determining the hurdle rate for investment projects to influencing the composition
of the firm’s capital structure), it is surprising that the link between cost of capital and
earnings management has received little formal scrutiny. To date, the theoretical literature
has primarily focused on identifying conditions under which earnings manipulation emerges in
a single-firm setting. While this literature has provided many useful insights, its applicability
to cost of capital issues is limited. In a single-firm setting, firm-specific risk is priced, because
there are no alternative securities that would allow investors to diversify away idiosyncratic
risk. It is unclear, however, to what extent accounting information reduces non-diversifiable
risks in a multi-asset economy.
In this paper, we take up this task, and present a simple but rigorous model of earnings
manipulation with multiple firms whose cash flows are correlated. Important features of our
model are risk averse investors, myopic managers, and resource costs of manipulation. Man-
agers are concerned about short-term stock prices because of their compensation contract.
1According to the General Accounting Office, about 10 percent of all listed companies in the US announced
at least one earnings restatement between January 1997 and June 2002 (see GAO (2002)).
2See, e.g., the review of the earnings management literature by Healy and Wahlen (1999).
1

---

The use of accounting information by investors to value stocks therefore creates an incentive
for managers to manipulate earnings. We define earnings manipulation as any action that
enables low-value firms to report the same earnings as high-value firms.3 Such action is costly
for managers. The effort involved in manipulating earnings reports and the chance of being
caught and punished represent a disutility to managers. In addition, earnings manipulation
reduces the value of the firm, because resources are diverted from more productive endeavors.
This diverted-resource cost lowers the manager’s compensation payments.
Our analysis leads to two main findings. First, we demonstrate that when managers care
about short-term stock prices, the extent of earnings manipulation depends on the state of
the economy. In particular, we show that managers of low-value firms are more inclined to
engage in manipulation when the stock market is booming. The intuition for this result is as
follows. When business conditions are good, most firms have high earnings. Investors thus
correctly believe that few firms have an incentive to manipulate their accounting statements.
This means that reported earnings have a great effect on a firm’s stock price. A favorable
report leads to a significantly higher price than an unfavorable one. However, this is exactly
when the incentives for a manager of a low-value firm to issue an upwardly biased report are
highest. In bad times, on the other hand, incentives are low, because investors expect a large
number of firms to manipulate their earnings and, hence, put less emphasis on the observed
reports. Our model therefore predicts the most severe manipulation attempts to occur when
the economy is performing well. This prediction seems to be consistent with recent corporate
scandals. Many firms (including Enron, WorldCom, and Global Crossing) were found guilty
of fraudulent accounting during the economic boom of the late 1990s.
Our second result shows that earnings management influences a firm’s cost of capital,
despite the forces of diversification. We find that in large economies, manipulation can sig-
nificantly lower the risk premium that investors require to hold a stock in equilibrium by
3This includes cases of fraudulent accounting in which firms get auditors to approve statements that are
inconsistent with accounting standards, as well as cases in which firms take actions within accepted accounting
and legal standards to improve their accounting performance.
2

---

reducing the stock’s beta.4 This result is driven by the more frequent occurrence of manipu-
lation attempts in periods of economic expansion and crucially depends on the negative effect
of manipulation on firm value. If managers follow different strategies across the business cycle,
the diverted-resource cost of manipulation will have an impact on the correlation structure
of the firms’ cash flows. Assuming that managers are less likely to manage earnings in bad
times, earnings manipulation leads to a smaller reduction in firm value when the economy is
performing poorly and the marginal utility of consumption is high. This reduces a stock’s
required return in equilibrium. Viewing this result from the perspective of the firm, a firm
with relatively stronger incentives to engage in manipulation during a boom, compared to its
incentives during a recession, thus faces a lower cost of equity capital. We want to emphasize
that this result does not rely on the existence of naive investors who underestimate the extent
of manipulation. Rather, it is derived under the assumption that all investors are perfectly
rational and correctly anticipate the extent of manipulation in equilibrium.
Our model highlights the importance of disclosure requirements for the manager’s manip-
ulation decision. Common intuition suggests that forcing firms to disclose more information
to the public reduces the incidence of manipulation. Our analysis shows that this conclu-
sion is not always correct. All else equal, more informative accounting disclosures (in the
absence of manipulation) increase the expected quality of a firm that reports high earnings,
and, hence, lead to a higher stock price. This, in turn, increases the manager’s incentives
to engage in manipulation when earnings are low. Thus, our results suggest that stricter
disclosure requirements can be counterproductive in terms of reducing the extent of manipu-
lation. Unless disclosure laws make manipulation more costly for managers, more disclosure
may actually lead to more manipulation. This adds a new perspective to the debate on how
disclosure regulation can be used to curb earnings manipulation.
Prior theoretical work has primarily focused on managerial incentives to manage earnings
4We want to point out, however, that, controlling for beta, there is no cost of capital effect in the cross
section.
3

---

in a single-firm setting. Dye (1988), Evans and Sridhar (1996), and Demski (1998) present
models in which the manager is, by assumption, unable to communicate relevant private
information to shareholders. In these models, the revelation principle is not applicable and
the optimal managerial compensation contract encourages earnings management. Bar-Gill
and Bebchuk (2003) develop a model in which firms misreport their performance in order
to obtain better terms when raising funds for new investments. Goldman and Slezak (2006)
consider a variation of the principal-agent model where the agent can take a costly but unob-
servable action to manipulate disclosed information. They show that the optimal managerial
compensation contract balances incentives to exert effort against incentives to commit fraud.
However, none of these papers examines the effect of changing economic conditions on the
manager’s incentives to engage in manipulation, and thus do not shed any light on the ques-
tion of how earnings management affects a firm’s cost of capital. In this respect, our approach
is closer to the work of Povel, Singh, and Winton (2007), who study fraud and monitoring
decisions in a setting with multiple firms that seek financing from outside investors. Their
results show that a firm’s decision to commit fraud depends on the investors’ beliefs about
the state of the economy. Firms are more likely to manipulate their financial reports in rela-
tively good times (as measured by the average quality of firms seeking financing). However,
Povel, Singh, and Winton do not analyze how the firms’ manipulation decisions affect security
prices.
The remainder of this paper is organized as follows. Section 2 presents the economic
setting. Section 3 describes the equilibrium of the model and analyzes the manager’s ma-
nipulation decision and its effect on the firm’s stock price. Section 4 considers the impact of
earnings manipulation on a firm’s cost of capital in a large economy. Section 5 provides a
short summary and conclusion. All proofs are contained in the Appendix.
4

---

2
The Model
We consider an economy with N all-equity financed firms indexed by i = 1, . . . , N . The
model takes place over times 0, 1, and 2. At time 0, the initial owners of a firm sell their
shares to risk averse outside investors. Prior to trade at time 1, a certified third-party monitor
issues a (potentially misleading) report concerning the future value of the firm. Based on
this information, the market determines the intermediate stock price at which the manager
sells her equity stake in the firm. At time 2, conclusive public information arrives and the
firm is liquidated. Besides the firm’s shares, market participants can also invest in a riskless
bond. The bond is in perfectly elastic supply and its interest rate is normalized to zero. The
structure of the economy is common knowledge.
2.1
Firms and Managers
Each firm i is controlled by a manager who owns sM shares of the firm. The remaining s
i
i
shares are issued to the public at time 0. Each share of firm i’s stock pays a liquidation
value of P2,i = e1,i + e2,i at time 2, where et,i is referred to as the firm’s economic earnings in
period t = 1, 2. To focus on the manager’s incentives to manipulate earnings, we consider a
simple setting in which earnings as well as informational variables are binary. In particular,
we assume that, before any action is taken by the manager, the earnings distribution is
described by the following factor structure:
+1, with probability (1+βiF)/2,
et,i = µi + θt,i σi,
θt,i = 
(1)
−1, with probability (1−β

i F )/2,
where βi ∈ (0, 1) measures the sensitivity of firm i’s expected payoff with respect to the
systematic factor F . We assume that F is equally likely to be +1 (“good state”) or −1 (“bad
state”), and that the firms’ earnings are stochastically independent conditional on the factor
5

---

realization, i.e., Cov[et,i, es,j | F ] = 0 for all s, t ∈ {0, 1} and i = j. As is standard in factor
models, any comovement of firm values is thus captured by the common factor F . The values
of µi, σi, and βi are common knowledge, but the realizations of θt,i and F are not revealed
until time 2. The assumption that the same factor realization determines the firms’ earnings
in both periods is only made for tractability and is not crucial to our results. Shares of the
stock are infinitely divisible and are traded competitively in the stock market. The price of
stock i at time t = 0, 1 is denoted by Pt,i.
Before the stock market opens at time 1, a certified third-party monitor provides a (noisy)
signal ri to the market concerning firm i’s terminal value. This signal, which we refer to as an
earnings report, can take on one of two values, +1 or −1. Absent any managerial intervention,
the report ri is correlated with firm i’s first-period earnings as follows:
P r(ri = θ1,i | θ1,i, no manipulation) = (1 + δi)/2,
δi ∈ (0, 1).
(2)
The parameter δi measures the quality of the auditor’s report. It represents various account-
ing standards and conventions in the economy as well as firm- and auditor-specific factors
such as the auditor’s experience in the industry and the transparency of the firm’s operations.
Although the report is made by a third-party monitor, we assume that the manager can in-
fluence its outcome—for example, by hiding information from the auditor or by colluding
with the auditor to issue a biased report. Such an effort increases the probability that a firm
with low economic earnings generates a favorable signal:
P r(ri = +1 | θ1,i = −1, manipulation) = (1 − δi)/2 + κi(F )
(3)
We assume that manipulation does not affect the earnings reports of firms with high first-
period earnings. Allowing these firms to manipulate their reported earnings would not alter
our basic conclusions, so long as low-payoff firms benefit more from manipulation than high-
6

---

payoff firms. For expositional purposes, we refer to κi(F ) as the amount or the extent of
earnings manipulation. We allow the manager to optimally choose the amount of manipula-
tion after observing the factor realization F . This dependence on the state of the economy
reflects the fact that the manager’s incentives to issue a biased report will depend on the
investors’ beliefs about the firm’s earnings and, hence, on the general business conditions.
Since F is binary, the manager’s manipulation strategy is fully characterized by the vector
κi = (κ+, κ−), where κ+ = κ
= κ
i
i
i
i(+1) and κ−
i
i(−1).
Earnings manipulation is costly to the manager. Specifically, we assume that the man-
ager’s private utility cost associated with the amount of manipulation κi(F ) is equal to
ci κ2(F ). This cost includes both the effort involved in manipulating the auditor’s report and
i
the chance that the manager is later caught and punished.5 The cost parameter ci is related
to the legal environment in the economy, but may also depend on firm-specific factors such as
the intensity of outside monitoring (by analysts, the media, etc.) that affect the manager’s
opportunities to misreport the firm’s performance.
In addition, by diverting resources from more productive uses, earnings manipulation is
costly, because it reduces the value of the manager’s equity stake. In particular, we assume
that the actions taken by the manager to bias the auditor’s report carry an opportunity cost
that lowers the terminal value of the firm as follows:
P r(θ2,i = +1 | F, manipulation) = (1 + βi F )/2 − ξi κi(F ),
(4)
where ξi ≥ 0 denotes the marginal resource cost.6 This cost includes the opportunity cost
of the time managers spend hiding information from auditors, the cost of bribing auditors,
5One justification for assuming a convex cost function could be that both the detection probability and
the penalty are increasing in the extent of the manipulation.
6This assumption seems reasonable considering the effort Congress has invested in devising new measures to
combat earnings management. For instance, the Sarbanes–Oxley Act of 2002, which increased enforcement and
introduced new penalties for fraudulent behavior, is the result of an on-going attempt by Congress to mitigate
earnings management. Even NASDAQ has issued new guidelines requiring its listed firms to have financially
literate audit committees. Thus, there seems to be a widespread perception that earnings management is
detrimental to shareholders’ interests.
7

---

and the cost of making inefficient decisions based on inaccurate information. The following
assumption ensures that the probability in (4) is positive for any κi(F ) < δi.7
Assumption 1 ξi ≤ (1 − βi)/(2 δi).
Managers are assumed to be risk neutral. They own sM shares of their respective com-
i
panies, which they are not allowed to sell prior to time 1. In addtion, they are assumed to
be cash-constrained and thus cannot purchase additional equity at time 0. We take sM as
i
given and do not solve for the optimal compensation contract. Goldman and Slezak (2006)
show how sM can be endogenized in a model with unobserved managerial effort. If man-
i
agers were forced to hold their shares until the liquidation value P2,i is realized (which is,
by assumption, not subject to manipulation), then no manipulation would occur. In most
situations, however, such long-term incentive schemes are impractical. We therefore assume
that managers sell their stake in the firm at time 1.
2.2
Investors
Our economy is populated by M risk averse investors. For tractability, we assume that these
investors are short-lived. An agent born at time t = 0, 1 can buy and sell securities at
time t and has to unwind her position at time t + 1. Investors born in the first period do
not receive any information about the firms’ cash flow realizations before submitting their
orders. Investors who enter the market in the second period, on the other hand, can base
their asset demands on the firms’ earnings reports that are issued at time 1. However, they
do not observe the extent of earnings manipulation.
All investors are assumed to have mean-variance preferences with an identical risk aversion
coefficient of γ. They maximize their expected utility from end-of-period consumption given
by:
γ
E[U (Wt+1) | Ft] = E[Wt+1 | Ft] −
V ar[W
2
t+1 | Ft],
t = 0, 1,
(5)
7In Section 3.2, we will show that the equilibrium extent of manipulation κi(F ) never exceeds δi.
8

---

where Ft denotes the investors’ information set at time t and Wt+1 denotes their wealth at
time t + 1.
For simplicity, we assume that investors behave competitively. They take equilibrium
prices as given even though their aggregate trades affect market prices. Such behavior can
be justified by assuming that there is a large (to be precise, an infinite) number of investors,
so that no single trader can influence the price. This assumption seems quite reasonable, as
our analysis will, for the most part, focus on a large economy (in the sense that the number
of firms goes to infinity). In fact, this assumption is necessary to obtain a finite market risk
premium in equilibrium. As the number of firms in the economy increases without bound,
each of a finite number of investors would have to bear an infinite amount of risk and would
therefore demand an infinite risk premium. To rule out such an unrealistic scenario, we
require that the economy’s risk-bearing capacity increase at the same rate as the total risk
in the economy. More precisely, we assume that the number of investors expand at the same
rate as the number of firms so that, in the limit, M approaches a constant.8 Without loss of
N
generality, we assume that this constant is unity. In the ensuing analysis, we use the term
“large economy limit” to refer to the case where M and N approach infinity at the same rate.
3
Manipulation and Asset Prices in Equilibrium
In this section, we solve for the equilibrium of the economy defined above. We will focus on
the relationship between earnings manipulation and asset prices in the large economy limit.
The equilibrium concept we use is that of a Perfect Bayesian Equilibrium (PBE). Formally,
a PBE of our economy is defined by a manipulation strategy κ = (κ1, . . . , κN ), by the
market’s beliefs about κ, and by demand functions of time 0 and time 1 investors, such that:
(i) manager i’s manipulation strategy maximizes her expected utility, for all i = 1, . . . , N ; (ii)
for each price-taking investor, the trades specified by her demand function at a given date
8Similar restrictions have been adopted by Hughes, Liu, and Liu (2007) and Lambert, Leuz, and Verrecchia
(2007).
9

---

Document Outline

• Introduction
• The Model
  • Firms and Managers
  • Investors
• Manipulation and Asset Prices in Equilibrium
  • Investors' Portfolio Problem and Asset Prices
  • Optimal Earnings Manipulation
• Manipulation and the Cost of Capital
• Conclusion
• Appendix

---