Earnings Manipulation and Incentives in Firms

Earnings Manipulation and Incentives in Firms∗
Guido Friebel†and Sergei Guriev‡
First version: June 2004
This version: May 2005
Abstract
We study the effect of earnings manipulation on incentives within the corporate
hierarchy. When top management manipulates earnings, it must prevent informa-
tion leakage from corporate insiders to the outside world. If an insider (e.g. a
division manager) gains evidence about earnings manipulation, the threat to blow
the whistle can provide the division manager with an additional payment. We show
that it is easier for division managers to prove top management’s manipulations
when the performance of their own divisions is low. Earnings manipulation there-
fore undermines incentives of division managers and other insiders to exert effort
and destroys value. We show that earnings manipulation is more likely to occur in
flatter hierarchies; we also discuss implications of the auditing and whistle-blowing
regulations of the Sarbanes Oxley Act.
Keywords: agency costs, Sarbanes Oxley Act, whistleblowing, flat hierarchies.
JEL Classification Codes: D23, G30, M40, M52
∗We thank Patrick Bolton, James Brickley, Mike Burkart, Kong-Pin Chen, Mathias Dewatripont, De-
nis Gromb, Barry Ickes, Andy Leone, Fausto Panunzi, Paul Pover, Canice Prendergast, Michael Raith,
Patrick Rey, Andrei Shleifer, Jean Tirole, Dimitri Vayanos, Joanna Wu, and Luigi Zingales for helpful dis-
cussions. We also thank seminar participants at the ESSFM at Gerzensee 2004, CEFIR/NES (Moscow),
University of Rochester, London School of Economics, University of Frankfurt, Victoria University at
Wellington, Free University of Brussels, Stockholm School of Economics. All errors are ours.
†University of Toulouse (EHESS and IDEI), CEPR. E-mail: friebel@cict.fr
‡New Economic School, Moscow, and CEPR. E-mail: sguriev@nes.ru

---

1
Introduction
Recent corporate scandals have reinforced existing skepticisms about the efficiency of fi-
nancial markets and have triggered regulatory reforms like the Sarbanes-Oxley Act. While
these events spurred a number of research avenues in finance, law, and economics, our
understanding of the numerous and sizeable corporate misfortunes is far from complete.
In this paper, we offer a new perspective on earnings manipulation. We look inside the
corporate hierarchy, and investigate explicitly the intra-firm agency conflicts that are
associated with earnings manipulation. Our paper is motivated by a number of ques-
tions that are now increasingly discussed in economics, accounting and finance: Why did
the mechanisms of corporate governance fail to detect massive earnings manipulation by
management? What are the costs of earnings manipulation for society? What effect can
regulatory reforms have on avoiding earnings manipulation in the future?
The corporate governance literature has traditionally focused on external gatekeepers
such as auditors, non-executive board members, and financial institutions. We look at
the role of corporate insiders as internal gatekeepers. As they often have direct evidence
that executives engage in manipulation, they can play a crucial role. However, it appears
that this inside information rarely reaches the outside world. We do not investigate the
external costs of earnings manipulation, such as the transfer of utility from less informed
investors to more sophisticated players or the shattered confidence of shareholders in
financial markets. Rather, we identify an effect related to productive efficiency: earnings
manipulation undermines the functioning of companies, because it destroys incentives
within firms. Hence, it obstructs the very creation of value, rather than affecting its
distribution. Our perspective also provides some new insights about the role of regulations
like Sarbanes-Oxley, and it highlights the importance of organisational structure as a
determinant of earnings manipulation. In particular, we show that the recent tendency
to flatten hierarchies may have contributed to the widespread earnings manipulation.
A growing literature has shown that executive stock options and other short-term
incentives have played an important role in explaining the incidence of earnings manipu-
lation.1 In our theory, both top management and initial shareholders have an interest in
1 Peng and Roell (2004) provide an overview of empirical literature on the relationship between the
structure of managerial incentives and earnings manipulation. Bergstresser and Philippon (2002) and Gao
and Shrieves (2002) measure earnings manipulation through the value of discretionary current accruals
and find that it is related to stock holdings and options of executives. Interventions by the SEC are another
measure: Johnson et al. (2003) find that in 43 firms that committed fraud, stock-based compensation
and vested options were higher than in a matched control sample. Erickson et al. (2003) control for
endogeneity of stock-based compensation and find similar effects. Peng and Roell (2004) themselves use
1

---

overreporting short-term earnings, because they can sell stocks at inflated prices to un-
informed outside investors. We assume that the latter base their valuations of the stock
(at least to certain extent) on the accounting reports. This assumption is in line with
empirical evidence showing that even sophisticated investors may fail to fully account for
the possibility of earnings manipulation (for instance, Louis, 2004). Yet, it is important
to note that our results also hold in a game with rational investors and asymmetric in-
formation. In such a setting some bad firms pool with better firms and some investors
buy stocks based on reported earnings. This is, for instance, the case in Bebchuck and
Bar-Gill (2003) and Povel et al. (2003). Also, our results do not hinge on the original
shareholders’ incentive to sell to outsiders. Any other reason for short-term incentives
(discussed in the next section) will have similar effects in our model.
In the model, short-term incentives of top management affect the incentives inside
firms via two channels. First, as top managers have incentives to report high earnings
even if outcomes are low, they provide weaker incentives to their subordinates. The
second, more interesting, channel relates to the policies top management undertakes in
order to prevent the leakage of information to the outside world. A division manager may
have evidence that top management inflates earnings. Then, if she can prove that top
management tries to lie about earnings, the threat of blowing the whistle allows her to
claim a share in top management’s benefits of hiding the truth from the public. Our theory
thus explains why it is costly to maintain two accounting systems, one for the outside
world, potentially reporting inflated figures, and one as the basis of internal decisions and
incentives for insiders. Insiders with sensitive information would have a credible threat to
blow the whistle unless their compensation relates to the reported, rather than the true
earnings.
The important feature of the model is that it is easier for a division manager to prove
earnings manipulation by the top management if her own division has underperformed.
The knowledge about earnings manipulation and the threat of blowing the whistle can
provide division managers with an additional payment. Hence, there is a smaller difference
between the division’s payoffs associated with high vs low output. This distorts the
division managers’ effort choices.
Whistle-blowers do not have to be silenced ex post, at the bargaining table. Rather,
we show that top managers can neutralize the incentive to blow the whistle ex ante —
by providing lower level managers with short-term incentives. This may explain why
yet a different measure, namely, allegations of executive misbehavior in shareholder class actions. They
show that the incidence of such lawsuits are positively related to stock option components in executive
compensation.
2

---

in recent booms, stock options and other short-term incentives propagated in corporate
hierarchies. In the light of our theory, short-term incentives align insider incentives with
those of top managers and ensure that sensitive information does not reach the outside
world. Our theory hence contributes to a better understanding of why many firms provide
even lower-level employees with stock options in times of booms (see Oyer and Schaefer,
2005, and Jensen and Murphy, 2005).
From the point of view of initial shareholders, the trade-off is as follows. On the one
hand, if top managers manipulate earnings, stocks can be sold at inflated prices. On
the other hand, earnings manipulation weakens internal incentives and reduces firm value
in the long run. We show that in some parameter range, shareholders may encourage
short-termism and manipulation through the provision of short-term incentives. Then, in
equilibrium, division managers face suboptimal incentives and less value is created.
The problem we study — earnings manipulation and its effect on value creation — is
important, not only because the value of manipulations often amounts to several billions
of US dollars, but also because there are many more than a few isolated cases. Beyond
such household names as Enron, Tyco, WorldCom (in the US) or Ahold and Parmalat (in
Europe), there are many other, less known cases of earnings manipulation. The Forbes’
Corporate Scandal Sheet, for instance, lists more than 20 large corporate scandals (Pat-
suris, 2002). Xerox, for instance, inflated its earnings over five years. Its earnings ma-
nipulation reached $1.5 billion. HealthSouth Corp. overstated profits by $2.5 bn between
1997 and 2003; Waste Management manipulated earnings by $1.7 bn by overstating the
value of their trucks (Levitt and Dwyer, 2003).
To illustrate the results of our theory, consider the quintessential example of corporate
governance troubles: Enron, once the world’s seventh largest company, now bankrupt,
with several top managers facing legal charges. Earnings manipulation in Enron was so
widespread, that CFO Andy Fastow (backed by CEO/COO Jeff Skilling) perceived his job
being mostly, if not exclusively, about arranging “structured finance”: the use of aggressive
accounting to deliver high earnings quarterly reports (Maclean and Elkind, 2003, ch.
10).2 Enron’s ubiquitous book-cooking resulted in the inability to monitor projects even
internally. A deputy CEO once complained about Fastow’s aggressive accounting: “With
Fastow, you could never tell whether [individual] deals were clean because they were too
complicated” (Maclean and Elkind, 2003, p. 152). Healy and Palepu (2003) document
how widespread earnings manipulation induced Enron managers to take unfounded and
often excessively risky decisions. This may explain why despite very talented staff and
2 The sheer scale of Fastow’s activities is striking. Between 1997 and 2000 he created about 3000
corporate entities, including more than 800 of which were offshore (Maclean and Elkind, 2003, p. 310).
3

---

profitable core business, Enron ended up not only having its value below the reported
numbers but simply destroyed (Maclean and Elkind, 2003, Partnoy 2002, 2003).
Enron’s top management had substantial equity stakes or options (allegedly, the top
executives managed to cash stock options for $35-250 million each, Maclean and Elkind,
2003), and hence little incentives to communicate earnings truthfully. But, there were
a few attempts of blowing the whistle by those who were not enjoying the stock option
bonanza. The famous “smoking gun memo” by Vice President Sherron Watkins to the
Chairman Kenneth Lay opens with “Has Enron become a risky place to work? For those
of use who didn’t get rich over the last few years, can we afford to stay?” (Watkins, 2001,
our italics.) Even though it appears that Sherron Watkins had been aware of the scale of
Enron’s earning manipulation, she did not insist on disclosing to public; the memo never
reached outsiders before the collapse (Ackman, 2002).
Another (post Sarbanes Oxley) example of the threat of whistle-blowing is the case
of Matthew Whitley and Coca Cola. After being laid off, Whitley asked his former
employer for a settlement payment of $44.4 million. He argued he had been fired in
retaliation for raising concerns about accounting fraud (CFO Magazine, 2003). When
Coca Cola dismissed the settlement proposal, the case became public. Coca Cola and
other companies are now facing charges by the SEC (CFO Magazine, 2004a); Whitley
and Coca Cola have settled in the meantime. Many firms seem to be quite vulnerable
against whistle-blowing. New York’s attorney general, Eliot Spitzer, built his case against
Wall Street firms doctoring their reports on (e-mail) information from insiders. (Time
Magazine, 2002).
Whistle-blowing nonetheless seems a rather rare phenomenon given the size and range
of earnings manipulation and other corporate fraud, but this does not speak against
our theory. Whistle-blowing may occur seldom in equilibrium, because there are many
ways to avoid it. Top management can settle ex post with employees that have sensitive
information, or it may, ex ante, give informed insiders stock-price-based incentives. This
provides agents at different levels in the hierarchy with similar incentives to be silent.
The paper proceeds as follows. In Section 2, we discuss related literature. In Section 3,
we set up a simple model. In Section 4 we establish the main result. In Section 5 we discuss
extensions that allow to investigate the role of organisational structure and monitoring
intensity. In particular, we argue that at given size, it is harder for firm insiders to provide
evidence of earnings manipulation in flatter hierarchies. Section 6 discusses the Sarbanes-
Oxley Act in light of our theory. We argue that the Act will reduce the frequency of
earnings manipulation, because it decreases insiders’ costs of whistle-blowing and induces
auditors to monitor more carefully. This reduces top management’s benefit from earnings
4

---

manipulation as they have to share with a larger number of firm insiders and auditors.
Section 7 concludes.
2
Related literature
There is a growing literature that investigates the effects of short-termism and earnings
manipulation on managerial incentives. In Stein (1989), both short-termism and earnings
manipulation emerge in a non-cooperative equilibrium between managers and rational
investors. Bebchuk and Stole (1993) show ambiguous effects of short-term incentives on
managerial effort depending on the structure of asymmetric information. Bolton et al
(2004) show that in the presence of overconfident shareholders, managers may choose
sub-optimal projects with higher variance. Jensen (2004) argues that overvalued equity
aggravates the agency problems between investors and managers.
While these papers point to the agency relationship between the outside world and
corporate management, we explicitly analyze the effect of earnings manipulation on the
agency relationship inside firms. Our model of a corporate hierarchy highlights the poten-
tial role of corporate insiders as gatekeepers; it shows how distortions propagate through-
out the hierarchy and how earnings manipulation obstructs value creation by undermining
internal incentives. It also highlights the importance of organisational structure for the
feasibility of earnings manipulation.
Jensen and Murphy (2005) argue that the main reason why CEOs were granted ex-
cessive compensation based on stock market prices was that shareholders did not fully
account for the costs of such incentives. Our, formal, theory is to a large extent comple-
mentary, except that in our theory initial shareholders are fully rational. However, we
could readily reinterpret our model in a way that the shareholders who design (directly
or indirectly) the compensation of CEOs may make mistakes. Then, the problem would
be further amplified, but the value of insider information would be quite similar. If an
inside manager realizes that something goes wrong in the company, the CEO would still
not want them to reveal it, because CEO compensation depends on stock market price.
In our theory, top management has short-term incentives to manipulate earnings be-
cause outside investors value the company on the basis of accounting reports. There are at
least two explanations why full unraveling of earnings manipulation does not occur. First,
in the presence of short-sale constraints and some overconfident investors, the stock price
will to some extent reflect the company’s valuation by investors that rely on the most
optimistic signals (Harrison and Kreps, 1978, Scheinkman and Xiong, 2003, Bolton et al,
2003). The other prominent explanation is based on rational but uninformed investors
5

---

(Bebchuk and Bar-Gill, 2003, Povel et al., 2003). The equilibrium in an asymmetric
information game may involve partial pooling and therefore rewards for companies over-
reporting earnings. If overreporting is costly, investors know that the worst companies
will not report the highest possible outcome. This is why they do value positive reports
even though they know that the pool of companies with good reports include both good
and intermediate firms. For the sake of simplicity, we choose a setting with overconfident
investors. The asymmetric information framework would produce the same results; we
provide a sketch of such a model in the Appendix.
Managerial short-termism can also be driven by the costs of long-term arbitrage
(Shleifer and Vishny, 1990), and managerial risk aversion and demand for liquidity (Holm-
strom and Tirole 1998, Aghion et al, 2004, Axelson and Baliga, 2004). Whatever the
source of CEO short-termism, our theory would still imply weaker incentives for his sub-
ordinates.
To the extent that we look at rent-seeking and incentives inside a firm, our model has
some similarity to the model in Scharfstein and Stein (2000). However, in our model,
the potential rents for division managers are created by CEO’s earnings manipulation.
Moreover, we allow for cash compensations inside the firm; our main interest is information
diffusion to the outside world rather than allocation of capital between divisions. Also
related is the paper by Faure-Grimaud and Gromb (2004) who study the role of large
shareholders as a different type of insiders in providing information about a firm.
Earnings misreporting may also be carried out for tax optimization purposes (Misai,
2003). In this case, managers tend to underreport rather to overreport the earnings.
Desai and Dharmapala (2004) build a model of interaction between managers, owners
and government. The effect of high-powered incentives on tax avoidance is ambiguous;
tax sheltering is complementary to diversion of profits from shareholders. On the other
hand, by paying less taxes, manager may increase shareholder value. Hence, the effect
of incentives on reported earnings depends on corporate governance, namely to what
extent management is controlled by shareholders. Desai and Dharmapala do not model
corporate hierarchy, but our analysis suggests that tax avoidance need not distort internal
incentives. Indeed, as management tends to underreport earnings, keeping two sets of
books is incentive compatible. The potential whistle-blowers are the successful rather
than failing divisions; these divisions get a bonus on top of their official compensations
and are happy not to report the fraud.
This is similar to the result in Chen and Chiu’s (2005, Proposition 1). They look at
a principal-agent model of tax evasion, in which having two sets of books is feasible. In
their model, managerial effort can be distorted in both directions: relative to the second-
6

---

best, there is too much (too little) effort, when the manager has decreasing (increasing)
absolute risk-aversion. In our model, however, effort is always distorted downwards,
because insiders learn more about the true state of the firm (or have higher incentives to
learn about it) when they themselves have underperformed.
Our paper contributes to literature on whistle-blowing (Miceli and Near, 1992) that
discusses costs and benefits of blowing the whistle. This literature has been largely de-
scriptive except for Leppamaki (1998) who studies whistle-blowing in a rather specific
context of bilateral monitoring and for Friebel and Raith (2004) who look at whistle-
blowing and conflicts between different layers in the hierarchy. Our paper complements
this literature by modelling the effect of potential whistle-blowing on corporate gover-
nance. The prediction that potential leakage of inside information is prevented by profit
sharing is not new in the literature on innovation and unpatentable knowledge (see Anton
and Yao, 1994, Baccara and Razin, 2002, Bhattacharya and Guriev, 2004). Yet, it has
not been explicitly modelled in the framework of corporate governance. While testing the
alternative explanations of provision of stock options to employees, Oyer and Schaefer
(2005) reject the argument that stock options are given to provide incentives to work
harder. They do, however, not discuss stock options as incentives to prevent leakage of
sensitive information.
Finally, our paper is related to the problem of leniency programmes as deterrence
device for cartels and other criminal associations, see for instance Spagnolo (2000). How-
ever, our whistle-blower problem is more common as it also concerns legal organisations,
and it looks at a complex interplay between efficiency and whistle-blowing which is not
present in criminal organisations.
3
The model
We consider a publicly traded firm that is held by initial shareholders (“S ”). The firm
is run by a CEO (“M ”) who reports to S, and there are two division managers, A and
B, who report to M . There are also potential new investors (“I ”). Agents M , A and
B are risk-neutral, but have limited liability; their reservation payoffs are normalized
to zero. Both initial investors S and overconfident potential investors I are financially
unconstrained and, in dealing with M , risk-neutral.
7

---

3.1
Production and information
The division managers exert effort ei, i = A, B, which increases the expected value of the
firm. Output of each division yi can take two values: 1 (with probability ei) and 0 (with
probability 1 − ei). The aggregate output of the divisions is
y = yA + yB.
Effort is private information of the respective division manager. The cost of effort is c(ei),
which is an increasing, convex function. For simplicity, we assume that c(e) ∼ e1+σ, so that
the elasticity of the marginal cost of effort σ = ec00(e)/c0(e) is positive and independent of
the level of effort. Division manager i observes her own output, but not that of the other
division manager. In what follows, we assume that division managers cannot engage in
any side-contracting and, in particular, that they cannot share information.3
The CEO observes both divisions’ outputs, while the outside world — S and new
investors I, who may buy stocks from S — cannot observe the true value of the firm.
Rather, they have to rely on M ’s report, which may or may not be truthful. This setting
captures the idea that certain information is only available within the boundaries of a
firm. The role of auditors, who at some cost can generate additional information about
the true value of the firm, is discussed in Section 6.
Besides gathering and reporting information, M also exerts effort E = {0,1} which
is complementary to that of division managers. For simplicity we make an extreme as-
sumption of perfect complementarity: the firm’s gross output is yE. If M chooses high
effort, E = 1, the output of divisions adds value y. If the top manager shirks, the firm’s
value is nil. The top manager’s cost of exerting E = 1 is cM . We shall assume throughout
the paper that this cost of effort is sufficiently low and that shareholders will provide the
top manager with incentives (short-term, long-term, or both) such that M never shirks.
Thus we will solve the model for the case E = 1, and then check for M ’s incentive not to
deviate to E = 0.
3.2
Timing
There are three dates: t = 0 (contracting stage), t = 1 (short run), and t = 2 (long run).
At t = 0, S hold 100% shares. They keep at least αS2 ≥ 0 shares in the long run (e.g. for
risk management purposes, or for preserving non-pecuniary benefits of control). In our
model, αS2 is an exogenous parameter. The initial shareholders offer M a contract; then
M offers a contract to A and B. The compensation package of M includes a fixed salary
3 We discuss this issue further in 4.7.
8

---

and bonuses that are contingent on earnings reported at t = 1 and t = 2. These bonuses
can be interpreted as stock grants or stock options to be exercised at t = 1 (short term)
and t = 2 (long term). For simplicity, we assume that the manager receives αM1 shares
that can be sold in the short term (to new investors I ), and αM2 shares that must be
kept until the end; β is the present value of all fixed salaries M receives. The division
managers’ contracts are contingent on their individual output yA,B. We assume that either
M or a division manager can verify the division’s output before a court if they wish to.
Date t = 1 has several stages:
1. The CEO chooses E.
2. Division managers A and B choose their effort levels eA, eB and division outputs
realize.
3. The CEO observes yA and yB, and prepares a report about aggregate earnings. The
reported earnings x may or may not be manipulated by M . Given W, the sum of
all wage payments to M , A and B, the reported short-term value of the firm is
V1 = x − W.
Here x = y if there is no overreporting, and x > y otherwise. Obviously, W depends
on whether or not there is earnings manipulation. We assume that inflating short-
term earnings by x − y units has costs Cx−y with C2 > C1 > 0 = C0. These costs
reduce the firm’s long-run V
4
2, but do not affect short-run value V1.
4. The division managers learn about the report before it is sent to potential investors.5
Subsequently, if they believe that there is some overreporting, they can bargain
individually with M for a wage raise. The bargaining power of either division
manager (vis-a-vis the CEO) is γ ∈ [0,1]. If M and one or several of the division
managers disagree, they can make available to the public verifiable information
about their division output. This whistle-blowing drives down the stock price that
the potential investors are willing to pay. We further specify the whistle-blowing
game in Section 4.
4 Two comments: First, this may be considered as a reduced form of the net present value of stochastic
costs of earnings manipulation. Second, the costs may be incurred by the firm, for instance, as consulting
fees to auditors or misallocation of resources, or by the manager, for instance, as an effort cost or the
expected costs of potential imprisonment. We take the former route, as it is more in line with the
literature, e.g. Stein (1989).
5 We could also allow for A and B to receive noisy signals about the report.
9

---