THE JOURNAL OF FINANCE
•
VOL. LXVI, NO. 5
•
OCTOBER 2011
Overconfidence, Compensation Contracts,
and Capital Budgeting
SIMON GERVAIS, J. B. HEATON, and TERRANCE ODEAN
∗
ABSTRACT
A risk-averse manager’s overconfidence makes him less conservative. As a result, it
is cheaper for firms to motivate him to pursue valuable risky projects. When com-
pensation endogenously adjusts to reflect outside opportunities, moderate levels of
overconfidence lead firms to offer the manager flatter compensation contracts that
make him better off. Overconfident managers are also more attractive to firms than
their rational counterparts because overconfidence commits them to exert effort to
learn about projects. Still, too much overconfidence is detrimental to the manager
since it leads him to accept highly convex compensation contracts that expose him to
excessive risk.
AVAST EXPERIMENTAL LITERATURE finds that individuals are usually overconfi-
dent in that they believe their knowledge to be more precise than it actually
is. The incidence of overconfidence is likely to be even greater among CEOs
than among individuals at large; for example, Goel and Thakor (2008) show
that overconfident individuals are more likely to win the intrafirm tourna-
ments that lead to the rank of CEO. Since overconfidence directly influences
decision-making, it is logical to investigate the effects that overconfident man-
agers have on corporate policies and firm value. How does overconfidence affect
the investment decisions that managers make on behalf of shareholders? How
do compensation contracts optimally adjust to these effects? Do firms benefit
from managerial overconfidence? Can overconfidence ever benefit the biased
∗
Simon Gervais is from Duke University. J. B. Heaton is from Bartlit Beck Herman Palenchar
& Scott LLP. Terrance Odean is from the University of California at Berkeley. This paper is an
updated version of a previous working paper, “Overconfidence, Investment Policy, and Manager
Welfare,” by the same authors. The authors would like to thank Franklin Allen, Jonathan Berk,
Bruce Carlin, David Denis, Robert Goldstein, David Ross, Jacob Sagi, Marti Subrahmanyam, Peter
Swan, two anonymous referees, and the Acting Editor, David Hirshleifer, for their comments and
suggestions. Also providing useful comments and suggestions were seminar participants at the
Meetings of the European Finance Association, the Meetings of the American Finance Association,
the NYU Stern Five Star Conference on Research in Finance, the Laurier Finance Conference, the
Conference of the Financial Intermediation Research Society, Baruch College, Cornell University,
Emory University, Insead, London Business School, London School of Economics, the University of
Amsterdam, the University of British Columbia, the University of California at Los Angeles, the
University of Florida, the University of Maryland, the University of Minnesota, the University of
Oregon, the University of Toronto, Vanderbilt University, and the Wharton School. J. B. Heaton
acknowledges that the opinions expressed here are his own, and do not reflect the position of
Bartlit Beck Herman Palenchar & Scott LLP or its attorneys. All remaining errors are the authors’
responsibility.
1735
1736 The Journal of Finance
R
manager himself? How does labor market structure influence the effects of
overconfidence?
We develop a model of capital budgeting that enables us to answer these
questions by contrasting the decisions of overconfident managers with those
of rational managers. In our model, a risk-neutral firm hires a risk-averse
manager to make an investment decision on its behalf. The manager is skilled
in that he has private access to a noisy signal about the quality of the new
project before he decides whether to invest in it. An overconfident manager
overestimates the precision of his signal, and so is overly inclined to undertake
(abandon) the project when his information is positive (negative). In a context
where the manager is both risk-averse and overconfident, the compensation
contract that maximizes shareholder wealth serves two purposes: to increase
the manager’s appetite for undertaking risky projects, and to curb the impetu-
ousness that his overconfidence creates. We analyze this contracting problem
first in a setting where the manager’s opportunities outside the firm are ex-
ogenously fixed, and then in a labor market equilibrium that endogenizes the
manager’s reservation utility.
When the manager’s outside opportunities are fixed, the firm captures the
entire economic surplus that the manager’s skill creates. We show that this sur-
plus is increasing in the manager’s overconfidence. The inability to diversify
their human capital causes risk-averse managers to act more conservatively
than is in the best interests of shareholders (e.g., Jensen and Meckling (1976),
Treynor and Black (1976)). Because overconfident managers overestimate the
amount of risk that their information eliminates, they are less prone to this
conservatism. The form of compensation contract that is offered by the firm de-
pends on the degree of the manager’s overconfidence. If the manager is mildly
overconfident, the firm increases its value by reducing the convexity of his
compensation contract relative to a rational manager’s contract. In this case,
less performance-based compensation (e.g., bonuses, stocks, and options) is re-
quired to realign the manager’s incentives to undertake valuable risky projects
because the overconfident manager perceives less risk. If, instead, the manager
is extremely overconfident, the firm takes advantage of his bias by offering him
a highly convex contract that he overvalues. The performance-based compen-
sation then further benefits the firm by allowing it to arbitrage the manager’s
excessive beliefs that his decisions will lead to good outcomes.
When a competitive labor market is introduced into the model, firms must
compete to hire managers who then capture some of the economic surplus they
create. In this context, we show that an overconfident manager may benefit
from his own bias; that is, he may end up capturing more economic surplus
than an otherwise identical but rational manager does. The realization of this
benefit depends upon the degree of the manager’s overconfidence and the set
of firms that seek to hire him. When similar firms compete for the services
of a mildly overconfident manager, they rely on the manager’s overconfidence
to guarantee his commitment to undertake valuable risky projects. A mod-
est amount of performance-based compensation is then sufficient to realign
the manager’s incentives, and the firms end up competing to attract him by
Overconfidence, Compensation Contracts, and Capital Budgeting 1737
increasing the safer portion of his compensation. In equilibrium, the overcon-
fident manager’s compensation is closer to first-best, and he is better off than
his rational counterpart. Specifically, the manager’s overconfidence allows for a
more efficient transfer of economic surplus via a flatter schedule that improves
the risk-sharing arrangement between the manager and the winning bidder.
If the manager’s overconfidence is extreme, however, it becomes optimal for
the firms to compete for his services by increasing his performance-based com-
pensation. In this case, the manager overvalues this type of compensation to
such an extent that firms gain from shifting risk onto the manager. This al-
ways makes the manager worse off. These results complement the work of Goel
and Thakor (2008), who show that moderate levels of manager overconfidence
benefit the firm, while extreme levels of overconfidence are detrimental. Our
analysis shows that the presence of labor markets leads to similar results about
the welfare of managers.
When the firms competing for the manager’s services differ in their growth
prospects and risk, the eventual matching of managers with firms depends on
the manager’s overconfidence. Rational and mildly overconfident managers are
more likely to end up working at safe, diversified, value firms that offer rela-
tively flat compensation contracts. In contrast, highly overconfident managers
are likely to be attracted by the compensation convexity that risky, focused,
growth firms can offer. Our model thus predicts that changes in the cross-
sectional composition of labor and compensation contracts vary with changes
in firms’ investment opportunity sets (e.g., resulting from the emergence of
a new growth industry), changes in individual overconfidence (e.g., resulting
from self-attribution bias), and the extent of competition for a given skill (e.g.,
individuals with industry-specific skills do not attract bidders from other in-
dustries to the extent that individuals with general, portable skills do).
Finally, we show that managerial overconfidence can serve as a commitment
to exert costly effort. In this final extension of the model, compensation con-
tracts must serve a dual purpose, as in the work of Lambert (1986), Hirshleifer
and Suh (1992), and Diamond (1998): to realign the manager’s incentives to
make investment decisions in the shareholders’ interest, and to ensure that the
manager exerts the effort necessary to investigate investment opportunities.
Our analysis departs from earlier studies by adding managerial overconfidence
to this dual agency problem. Because overconfident managers overvalue the
benefit of the effort needed to learn about risky projects, they can be motivated
to exert this effort more easily than rational managers. In some cases, dual
realignment is only possible with overconfident managers and so hiring these
managers, as opposed to otherwise identical but rational managers, becomes
a priority for firms. This again benefits managers who exhibit some overconfi-
dence.
Although a large body of theoretical literature studies the implications of
overconfidence for financial markets,
1
relatively few theoretical studies look at
1
See Hirshleifer (2001) for a survey of this literature.
1738 The Journal of Finance
R
overconfidence in corporate settings.
2
Roll (1986) suggests that overconfidence
(hubris) may motivate many corporate takeovers. His conjecture that success-
ful investments require CEO skill and some risk-taking is consistent with our
model. As we show, however, the conclusion that overconfidence leads to over-
investment implicitly relies on suboptimal contractual arrangements between
the firm and its decision-makers. Our analysis departs from the existing litera-
ture by assuming that firms can identify overconfident managers and therefore
adjust their contracts to properly account for their biases. In this context, the
incidence of overinvestment need not be affected by overconfidence, and more
economic surplus can sometimes be shared by the firm and the agent.
3
Building on the work of Fershtman and Judd (1987) and Sklivas (1987)
about the optimal incentives of agents whose firms compete in product markets,
Englmaier (2006) shows that overconfidence commits agents to be more aggres-
sive, making their firm more profitable in the process.
4
Hackbarth (2008) finds
that managerial overconfidence leads to greater debt financing and that over-
confidence, by acting as a commitment device, can also ameliorate bondholder
and shareholder conflicts such as debt overhang. Our paper differs from these
in that we consider the interaction of the manager’s bias with the compensation
contract and incentives provided by his firm.
The work of Adrian and Westerfield (2009) and Palomino and Sadrieh (2011)
is also related to this paper. Adrian and Westerfield (2009) develop a model
of dynamic contracting with disagreement and learning, and show that there
are gains from shifting the agent’s consumption—via contracting—to states
the agent considers more likely. Palomino and Sadrieh (2011) show how the
principal can benefit from the agent’s overconfidence in a delegated portfolio
management setting if he knows that the agent is overconfident. Like these
papers, we also show that the principal can benefit from the agent’s overcon-
fidence via contracting when he is aware of the agent’s overconfidence. The
difference is that we focus our analysis on capital budgeting issues and exam-
ine the labor market conditions under which the agent also benefits from his
overconfidence.
Most closely related to our paper is the work of Goel and Thakor (2008).
They model the firm’s internal promotion process as an intrafirm tournament
and show that overconfident managers are more likely than rational managers
2
Our notion of overconfidence captures the idea that individuals overestimate the precision of
their information or their ability to interpret that information when they make economic decisions.
A related concept is the idea that people are optimistic in the sense that they expect future
outcomes to be better than they really are. Such a bias has been studied in the context of financial
intermediation (Manove and Padilla (1999), Coval and Thakor (2005)), entrepreneurship (Landier
and Thesmar (2009)), and capital budgeting (Heaton (2002)), among others. For surveys of the
effects of behavioral biases in corporate finance, see Baker, Ruback, and Wurgler (2007) and
Gervais (2010).
3
This possibility is in fact documented by Graham, Li, and Qiu (2012), who find that manager
fixed effects explain over half of the variation in executive compensation. They also show that
these fixed effects relate to unobservable manager characteristics, which may include skill, risk
aversion, and overconfidence.
4
Kyle and Wang (1997) make a similar point in the context of money management.
Overconfidence, Compensation Contracts, and Capital Budgeting 1739
to be promoted to CEO. One of their main results, namely, that increases in
managerial overconfidence lead to higher firm value but only up to a point (after
which excessive overconfidence destroys firm value), is also encountered in our
analysis. The key differences are threefold. First, in contrast to that paper, we
assume that firms know they are dealing with overconfident managers and thus
write incentive contracts accordingly, that is, we focus on the design of optimal
contracts when the principal can distinguish among rational and overconfident
agents. Second, we show that the result that an overconfident manager is made
worse off by his bias, a finding common to most papers in the literature, is a
special case. Specifically, this result occurs only when the structure of the labor
market is such that all the surplus created by the manager’s overconfidence
goes to the firm. In the more general case in which the surplus is shared,
both the firm and the manager can benefit from managerial overconfidence.
Third, unlike the previous literature, we also examine how CEOs are matched
with firms based on both managerial overconfidence and firm attributes. For
example, our model shows that the most overconfident executives will tend to
end up in risky growth firms, as documented by Graham, Harvey, and Puri
(2009).
5
Recent empirical studies also document the presence of managerial overcon-
fidence and its effects on corporate policies. For example, Malmendier and Tate
(2005, 2008) use the tendency of CEOs to delay the exercise of their stock op-
tions to proxy for overconfidence, and show that this measure correlates with
the intensity of firm investments. Ben-David, Graham, and Harvey (2010) and
Sautner and Weber (2009) use survey evidence to show that the overconfidence
of top executives affects various corporate decisions, including the investment
policy of the firm. Liu and Taffler (2008) use formal content analysis of CEO
statements to measure CEO overconfidence, and find that high ratings of this
measure correlate with investment activity. As we show, overconfidence cou-
pled with inefficient contracting can lead to overinvestment. However, when
contracts optimally consider overconfidence, our model makes a number of new
empirical predictions that are readily testable using the same data as in these
existing studies. For example, our model predicts that overconfident execu-
tives should receive more performance-based compensation than their rational
counterparts in young, risky, growth firms and less performance-based com-
pensation in older, safer, value firms.
The rest of our paper proceeds as follows. Section I introduces the model,
presents the first-best solution, and solves the firm’s problem of choosing a
value-maximizing compensation contract. In Section II, we introduce labor
market considerations and show how overconfidence can end up benefiting the
manager. Section III provides an extension of our model that accommodates
moral hazard resulting from costly effort. Finally, Section IV summarizes our
findings and concludes. All the proofs are contained in Appendix A.
5
Our work also complements that of Bernardo and Welch (2001), who provide an evolutionary
rationale for the presence of overconfident entrepreneurs in a society, and that of Wang (2001),
who shows how overconfident traders can survive in the long run.
