172 M. Baker et al.
model is itself hard to distinguish from models of costly external finance built on asym-
metric information. Thus, to test the behavioral theories, one must separate theγrelated
to overconfidence and optimism from theγthat arises from agency or asymmetric in-
formation problems.
3.3. Investment policy
Despite the obvious difficulty of obtaining direct, manager-level measures of optimism
and overconfidence, evidence is accumulating that these biases do affect business in-
vestment.
3.3.1. Real investment
We begin with startup investments. The evidence indicates that entrepreneurial startups
are generally made under a halo of overconfidence and optimism.Cooper, Woo, and
Dunkelberg (1988)find that 68% of entrepreneurs think that their startup is more likely
to succeed than comparable enterprises, while only 5% believe that their odds are worse,
and a third of entrepreneurs view their success as essentially guaranteed. The survey
responses of French entrepreneurs tabulated inLandier and Thesmar (2005)also seem
consistent with an initial underestimation of the task of starting a firm: at startup, 56%
expect “development” in the near future, and 6% expect “difficulties”.
The actual performance of startup investments is more sobering. Landier and Thes-
mar find that when surveyed three years into their endeavor, only 38% of French
entrepreneurs expect further “development” while 17% anticipate “difficulty”. Leav-
ing profitability aside entirely, only half of all startups survive more than three years
(Scarpetta et al., 2002). Moskowitz and Vissing-Jorgensen (2002)argue more generally
that the return on private equity in the US between 1952 and 1999 is lower than seems
justified given the undiversified nature of entrepreneurial investment. As a whole, the
evidence on startup investments seems consistent with the overconfidence thatCamerer
and Lovallo’s (1999)experimental subjects display when making entry decisions.
Optimism also appears to influence investment in more mature firms.Merrow,
Phillips, and Myers (1981)compare forecast and actual construction costs for pioneer
process plants in the energy industry. There is a strong optimism bias in project cost
forecasts, with actual costs typically more than double the initial estimates.Statman
and Tyebjee (1985)survey several other studies of this sort, involving military hard-
ware, drugs, chemicals, and other development projects, and conclude that optimistic
biases in cost and sales forecasts are fairly widespread.
Malmendier and Tate (2005)provide cross-sectional tests of the effects of optimism
in a broader sample. They form a clever manager-level proxy for optimism: the propen-
sity for a manager to voluntarily hold in-the-money stock options in his own firm. The
intuition is that since the CEO’s human capital is already so exposed to firm-specific
risk, voluntarily holding in-the-money options can be seen as a strong vote of opti-