Ch. 2: Self-Selection Models in Corporate Finance 59
II. EMPIRICAL APPLICATIONSThis part reviews empirical applications of self-selection models in corporate finance.
We limit our scope to papers in which self-selection is an important element of the
econometric approach or substantive findings. We begin with applications in event-
studies. Here, the specifications are related to but differ from standard selection models.
We then review applications in security offerings and financial intermediation, where
more conventional selection models are used to characterize how private information
affects debt issue pricing. We then turn to the diversification discount literature, where
a range of methods have been used to address self-selection issues. The remaining sec-
tions include a collection of empirical applications based on selection and propensity
score based matching methods. A last section covers Bayesian techniques. As will be
clear from the review, most applications are relatively recent and involve a reasonably
broad spectrum of approaches. In most cases, the model estimates suggest that unob-
served private information is an important determinant of corporate finance choices.
- Event studies
Event studies are a staple of empirical corporate finance. Hundreds of studies routinely
report the stock market reactions to announcements such as mergers, stock splits, div-
idend announcements, equity issues, etc. Evidence in these studies has been used as a
basis for testing and generating a wealth of theories, policies, and regulations.Chapter 1
in this volume (Kothari and Warner, 2007) overviews the literature.
Self-selection entered the event-study literature relatively recently. Its main use has
been as a tool to model private information revealed in events. The basic idea is that
when firms announce events, they reveal some latent “private” information. If the private
information has value, it should explain the announcement effects associated with an
event. Selection models are convenient tools to model the information revelation process
and estimate “conditional” announcement effects.
7.1. Conditional announcement effects:Acharya (1988)
Acharya (1988)introduces the self-selection theme to event-studies, using a version of
the standard Heckman specification to model calls of convertible bonds. In Acharya’s
model, firms decide whether to call an outstanding convertible bond (eventE)ornot
(eventNE) according to a probit model, viz.,
E ifWi=Ziγ+ηi> 0 , (31)
NE ifWi=Ziγ+ηi 0 , (32)whereZdenotes known observables andη, the probit error term, is private information.
Ex-ante, private information has zero mean, but ex-post, once the firm has announced