Ch. 2: Self-Selection Models in Corporate Finance 81
capital he has available. The characteristic of the market is the year of the investment.
There are two parameters of central interest. One is the access of better venture capi-
talists to deal flow, which is captured by the experience of the venture capitalist. The
other is the synergy between venture capitalists and their target investments or the value
added by venture capitalists, which is captured by the correlation between the private
information in the decision to invest and the probability of going public.
Sørensen’s final sample includes 1,666 investments made by 75 venture capitalists
between 1975 and 1995 in the states of California and Massachusetts. Experience is
proxied by the total and stage-of-life-cycle-specific number of deals done since 1975.
Sørensen reports a number of interesting findings. He finds evidence for sorting. Expe-
rienced investors are more likely to have access to the better deals whose probability of
going public (and doing so faster) increases by about two-thirds. This type of sorting
explains about 60% of the increased probability of success, leaving about 40% for the
synergies, or the value added by venture capital investors. Sørensen explains why one
might get different results from estimating a standard selection model compared to one
with sorting.
13.2. Switching regressions:Li and McNally (2004),Scruggs (2006)
Li and McNally (2004)andScruggs (2006)offer interesting applications of Bayesian
methods to estimate switching regression models of self-selection. Both papers empha-
size that the value of the Bayesian approach is not merely the difference in philosophy
or technique; rather, the techniques offer insights not readily available through classical
methods. The application inLi and McNally (2004)is the choice of a mechanism to ef-
fect share repurchases, while the application in Scruggs relates to whether convertibles
are called with or without standby underwriting arrangements. For convenience, we fo-
cus on Li and McNally, but substantially similar insights on methodology are offered in
the work by Scruggs.^31
Share repurchases started becoming popular in the 1980s as a way to return excess
cash to shareholders in lieu of dividends. Repurchases tend to be more flexible in timing
and quantity relative to the fixed cash flow stream expected by markets when companies
raise dividends. Share repurchases can be implemented in practice as a direct tender
offer or more open-ended open market repurchases.Li and McNally (2004)investigate
the choice between the two mechanisms and their impact on share price reactions to
announcements of repurchases using Bayesian self-selection methods.
Li and McNally propose the following system of equations to analyze the choice of
a repurchase mechanism
I∗=Ziγ+ηi, (69)(^31) Wald and Long (2006)present an application of switching regression using classical estimation methods.
They analyze the effect of state laws on capital structure.