Ch. 2: Self-Selection Models in Corporate Finance 77
the bankruptcy court; and the recovery rates of creditors. These are modeled as a func-
tion of a comprehensive set of regressors that include linear and non-linear functions
of firm size, various proxies for the structure of the filing firm and managerial owner-
ship. Because the variables in the two stages are similar, the study essentially relies on
non-linearity for identification.
Bris et al. find no evidence that firms that were more likely to self-select into Chapter
11 were any faster or slower in completing the bankruptcy process. Controlling for self-
selection, Chapter 11 cases consumed more fees, not because Chapter 11 is intrinsically
the more expensive procedure, but because of intrinsic differences in firms that choose
to reorganize under this code. After controlling for self-selection, Chapter 11 emerges
as the cheaper mechanism, and Bris et al. report that self-selection explains about half of
the variation in bankruptcy expenses. With self-selection controls, Chapter 11 cases had
higher recovery rates than Chapter 7 cases. In sum, selection has a significant impact
on estimates of reorganization costs under different bankruptcy codes. After controlling
for selection, Chapter 7 takes almost as long, consumes no less and probably more in
professional fees, and creditors rarely receive as much, so there is little evidence that it
is more efficient than Chapter 11 reorganizations.
11.3. Family ownership and value:Villalonga and Amit (2006)
Villalonga and Amit (2006)examine the effect of family ownership, control, and man-
agement on value for a sample of Fortune 500 firms from 1994 to 2000. The specifica-
tion is a standard Heckman style selection model of Section2 with a treatment effect.
The first step is a probit specification that models whether a firm remains family
owned or not. Family ownership is defined as firms in which the founding family owns
at least 5% of shares or holds the CEO position. In the second step, value, proxied by
Tobin’sQ, is regressed on a dummy variable for family ownership, industry dummy
variables, theGompers, Ishii and Metrick (2003)shareholder rights index, firm-specific
variables from COMPUSTAT, outside block ownership and proportion of non-family
outside directors, and, of course, the inverse Mills ratio that corrects for self-selection.
To assist in identification, Villalonga and Amit include two additional instruments in
the selection equation laggedQand idiosyncratic risk. Idiosyncratic risk is presumably
related to family ownership but not toQif only systematic risk is priced by the market.
Villalonga and Amit report that family ownership has a positive effect on value in
the overall sample and in sub-samples in which the founder is the CEO. Interestingly,
the sign is negative when the founder is not the CEO. Villalonga and Amit interpret
their findings as evidence that the benefits of family ownership are lost when the family
retains control in the post-founder generation. Their results strengthen when they incor-
porate a control for self-selection. In the self-selection specification, the inverse Mills
ratio is significant and negative in the overall specification and sub-samples in which the
CEO is the founder. In these samples, family ownership appears to be associated with