74 K. Li and N.R. Prabhala
work does make an important point. Specifically, the statistical significance of discount
based on industry/size matching methods is not a given fact, but is an open question in
light of her results.
10.3. Refocusing and the discount:Çolak and Whited (2005)
If one accepts the diversification discount as a fact, then the question is what causes
the discount. One view is that conglomerates (i.e., diversified firms) follow inefficient
investment policies, subsidizing inefficient divisions with cash flow from the efficient
divisions.Çolak and Whited (2005)evaluate the efficiency of investment in conglom-
erate and non-conglomerate firms by comparing investments made by focusing firms
with those made by firms that do not focus. The focusing sample inÇolak and Whited
(2005)consists of 267 divestitures and 154 spinoffs between 1981 and 1996. Control
non-focusing firms are multi-segment firms in similar businesses that do not focus in
years−3 through+3 where year 0 is the focusing event for a sample point.
The main specification used inÇolak and Whited (2005)employs propensity scores
to match focusing and non-focusing firms. As in standard propensity score method im-
plementations,Çolak and Whited (2005)estimate the propensity score as the probability
that a given firm will focus in the period ahead. The probit estimates broadly indicate
that firms are more likely to focus if they are larger, have less debt, diversity in segments
(entropy), and have had recent profit shocks.
The central issue in Çolak and Whited is, of course, on change in investment effi-
ciency after a focusing activity. Çolak and Whited use several measures of change in
investment efficiency, including investmentQ-sensitivity, the difference in adjusted in-
vestment to sales ratio between high and low growth segments, and the relative value
added, which is akin to weighted investment in high minus lowQsegments. Çolak and
Whited find that the changes in these measures are not significant relative to changes in
firms that do not focus and that have similar propensity scores, using theDehejia and
Wahba (1999)matching procedure and theAbadie and Imbens (2004)implementation.
There is no evidence that post-spinoff efficiency improves once the focusing firms are
matched by propensity score to the non-focusing firms.
For robustness, Çolak and Whited also report estimates of a treatment effects model,
equation (68) ofCampa and Kedia (2002). There is little evidence for efficiency gains,
except for one case in which the investment efficiency has a significance level of 10% for
focusing firms. This could, however, arise due to pure chance given the wide number
of dependent variables and specifications examined. While the paper does not report
the coefficient for the inverse Mills ratio in the treatment effects model, Toni Whited
confirms to us in private communication, that this selection termissignificant. This
suggests that self-selection is the main explanation for why firms experience efficiency
gains after focusing. The unobserved private information that leads firms to focus ex-
plains post-focusing improvements in efficiency; controlling for self-selection, there is
little evidence of any additional efficiency gains.