Ch. 8: Conglomerate Firms and Internal Capital Markets 461
By contrast,Scharfstein and Stein’s (2000)model of intra-firm bargaining would
imply that the least productive divisions receive transfers from the most productive di-
visions. Again, an increase in diversity will lead to an increase in this transfer. That
model would predictγ<0 in cells (1) and (2) andγ>0 in cells (3) and (4).
RSZ test their model on 13,947 firm-years in the sample data is obtained from COM-
PUSTAT for the period 1980–1993. They separate regressions for each cell ofTable 1
and obtain parameter estimates that accord with the predictions in the table.
RSZ perform extensive robustness checks. They also verify that (a) investment devi-
ations that they that they classify as value increasing actually are positively related with
to diversified firms’ value and (b) that diversity itself is negatively related to firm value.
To summarize, even though some transfers in the right direction increase with diver-
sity (cells (2) and (3) in the table above), RSZ find that on average as diversity increases,
investment in segments with above-average opportunities becomes too small and in-
vestment in segments with below average opportunities becomes too large. This leads
reduces the value of the firm of diverse firms.
4.5. Investment under a profit—maximizing neoclassical model
TheMaksimovic and Phillips (2002)model differs form the preceding literature in sev-
eral regards. First, the tests are motivated by the neoclassical profit-maximizing model.
The model assumes that each firm has a corporate ability or talent, a fixed resource. It
chooses the industries in which it operates so as to extract the maximum value from
its ability, diversifying and focusing in response to demand shocks, and the consequent
changes in the opportunity cost of assets, across industries. Thus, the focus of the model
is not specifically on how well the internal capital market works, but on whether the di-
versified firms expand in segments in which they have a comparative advantage.
An implication of the MP model is that the decision to diversify is endogenous and
depends on segment productivity and industry demand shocks. This implies that the use
of single-segment firms as benchmarks for the values of conglomerates’ segments is
subject to selection bias.
Second, in empirical tests MP use plant-level Survey of Manufactures LRD data to
classify each firm’s plants into 3-digit SIC code industries. Thus, their classification
of firms’ assets in not subject to the same discretion that characterizes COMPUSTAT
segment data. Moreover, their sample is larger than that of comparable studies. How-
ever, the Survey of Manufactures only covers manufacturing industries, so MP, cannot
separately identify manufacturers who also operate outside manufacturing.
Third, instead of analyzing capital expenditures at the segment level, MP analyzes the
growth in value added. Thus, their measure takes into account growth through whole and
partial firm acquisitions as well as through direct capital expenditures.^33
(^33) Maksimovic and Phillips (2007)show that diversified firms are more likely to grow through acquisitions
than single-segment firms.