Ch. 8: Conglomerate Firms and Internal Capital Markets 465
and a control subsample of “regular” conglomerates that do not reduce the number
of segments so substantially. They run their regression separately on the two subsam-
ples. For the failed conglomerates the coefficientsφandθare not significantly different
from zero for the period prior to restructuring. The subsample of “regular” conglom-
erates these coefficients are negative and significant, as predicted by the model. Thus,
MP find evidence that there is subset of “failed” conglomerates that grow inefficiently,
and are subsequently broken up. However, even for these failed conglomerates MP
do not find a positive significant relation the segments’ growth rates and other seg-
ments productivity. Thus, they find no evidence that even these failed conglomerates
systematically grow their unproductive segments at the expense of productive seg-
ments.^38
MP also find that a segment’s relative size in the firm does affect its growth, even
controlling for productivity. Main segments of firms (i.e., segments that produce at least
a quarter of its value added) grow faster in response to positive demand shocks than
peripheral segments. In part this is because main segments are on average more pro-
ductive. However, a substantial growth differential remains even after controlling for
productivity.
The growth differential is especially pronounced in recessions. Rather than being
cushioned in recessions as predicted by models that stress bargaining within the firm,
peripheral segments of conglomerates are cut sharply in response to negative demand
shocks. These cuts are greater than predicted by MP’s simple neoclassical model. They
suggest that a more complex mechanism is at work. Thus, negative demand shocks may
cause diversified firms to reassess the prospects of their peripheral segments and to shift
resources into more promising ventures, as modeled byStein (1997).
The decline in peripheral divisions is also reflected in aggregate Census data. In the
beginning of the 1980s main divisions of diversified firms produced about half of the
value added by U.S. manufacturing and this share was maintained through the end of
the 1990ies. By contrast, the share of peripheral segments of diversified firms fell from
27.5% to 23.5% over that period.
In sum,Maksimovic and Phillips (2002)find that a simple profit maximizing neoclas-
sical model of firm growth across segments is consistent with plant-level data and that
there is little evidence of systematic resource misallocation by diversified firms. There
is some evidence that failed conglomerates that are subsequently broken up do not allo-
cate resources model efficiently. However, even these firms do not systematically grow
unproductive segments at the expense of productive segments. Instead, there is evidence
that smaller, less productive units of conglomerates grow more slowly than their main
divisions or similarly productive stand alone firms.
(^38) MP perform robustness tests using several alternative measures of productivity and investment. Their
model predicts, for example, predicts that segment size is a proxy for segment productivity. The results using
segments size yield the qualitative results.