464 V. Maksimovic and G. Phillips
opportunity cost of staying in the industry rather than selling out to a high productiv-
ity producer. This effect is amplified because any reductions in growth or divestitures
in a segment in which the conglomerate is less productive will produce positive ex-
ternalities in its segments. Hence, a conglomerate which is a relatively unproductive
producer and therefore divests or grows more slowly in an industry that has received a
positive demand shock will grow faster than it otherwise would have in itsotherseg-
ments.
The predictions concerning cross-segment effects are derived inMaksimovic and
Phillips (1999, 2002)and briefly reviewed inAppendix A. They differ from predic-
tions of models that stress influence costs, which suggest that resources are trans-
ferred to unproductive segments, and empire-building models, such asLamont (1997)
that suggest that wealth generated by positive shocks is dispersed throughout the
firm.
To test their modelMaksimovic and Phillips (2002)run the following regression on
a sample of 270 thousand segment years over the period 1977–1992.
GROWTH=α+β(Industry shock)+γ(Segment TFP)
+δ(Industry shock)×(Segment TFP)
+φ(Other segments’ TFP)
+θ(Relative demand)×(Other segments’ TFP)+controls.
MP use TFP as measure of each segment’s productivity. The TFP takes the actual
amount of output produced for a given amount of inputs and compares it to a predicted
amount of output. The measure is computed at the plant level and aggregated up to seg-
ment level. “Predicted output” is what a plant should have produced, given the amount
of inputs it used. A plant that produces more than the predicted amount of output has
a greater-than-average TFP. This measure is more flexible than a cash flow measure,
and does not impose the restrictions of constant returns to scale and constant elasticity
of scale that a “dollar in, dollar out” cash flow measure requires. Demand shocks are
measured by changes in the industry real shipments.^37
Consistent with the model, MP find that productive segments grow faster(γ > 0 ),
especially in industries which have experienced a positive demand shock(δ > 0 ).Most
importantly, a segment’s growth rate is lower if the firm has more productive operations
in other industries(φ < 0 ). The segment’s growth is further reduced if these more
productive operations are in industries which have received a positive demand shock
(θ < 0 ). The last two finding are consistent with the cross-segment predictions of MP’s
neoclassical model but difficult to reconcile with an agency model in which the firm
invests inefficiently.
As a robustness check MP identify a subsample of “failed conglomerates” (diversi-
fied firms which restructure by decreasing the number of segments by at least a quarter)
(^37) MP show that their results also hold for several other measures of productivity and demand shocks.