462 V. Maksimovic and G. Phillips
Fig. 6. Productivity ordered by segment size. Segment productivity estimated from LRD plant-level data.Fourth, MP do not use Tobin’sqs of single-segment firms to proxy for a segment’s
growth options. Instead, they use the industry growth in real total value added to obtain
industry level measures of investment opportunities. To provide a measure of segment
productivity at the micro level, they benchmark each plant in an industry against every
other plant in the industry to obtain each plant’s predicted real value added in each
year given the inputs (capital, energy, labor and materials) the plant used in that year.
They use the difference between the plant’s actual value added and the predicted value
as a measure of the plant’s relative productivity. They aggregate up their measure plant
productivity to derive the segment-level Total Factor Productivity (TFP) for each year.^34
Using LRD data, MP can directly observe how the productivity of diversified firms’
segments varies by the number of segments and the segment’s relative size within the
firm.Figure 6summarizes the data. Controlling for the number of segments in a diver-
sified firm, TFP decreases as the segment’s relative size within the firm falls.
The pattern of productivity inFigure 6is consistent with a neoclassical model in
which firms spread their operations across a range of industries in which they have a
comparative advantage and in which they have decreasing returns to scale.^35 Consistent
with the model, larger firms are more productive on average than smaller firms. How-
ever,Figure 6can also be given an agency interpretation: large productive firms may
(^34) Schoar (2002)shows that TFP at the firm level predicts the conglomerate discount for diversified firms.
(^35) Decreasing returns to scale may arise from the production technology. More broadly, a firm may perceive
itself as having decreasing returns to scale because expansion may provoke a competitive response by rival
firms.