Good Dispersion, Bad Dispersion (joint with Nicolas Vincent)

Abstract:

Dispersion of marginal revenue products of inputs across firms are commonly thought to reflect misallocation. Consistent with that view, aggregate output monotonically declines in dispersion. We show that non-convex distortions to a firm's problem, however, break this monotonicity such that dispersion and efficiency are related in an inverted U-shaped fashion. Eliminating distortions may thus increase dispersion of marginal revenue products while improving the allocation and raising output. In a quantitative model of the U.S. manufacturing sector, we find that one quarter of the total variance of revenue products reflects “good dispersion,” while only the remaining three quarters are “bad dispersion” reflecting inefficient distortions. An important implication of this insight is that the welfare effects of eliminating distortions in emerging economies are larger than previously thought.