Abstract: We construct a new dataset using census, survey, and registry data from hundreds of sources to document a clear positive relationship between development and average establishment size in manufacturing across 134 countries. We rationalize this relationship using a standard model of reallocation among production units that features endogenous entry and productivity investment. The model connects small operational scales to the prevalence in poor countries of correlated distortions (the elasticity between wedges and establishment productivity). The model also rationalizes the finding in poor countries of low establishment-level productivity and low aggregate productivity investment. A calibrated version of the model implies that when correlated distortions increase from 0.09 in the U.S. to 0.5 in India, establishment size and establishment-level productivity fall by more than 82 percent and aggregate productivity falls by around 70 percent. Relative to the existing literature, these substantial size and productivity effects are more in line with cross-country data.
Keywords: misallocation, establishment size, productivity, investment, idiosyncratic distortions.
JEL Classification: O1, O4.