On the U.S. Firm and Establishment Size Distributions

This paper revisits the empirical evidence on the nature of firm and establishment size distributions in the United States using the Longitudinal Business Database (LBD), a confidential Census Bureau panel of all non-farm private firms and establishments with at least one employee. We establish five stylized facts that are relevant for the extent of granularity and the nature of growth in the U.S. economy: (1) with an estimated shape parameter significantly below 1, the best-fitting Pareto distribution substantially differs from Zipf's law for both firms and establishments; (2) a lognormal distribution fits both establishment and firm size distributions better than the commonly-used Pareto distribution, even far in the upper tail; (3) a convolution of lognormal and Pareto distributions fits both size distributions better than lognormal alone while also providing a better fit for the employment share distribution; (4) the estimated parameters are different across manufa cturing and services sectors, but the distribution fit ranking remains unchanged in the sectoral subsamples. Finally, using the Census of Manufactures (CM), we find that (5) the distribution of establishment-level total factor productivity---a common theoretical primitive for size---is also better described by lognormal than Pareto. We show that correctly characterizing the firm size distribution has first order implications for the effect of firm-level idiosyncratic shocks on aggregate activity.