The Trouble with Approximating Industry Concentration from Compustat

Abstract

Using Compustat to measure industry concentration in the U.S. is problematic. Popular approximations of the Herfindahl Index based on this data set have a vanishingly low correlation with the more comprehensive Census measure. As a result, major variables of interest in corporate finance show markedly different correlations with these indicators, which can lead to a breakdown of regression results. This confirms findings of an earlier article that has been ignored by recent empirical studies. I show that Census concentration ratios are excellent measures and provide a newly compiled data set based on all concentration data ever produced by the Department of Commerce – eliminating the need to use the convenient but flawed Compustat files.

Keywords: Industry concentration, Herfindahl-Hirschman Index, concentration ratios, Compustat, Economic Census

JEL Classification: C81, G00, L11