A Tale of Two Samples: Unpacking Recent Trends in Industrial Concentration

We use publicly available data from the U.S. Census Bureau's quinquennial Economic Censuses to examine trends in industrial concentration in the U.S. economy from 2002 to 2017. We find that, contrary to the popular narrative, industrial concentration is not rising and actually declined from 2007 to 2017. A notable difference between our methodology and previous studies using Economic Census data is that we use all available six-digit NAICS industries to conduct a cross-sectional analysis of industrial concentration levels in each Economic Census year, while previous studies have excluded industries subject to redefinition over time. When attention is restricted to only "comparable industries," we find, as with previous studies, that average concentration increased modestly from 2002 to 2017. We then show that restricting attention to comparable industries is problematic from a sample selection perspective as the sample of comparable industries exhibits substantially lower levels of concentration in 2002 than the set of industries later subject to redefinition in subsequent Economic Census years. The problematic nature of relying on the comparable industries sample to characterize economy-wide trends in concentration is confirmed by evidence that concentration levels in the comparable industries sample display significant mean reversion over time. Thus, finding a trend towards increasing concentration in this sample may, to a significant extent, reflect the role of transient economic shocks. Furthermore, to the extent changes in concentration are systematically related to competition and economic outcomes, we find that increases in concentration are correlated with increased output, increased employment, and higher wages. Case studies from the retail sector and the taxi industry provide examples of situations where increasing industrial concentration is the direct result of increasing market competition.