The Credibility Revolution Meets Research Design: Considerations for Antitrust Practitioners

Merging parties continue to advocate using cross-sectional regressions of price against Herfindahl–Hirschman Indices (HHIs) despite warnings regarding their inability to predict anticompetitive merger effects. Using the FTC vs. Kroger/Albertsons as a case study, we illustrate how regression design, in general, is important for generating reliable results. We show how demand and supply conditions can vary systematically across areas with different numbers of competitors such that cross-sectional comparisons of price and concentration confound multiple explanations, biasing results. We argue that different study designs control for changes in competitive conditions within local areas, but may be difficult to implement with available data.