Missing Missingness in Merger Analysis

Data and statistical modeling have played an increasingly important role in analysis across disparate areas of law. But courts' ability to assess the validity and reliability of the analyses that rely on these data has not kept apace. This mismatch between the law's reliance on data and an ability to appropriately evaluate analyses using these data is especially acute in antitrust challenges to horizontal mergers by the U.S. Department of Justice (DOJ) and the Federal Trade Commission (FTC). Whereas enforcement agencies and courts once applied relatively simple rules about the structure of a market, the analytical landscape has become more dependent on sophisticated economic theories and data analysis techniques. Increased reliance on such models presents opportunities for creating better economic outcomes on average. But the use of observational data also carries with it often unacknowledged hazards. This is a problem. Observational data often suffer from missingness, meaning these data may be randomly or systematically incomplete. Whereas random missingness creates imprecision, systematic missingness results in bias, which may lead a court or agency to improperly enjoin or allow a merger. This Article explores the conditions under which data vital to merger analysis may be missing, as well as its effects. As an illustration, this Article evaluates the court's discussion of data in F.T.C. v. Sysco through the lens of missingness and conducts simulations to examine how more complete data would have altered the court's analysis. Finally, this Article offers changes to current practice to both increase transparency of and public confidence in the courts' use of these data in merger review