Most civil cases settle. Yet generating a definitive settlement rate presents complex definitional and empirical problems, both in what should count as a settlement and how to count it. This Essay makes three contributions to better understanding and defining settlement. First, we propose a flexible, empirically informed, operationalizable definition of settlement as party resolution. Second, we exploit a new federal litigation data source to count party resolutions using machine learning models trained on 11 million docket sheet entries. Third, we offer new findings on party resolution frequency and distribution in the federal courts. Settlement is more widely and differently deployed than previously understood. We uncovered almost 40,000 additional party resolutions that were missing from the main existing source of administrative data on federal litigation. We also observe party resolution appearing alongside other dispositive events, functioning as a trimming device to drop parties, claims, or both as a lawsuit proceeds. We conclude by mapping directions for future work, integrating our settlement findings into a larger discussion of litigation and the courts’ role in achieving resolution, whether by a judge, jury, or the parties themselves.
