Pulling Back the Curtain on the Federal Class Action

• Published in: Journal of tort law Vol. 17; no. 2; pp. 117 - 148
• Main Authors: Gelbach, Jonah B., Hensler, Deborah R.
• Format: Journal Article
• Language: English
• Published: Berlin De Gruyter 28.10.2024; Walter de Gruyter GmbH
• Subjects: Certification; civil litigation; Class action lawsuits; class actions; Debates; dockets; large language models; machine learning
• ISSN: 1932-9148; 2194-6515; 1932-9148
• DOI: 10.1515/jtl-2025-0002

Practitioners, judges, and scholars have long debated numerous aspects of class litigation policy, with many disagreements involving competing empirical assertions. Astonishingly, given the long-running debate about the virtues and vices of class actions, we lack basic information about the federal class action caseload that would allow us to assess these claims, largely because the federal judiciary does not report even the most basic facts about class action litigation. Despite the critical role of class certification in the design of Rule 23’s scheme for class litigation, for example, we do not know how many cases have had class certification motions filed – not over time, not by the nature of suit, and not even in any particular year. What evidence we do have about class actions is fragmented in one way or another, e.g., involving only securities cases (which, unlike other cases, are governed in part by the PSLRA), or involving only cases with proposed settlements, as discussed in Fitzpatrick (2024). The lack of basic information about the federal class action universe facilitates a policy reform debate that enables contending parties to endlessly argue about the virtues and vices of the class action procedure without needing to account for how the procedure works in practice. The debate takes place in public while the class action process itself proceeds behind a curtain of ignorance. The goal of this paper is to start pulling back that curtain by reporting some basic information about federal class action filings and key events in the class action litigation process. Using docket entry text for cases filed between January 1, 2005, and December 31, 2014, we report some basic statistics regarding (1) how many putative class action cases there are, (2) how many have motions for class certification filed, (3) how many have an order on class certification, and (4) the timing of these events. We focus on certification because it is the key event in the life of class litigation: without it, the outcome of the litigation cannot bind the proposed class. Not surprisingly then, claims about certification rates have been central to policy debate about class actions. But except in limited circumstances, these claims have not previously been subject to rigorous analysis. Our approach is based on using a fine-tuned large language model to classify docket entries, as other recent work has done. We believe this will be a fruitful approach for future work in the area.