Unlike my prior posts, this post is about analysis, not data. Apart from data deficits, there are analytic challenges that have held back study of trial-level courts. In particular, the existing scholarly literature on litigation tends to segregate two aspects of the litigation process: judicial decision-making and litigant choices. Legal scholars have been most focused on judicial decisionmaking, using the snapshot of (some non-random portion of) litigation available from opinions. Likewise, the judicial politics literature focuses on judicial decision-making, though in more quantitative and large-n ways, but tends to ignore how litigant choices shape the mix of cases actually heard and decided by the courts. Conversely, law-and-economics work has theorized about litigants’ decisions to settle before trial, but has not generally considered the known or anticipated identity of the judge.
In a proposed study of EEOC enforcement litigation (cases brought in federal district court by the EEOC as plaintiff), my colleagues Pauline Kim, Andrew Martin, and I hope to bridge these separate literatures in a large-n study, by modeling judge’s and litigants’ choices together in order to understand how they interact to structure the litigation process and its outcomes. The key insight is that litigation is iterative and dynamic: at each point in a sequence, the parties’ prospects, stakes, and costs may change, making settlement more or less likely, and the possibility of final adjudication exists at several of the points.
Our plan is to collect data that includes case-specific information not only about the fact of settlement, but also about the pretrial activity that preceded it and the nature of relief obtained, permitting a more dynamic model of the litigation process. For a substantial portion of the cases, particularly in later years, we will also be able to read the complaints, enabling us to look at the relationship between the monetary stakes and outcomes. Our sources are a list of the EEOC's litigated cases, and then the dockets and other documents from those cases, available via PACER.
Thus, we will be able to look not only at whether settlement occurred, but when and on what terms In addition, docket information can be analyzed to determine what types of litigation events (e.g. filing or denial of motions to dismiss or summary judgment motions, intervention by the complainant, discovery disputes) occurred prior to settlement and how those events influenced the likelihood or timing of settlement.
The statistical methods we'll pick should then take account of the progress over time of the litigations we study, as well as the stage at which cases settle. We're hoping, as well, to look at the amount rather than just the fact of settlement.
The point is that thinking about trial-level courts requires different analytic approaches as well as different (and more) data. We're going to take a stab at developing some; any thoughts are much appreciated.