Measuring the Work of the Federal District Courts

The federal district court system is one of the largest and most impactful organizations in the United States. The nation’s ninety-four district courts resolve hundreds of thousands of cases, large and small, each year. Yet surprisingly little scholarly attention has been paid to measuring the work of the federal district courts—a vitally important task. It’s important because it affects substantive law; judges routinely decide the merits of issues based on how busy they think they and other judges are. It’s important because the law relaxes procedural protections—such as the Speedy Trial Act’s seventy-day deadline to bring criminal defendants to trial—if a court is perceived to be exceptionally busy. And, perhaps most directly, it’s important to the operation of the district courts; we can’t allocate judicial resources where they are most needed if we can’t accurately assess each district’s workload. The problem of judicial resource allocation is particularly acute today because the number of district court judgeships has remained stagnant for over two decades, the longest period in this country’s history. The backlog of new judgeships to be created provides all the more reason that we accurately measure the workload of the district courts. This Paper identifies problems with—and proposes improvements to—the way in which the federal courts measure their workload. For the last fifty years, the federal courts have used a system of case weights to measure their work. The system is simple: each type of case (patent, antitrust, etc.) gets a numerical weight, and the weights of all cases filed in a district are added together to determine the workload of that district. Through an exhaustive empirical analysis—spanning tens of thousands of cases, hundreds of thousands of pages of judicial opinions, and millions of docket entries—I find that the current system of case weights may significantly mismeasure court workload. Because the current system assigns a single, static weight (e.g., 4.72) to all cases of a particular type (e.g., patent cases) it assumes that cases of that type take, on average, the same amount of judicial work in every district court. But I find that patent case workload varies significantly between districts. I further analyze some of the most common types of cases in the federal courts (product liability cases and two types of civil rights cases) and find that they, too, exhibit significant district-todistrict variation in case workload. My results suggest that the current system overestimates the workload in some districts and underestimates it in others. This Paper’s theoretical contributions are anchored in these empirical results. The Paper is the first to pose—and the first to attempt to answer—a fundamental question: what should we be measuring when we measure a court’s work? I consider three distinct ways of conceptualizing judicial workload based on: tasks (e.g., number of cases resolved), outputs (e.g., number of pages of opinions issued), or resources (e.g., judge time). I categorize existing workload metrics into one of these categories. And I explain why, for courtadministration purposes, a resource-based metric (specifically, one based on the amount of time judges spend on cases) is better than other potential metrics. Building on this analysis, the Paper proposes a more accurate measure of workload that is just as easily administrable. Instead of the current system, which relies on an estimate of a case’s workload based solely on the case’s type (i.e., a static case weight), I propose measuring workload dynamically using evidence of actual work performed for each case. More specifically, I suggest calculating a dynamic weight for each case using, with minor modification, the same information the federal courts already use to calculate static case weights. My proposal addresses the issue identified in this Article and improves the way we measure the work of the federal district courts.