Dean Knox (UPenn)

Title

Monitoring racial bias in police traffic enforcement with imperfect proxies of driving behavior

Abstract

Decades of work has sought to assess whether police traffic enforcement is discriminatory by comparing the racial composition of those who are stopped to some external benchmark, such as the composition of local residents in a police jurisdiction. Such "benchmark analyses" have been criticized by statisticians and law-enforcement officials alike for yielding misleading results that fail to accurately represent the population at risk of police traffic stops—namely, individuals actually engaged in dangerous driving behavior. In this work, we present a causal framework for benchmark analysis that clarifies the distinction between valid and invalid benchmarks. We use this framework to formalize implicit assumptions in prior work and demonstrate how these assumptions can be empirically falsified. By drawing connections between benchmark analysis and a recent literature on negative control outcomes, we derive new estimators and partial identification results, with extensions to address a host of common statistical challenges that arise in benchmark analyses. Among other challenges, we consider scenarios where (1) "false positives" induce distortions in benchmark composition; (2) civilian race cannot be directly observed and must instead be inferred from surnames listed on citations; and (3) officer decisions lead drivers to modify the behavior that is captured by benchmarks. We demonstrate the generality of the approach with applications across multiple jurisdictions with differing data availability, leveraging records from red-light cameras, traffic collisions, and DUI checkpoints.