Measuring Racial Disparities in Police Stops: Evidence from Telemetric Mobility Data
We establish a reliable counterfactual for evaluating racial disparities in traffic stops using telemetric data to estimate the racial composition of motorists. We measure of motorist racial composition using telemetric mobility data and a novel two-stage correction methodology anchored by an independent measure, racial composition of motorists involved in accidents. Applying this to Massachusetts State Police stops, we find non-White motorists are stopped at rates exceeding their roadway presence by 6.7 percentage points. We demonstrate that uncalibrated telemetric data significantly understates minority presence (15.2 vs. 28 percent), and the uncalibrated telemetric data yields disparities that are over twice the size of our preferred estimates. We use this validated measure to assess the accuracy of several commonly-used disparity tests. We find that the widely used “Community Standard” test dramatically overestimates disparities, even when restricting to local roads or non-commuting hours. In contrast, the less-often used “Crash Benchmark” accurately captures disparities even when aggregated to higher temporal levels covering up to 94 percent of stops. Finally, the “Veil of Darkness” test yields smaller estimates than our telemetric measure, consistent with its focus on disparate treatment rather than the broader legal standard of motorist composition.
-
-
Copy CitationXiaofeng Gong, Yuling Han, Susan T. Parker, Matthew B. Ross, and Stephen Ross, "Measuring Racial Disparities in Police Stops: Evidence from Telemetric Mobility Data," NBER Working Paper 35727 (2026), https://doi.org/10.3386/w35727.Download Citation