A Few Bad Apples? Racial Bias in Policing

We provide new evidence on the presence and distribution of racial bias in the criminal justice system. In many states, the punishment for speeding increases discontinuously with the speed of the driver, exhibiting large jumps in fine amounts. It is a common practice for officers to reduce the charged speed to just below this jump, avoiding an onerous punishment for the driver. Using data from the Florida Highway Patrol, we find evidence of significant bunching in ticketed speeds below a jump in punishment for all drivers but significantly more for whites than for blacks and Hispanics. We estimate the bias of each officer by comparing his lenience towards whites and non-whites, allowing us to recover the full distribution of bias. The total disparity in lenience across races can be explained by a small percentage (~20%) of officers. Officers tend to favor drivers of their own racial group, and younger, female, and college-educated officers are less biased. We then estimate a model that allows for both heterogeneity in officer preferences and driver speeds across races. Because minorities tend to live in areas where officers are harsher to all drivers, policies targeting bias have little effect on the aggregate speed gap. We find that racial bias in lenience explains 16% of the minority-white speed gap, and spatial differences in race-blind lenience explain 30% of the gap.