This paper measures the effect of racial and ethnic diversity in police academy cohorts on officers' future job performance. I exploit the fact that entrance into the Chicago Police Department police academy is determined by randomly assigned lottery numbers to identify the long-run effect of peer diversity on policing outcomes. Using novel officer-level data on daily assignments and arrests, and by linking court outcomes to arrests, I construct metrics for individual officer arrest quantity and quality. I find being randomly assigned to an academy cohort with a 10 percent higher share of minorities decreases an officer's future arrests of Blacks by 3-11%. This is entirely driven by a decline in arrests for low-level crimes, with cohort diversity having no negative, and potentially a positive, effect on future arrests for serious crimes. The decline in low-level arrests also corresponds to an increase in average arrest quality, as cohort diversity has a large negative effect on low-quality arrests and a small negative effect on high-quality arrests. Notably, white and non-white officers are similarly affected by cohort diversity, suggesting that diversity influences how officers learn to police. These results indicate increased hiring of minorities in police departments may have positive spillovers on their peers.