Working Papers

The Boss in Blue: Supervisors and Police Behavior
[Supplementary Appendix]
I study how public sector supervisors influence the prioritization of agency objectives in the high-stakes setting of policing. Leveraging a context in which police officers frequently rotate between supervisors—sergeants—I estimate sergeant-specific effects on officer arrest decisions. I first show that sergeants substantially affect enforcement behavior: moving an officer from a 10th percentile to a 90th percentile sergeant in- creases monthly arrests by 42% relative to the mean. I then show that sergeants’ effects on serious and low-level arrests are weakly correlated and operate through distinct officer behaviors, suggesting variation in supervisors’ enforcement priorities. In particular, sergeants who increase low-level arrests do so primarily through discretionary drug enforcement, which disproportionately affects Black civilians and raises the incidence of police use of force. These findings underscore the importance of front-line supervisors in shaping public service delivery and identify a promising lever for police reform.

`Who You Gonna Call?` 911 Call Takers and Police Discretion
Police make high-stakes decisions under multiple sources of uncertainty, often resulting in adverse outcomes that are disproportionately experienced by minority citizens. What source of information do police use to make their enforcement decisions and does it generate accurate decisions? I examine a common information source for police officers throughout the United States: risk assessments conducted by 911 call takers. Exploiting random variation in the automated call distribution system used by a large metropolitan 911 center, I show that police are 37.7% more likely to make arrests at calls that the call taker deems risky. I demonstrate that officer perceptions are a primary channel through which the effects operate. This results in lower quality police decisions: misdemeanor arrests made at calls that are on the margin of being classified as risky are 14.9 percentage points less likely to result in a conviction. In cases where the race of the involved civilians is the same as the responding officer or the responding officer has more job experience, call taker information has a negligible effect on arrests. The results suggest that the strength of an officer’s priors determines the extent to which officers rely on information transmitted by call takers.

Teamwork in Policing (Draft available upon request)
I estimate the effects of teamwork in 911 call responses. In order to circumvent the endogenous assignment of more officers to calls with greater unobserved severity, I leverage exogenous variation in the number of double-manned patrol vehicles across shifts. I find that police are more likely to use their enforcement powers when dispatched in teams: an additional officer being dispatched to a call increases the likelihood of an arrest being made by 24% relative to the mean. I interpret these findings as evidence that officer teams reduce the costs of enforcement. Indeed, in the rare case where officers experience suspect resistance, the presence of more officers is associated with a lower likelihood of an officer being sent to the hospital. The results suggest that policing strategies which maximize patrol area to increase deterrence come at the cost of lower incapacitation capabilities.