Limiting aggressive policing can reduce police and civilian violence.

• This paper estimates the effect of a limit on police use of force on violence – both police and civilian violence – in Rio de Janeiro, Brazil. • Police killings and shootings plummeted after the limit. I argue that the internal oversight effectively tied the police's hands in the short run. • Civilian homicides, usually associated with organized crime, also fell. I show two ways that less police violence can slow criminal conflict. • A key takeaway is that restraining the police in high-crime settings may lower violence, both that committed by police and by criminal groups. • These results cast doubt on the efficacy of hard-on-crime policies and suggest that more police restraint, at a minimum, does no harm. Governments in the Americas rely on aggressive policing tactics to fight crime, despite scant evidence of impact. While recent studies depict militarized policing as a driver of violence, few governments have reconsidered their use of it. What impact does a restriction on aggressive policing have on violence, and why? This paper examines limits on police use of force and how they can be implemented to reduce both police and civilian violence. I argue that reforms that require internal, non-police oversight can be effective institutional constraints, minimizing police violence. In settings where organized crime is widespread, these limits can have spillover effects and further decrease civilian violence by (1) slowing the territorial diffusion of criminal conflict and (2) making conflict more predictable. I test these claims by examining an abrupt limit on police raids in Rio de Janeiro, Brazil. I find that limiting raids – militarized police strikes targeting criminal gangs and communities under their control – led to a 66% decrease in police killings and a 58% decrease in homicides. The effects were concentrated in police precincts where rival criminal groups are in close proximity. Limiting raids did not lead police to be more violent during ordinary patrolling duties, and did not affect property crimes. The implication is that restraining police use of force in high-violence settings may save lives and be no worse than hard-on-crime strategies. [ABSTRACT FROM AUTHOR]