This paper evaluates the causal impact of Rio de Janeiro’s Pacifying Police Units (UPPs), probably the largest–scale police reform initiative taking place in the developing world. The main goals of the UPPs were: 1) to regain control of territories previously dominated by armed criminal groups; and 2) to improve security for these communities through reduction of lethal violence. In the course of six years, more than 9,000 police officers were permanently assigned to the UPPs, servicing close to half million residents in the city slums (favelas). We are interested in understanding the process through which governments supply a basic public service –the police -- in poor urban neighborhoods that have long been abandoned to the arbitrary rule of non-state armed actors. Moreover, our paper documents Rio de Janeiro’s painful trajectory of police violence, illuminating some of it major institutional facilitators. Painstakingly geo-coding homicides and police killings from 2005 to 2013, we provide answers to some of the most critical questions about police use of lethal force, including the determinants of variations in who is targeted by police repression and how different strategies for policing the slums have impacted police killings. To evaluate the UPP impact on lethal violence, we use a variety of causal identification strategies that leverage spatial and temporal variation in the introduction of the UPP as well as geo-referenced data of more than 22,000 incidents of lethal violence. Our empirical models reveal that the UPP had mixed results. The introduction of the UPPs did not play a significant role in reducing murders in the favelas that were pacified. The UPP’s failure to reduce homicides imply that the poor in the slums continue to be subject to two or three times higher murder rates than the white middle class. Nonetheless, the UPP is breaking long-held practices of extreme use of police lethal violence. Our empirical results convincingly demonstrate that police killings would have been 60 percent larger without the UPP intervention.