When Evidence Becomes Practice: Hot Spot Policing in a Lower-Capacity Setting, (with Eduardo Fagundes, Gaston Pezzuchi, Rodrigo Serrano-Berthet, Claudio Todisco, Carolina Zancan and Reginaldo Santa Rosa) 2026. Submitted.

Public security agencies in Latin America operate under chronic human resource shortages, reactive deployment models, and fragmented analytic capacity, raising doubts about whether evidence-based policing can be effectively implemented in such environments. This article examines how organizational conditions shape the implementation and impact of a city-wide hot spot policing experiment — a police practice with solid evidence of effectiveness in the developed world but limited evidence of success in Latin America. A research–practice collaboration with the Military Police of Paraná in Brazil identified street segments that concentrate robberies and randomly selected segments to receive increased police visits. Patrol levels were measured for the first time using GPS data, revealing very low baseline levels of preventive patrolling and high compliance with the experimental protocol. The results indicate meaningful impacts: treated segments experienced a 22 percent reduction in robberies during treated days. No geographical displacement was found, but thefts increased, suggesting a shift from violent to nonviolent property crime. The study makes two contributions to public administration theory: it demonstrates how a research–practice collaboration can activate latent organizational capacity in lower-capacity settings by opening space for experimentation, and it identifies the internal capabilities that allowed the organization to convert evidence into practice: dedicated analytic units, repurposable operational units, mid-level managers who negotiated and sustained the innovation across organizational boundaries, and GPS-based accountability. 

Building Capacity from Within: Selective Segmentation and Police Reform in Latin America (with Michael Weintraub e Andre Mancha), 2025 Revised and resubmitted at the American Political Science Review.

Democracies in the Global South often face extreme violence alongside police forces that are ineffective, politicized, and abusive. We argue that governments can overcome dysfunctional policing through selective segmentation: creating new policing units with distinct recruitment, incentive, and oversight regimes that are insulated from dysfunctional parent institutions. We study this strategy in Ceará, Brazil, examining the Rapid and Visible Response Patrols (RAIO), a selective rapid-response squad. Using a staggered difference-in-differences design, we find that RAIO reduced homicides by 57 percent and robberies by 84 percent, with arrests falling alongside crime—a pattern consistent with deterrence as the primary operating mechanism. A survey of 2,000 residents with an image-based conjoint experiment shows that RAIO officers are viewed as more effective, less corrupt, and less abusive than conventional police. RAIO did not increase police killings and yielded electoral rewards for the incumbent governor, suggesting that selective segmentation can improve security while creating political incentives to sustain reform.

Regaining the Monopoly of Violence: Evidence from the Pacification of Rio de Janeiro’s Favelas (with Claudio Ferraz and Bruno Ottoni), 2023.

The proliferation of non-state armed groups and their local dominance of territories in in many developing countries has often resulted in a dramatic increase in criminal violence. This paper examines the effects of a large intervention aimed at retaking control of slums dominated by drug gangs in the city of Rio de Janeiro.We use the diferential timing in the implementation of this policy to asses its impact on crime both inside and outside slums. We find that territorial occupation by the police in slums significantly decreases violence and property crimes not only in these areas but also in their surroundings. We detect negative spillovers to neighboring municipalities, which suggests that criminal groups relocate and fight for turfs in regions previously liberated from their rule. 

Body-Worn Cameras and Racial Disparities in Police Violence: Evidence from Brazil (with Eduardo Fagundes, Leandro Piquet and Pedro Souza), 2026. Submitted.

We examine the effects of introducing body-worn cameras in one of the world’s largest police departments. Using the staggered rollout of the technology across units in São Paulo’s Police Department in 2022, we find that the cameras led to a 70.0% reduction in police lethality. This decline is entirely driven by a decrease in violence against Black individuals, who previously faced disproportionately high mortality rates. After the cameras were introduced, the lethality rate for Blacks aligns with that of Whites. Effects are concentrated in low-income census tracts, with a high proportion of Blacks, and with high baseline levels of police violence. We find no evidence that the cameras reduced police effort or altered citizen behavior.

Measuring the Expansion of Armed Criminal Groups in Rio de Janeiro (with Eduardo Fagundes, Mariana Carvalho and Ramon Chaves), 2023.

Many urban areas around the world are marked by the presence of armed criminal groups that often control territories and exploit several economic activities. Yet data on where these actors operate and what activities they engage in is scarce and rarely available to the public. This article introduces a new, comprehensive dataset with the presence and economic outreach of criminal groups that operate in Rio de Janeiro, a city where drug factions and militia groups have controlled territories for at least three decades. We use automated text analysis to read more than 420,000 citizen reports on criminal groups’ activities to map their presence, geographical expansion and characterize their economic activities in the metropolitan region of Rio de Janeiro over a 12 year period. We validate our measure of group presence with intelligence data and analyze how group presence correlates with measures of violence and criminal records.

Criminal governance in times of crisis: Evidence from the COVID-19 outbreak in Rio de Janeiro (with Benjamin Lessing and Santiago Tobon), 2023

In urban peripheries worldwide, and especially in Latin America, criminal groups use  coercive power to impose rules on and provide order to civilians. The reasons why gangs govern in particular ways, or at all, are poorly understood. Many charge taxes in exchange for governance provision—suggesting they act as stationary bandits—but some do not. Many control retail drug markets, but some also earn rents from licit goods and services like cooking gas and internet. During the COVID-19 crisis, anecdotes of gangs enforcing lockdowns and providing health-related public goods suggested they seized opportunities to consolidate their authority and perceived legitimacy.  We present novel, systematic data on  criminal governance practices in Rio de Janeiro, whose gangs are notoriously militarized, persistent, and—usefully, from our perspective—diverse. While many belong to prison-based drug syndicates, others are police-linked groups known as milicias. We surveyed residents from almost 200 favelas about local gangs’ type, economic and governance activities,  taxation, and pandemic response. Contrary to expectations, we find that drug gangs and  milícias alike earn rents from a range of licit products and services, enjoy similarly high levels of perceived legitimacy, and largely avoided involvement in pandemic response. Yet milicias are far more likely to tax, and seldom sell drugs. Our findings suggest that gangs’ core motives are economic rather than political, that they strategically distinguish between direct taxation and extracting monopoly rents from control over utilities. 

Learning to Punish. Resource Windfalls and Political Accountability (with Claudio Ferraz), 2014. 

We use exogenous variation from oil-based revenues paid to local governments in Brazil to test whether voters reward incumbent mayors for increasing government spending and whether this reward varies over time as voters learn about the source of revenues and public service delivery. We find that an unexpected increase in oil-based revenues increase reelection rates for incumbents in the short-run. However, when we look over different elections allowing voters to learn about the revenue windfall and public goods’ provision, we find different results. The reelection rates of incumbents, in elections that occur 10 years after the beginning of the oil boom, is lower when municipalities had an early oil windfall, when a large share of voters are informed (i.e.high media penetration and educated), and when public services and local development have not improved over the past decade

Media Networks and Political Accountability: Evidence from Radio Networks in Brazil (with Horacio Lareguy), 2014.

We examine how different types of media structure contribute to political accountability, analyzing the role of local radio stations, regional radio networks and television stations in compensating for the bias of Brazil’s federal government against non-aligned municipality mayors when allocating drought relief aid. Theoretical and empirical evidence shows that radio networks play a central role in reducing federal government bias due to their ability to exchange information among affiliated radio stations operating in different municipalities. Radio networks, however, are only effective when they contain non-aligned municipalities that received federal aid; only information from such municipalities reveals the federal government’s bias.  

The Effects of Pacification Policy on Drug Faction Conflicts in Rio de Janeiro, 2013.