Research
Crackdowns or Conciliation? How Anger and Fear Shape Public Security Preferences in Mexico (Job Market Paper)
Perceptions of crime shape citizens’ views on law enforcement and public security, yet we know far less about how those perceptions translate into concrete policy preferences. This paper opens that black box by centering emotions—unpacking how people feel about crime. My core claim is that two negatively valenced emotions, anger and fear, powerfully influence security preferences but in countervailing directions. Because anger is marked by blame attribution, a sense of control, and greater risk-seeking, it aligns with escalatory, coercive responses such as punitive crackdowns. Fear, by contrast, prioritizes restoring personal safety, risk avoidance, and stability; amid a surge in crime violence it inclines individuals toward therestoration of everyday order rather than retribution against cartel networks. Accordingly, they are more receptive to conciliation via state- mafia agreements—pragmatic pacts in which authorities trade limited leniency (e.g., amnesties, truces, tacit non-enforcement of circumscribed illicit activity) for credible cartel commitments to reduce violence. To test these claims, I conduct a lab experiment in Puebla, Mexico—an area experiencing a sharp surge in cartel violence—that isolates the causal impact of anger versus fear on downstream preferences. The findings align with the expectations. Taken together, these results challenge the presumption that emotional distress inevitably hardens punitive, iron-fist demands. Anger and fear, activated by different cues, channel citizens toward diametrically opposed security strategies.
When Rapacity Pacifies: Natural Resource Booms, Economic Diversification, and Criminal Violence.
Journal of Conflict Resolution (forthcoming)
Natural resource price shocks intensify competition for control of windfall sites. They also prompt criminal organizations to redeploy fighters toward those windfalls, thinning their presence in other resource regions that do not share the price surge. I test this in Mexico, focusing on drug cultivation and the extortion of mining operations. Exploiting exogenous variation in global gold prices (2001–2018) in a panel of 2,444 municipalities, I find that the gold boom led to a 15% decrease in homicides, a 19% drop in cartel-related executions, and a 19% reduction in active cartel cells in the average heroin-producing municipality. Conversely, when U.S.-bound heroin trafficking surged, violence in gold-mining areas fell. Together, the results reveal two linked mechanisms: rapacity in windfall zones and a peace dividend in resource areas bypassed by the boom. I further show that the decline is driven chiefly by reduced inter-cartel competition, with cartel–state clashes essentially unchanged.
Governing Crime with Code: Public Preferences for AI Use in Law and Order (With Gustavo Flores-Macías)
Law enforcement increasingly relies on artificial intelligence (AI) for a range of tasks, yet the fairness implications of delegating decisions formerly made by police officers, judges, and parole officers remain unsettled. On one hand, AI can encode or amplify existing biases and erode due-process protections; on the other, standardized and auditable procedures may curb arbitrary enforcement and promote consistency. Because authorization, funding, and citizen cooperation hinge on perceived legitimacy, mapping public preferences is a precondition for anticipating uptake, compliance, and equity impacts. To probe public attitudes, we field a large (n=10,000) single-profile conjoint experiment in Argentina, Brazil, Mexico, and the United States. The design varies four
attributes of a hypothetical program: the type of AI tool (facial recognition, social-media monitoring, predictive policing), the implementing agency (police vs. courts), the targets of use (general public vs. public servants), and the technology’s national origin. Preliminary results show the public is highly discriminating: support is strongest for facial recognition, followed by social-media monitoring, and weakest for predictive policing, with additional sensitivity to the agency, target, and origin. These preference patterns carry behavioral implications: they predict openness to adopting AI in public security, willingness to share personal information to train the models, and willingness to pay additional taxes to fund deployment. Overall, the findings map a nuanced demand curve for AI in policing rather than blanket endorsement or rejection.
Criminal philanthropists
(With Camila Angulo and Rodrigo Castro Cornejo)
Criminal organizations increasingly distribute private and public goods to civilians as part of constructing their own governance. Yet whether these non-coercive overtures actually purchase goodwill—and at what cost to the state—remains unclear. This project offers a systematic test via a nationwide survey experiment in Mexico. We present realistic vignettes in which cartels deliver food or medical assistance and then assess whether such provision elevates civilian support, erodes trust in the state, and increases tolerance toward neighbors who collaborate with cartels. We also probe a key condition: do any legitimacy gains evaporate once the same groups are portrayed as violent? By randomizing cues to non-coercive versus coercive provision, we isolate how “help” versus “harm” reshapes reliance, approval, and judgments about everyday criminal rule. The findings speak to civilian agency in high-crime settings and to how states can deliver services without ceding legitimacy to armed non-state actors.
It Takes a Criminal to Catch a Criminal: Experimental Evidence on Organized Crime as a Provider of Order.
(With Aldo Ponce Ugolini)
