State-Building in the City: An Experiment in Civilian Alternatives to Policing

State penetration varies widely within cities, with well-governed areas abutting persistently neglected ones. Governments are seeking ways to improve penetration, local security, and state legitimacy. We experimentally evaluate a 20-month non-police intervention in Medellín, Colombia, that dramatically increased municipal personnel and agency attention to 40 neighborhoods. Despite the intensity, average impacts on security and perceived legitimacy were negligible. Prespecified subgroup analysis reveals important heterogeneity, however. Where state governance began relatively lower, impacts were null to negative, but in initially high-governance sectors, security and state legitimacy significantly improved. These divergent impacts apparently resulted from city staff and agencies systematically underdelivering in low-initial-governance sectors. Bureaucratic capacity and incentives to deliver often depend on baseline state engagement, trust, and accountability. This could result in increasing marginal returns to statebuilding, which in turn would lead to persistent "neglect traps"—political attention and investment where state presence is already robust, reinforcing existing disparities.