Is it possible to reduce crime without exacerbating adversarial relationships between police and citizens? Community policing is a celebrated reform with that aim, which is now adopted on six continents. However, the evidence base is limited, studying reform components in isolation in a limited set of countries, and remaining largely silent on citizen-police trust. We designed six field experiments with Global South police agencies to study locally designed models of community policing using coordinated measures of crime and the attitudes and behaviors of citizens and police. In a preregistered meta-analysis, we found that these interventions led to mixed implementation, largely failed to improve citizen-police relations, and did not reduce crime. Societies may need to implement structural changes first for incremental police reforms such as community policing to succeed.