This paper examines the impact of the ‘Programa Escola da Família’ on school violence indicators at São Paulo state schools in 2002-2008. The number of violent acts is modeled as a Poisson count process and the empirical strategy – a non-linear fixed effects model – explicitly accounts for unobserved school heterogeneity, possibly correlated to program participation. The estimates suggest that the program reduces the number of cases of violence, particularly those associated with vandalism and interpersonal relationships problems among school community members.