Over the last 15 years, several Latin American cities have adopted dry laws, which restrain the sale of alcohol in bars and restaurants during specific hours of the week. Bogotá, in 1991, was the first. Several more have followed suit, or are likely to do so in the near future. Policy makers and the general press have argued that these measures reduce crime. In this paper, we use a particular feature of the adoption of laws in the São Paulo Metropolitan Area (SPMA) to estimate the effect of dry laws on the ultimate form of violent crime: murder. Between March 2001 and August 2004, 16 out of the 39 municipalities of the SPMA have adopted, at different dates, dry laws. By comparing the dynamics of homicide between adopting and non-adopting cities, we estimate that dry laws reduce homicides by at least 10%, with an even higher effect in high crime cities. Results are robust to inclusion of a large set of controls, to propensity score matching, to outliers, and to correction possible spillover effects from adopting to non-adopting cities.