Abstract
We evaluated whether the creation of Municipal Guards (GMs) improved the public safety indicators of the local governments that adopted them. Using municipal data, we tested several identification strategies based on the potential results model, focusing on binary treatments, discrete multivalued treatments and continuous treatments. The main results indicate that, in small and medium municipalities, having a GM with some years of activity may represent up to 30% fewer homicides than in their control group, which is equivalent to up to -4.8 deaths per 100 thousand inhabitants. On the contrary, the existence of GMs does not seem to impact property crimes.