This paper studies the effects of legislation in Brazil that banned the right to carry guns and provided for a voter referendum regarding whether to ban the sale of all firearms. Using a regression discontinuity design, I find that gun-related homicides decreased by 12.2% and gunshot wounds that were ‘intended to kill’ by 16.3%, with the reduction of the former being especially pronounced in high-crime areas and among black males. There is no evidence of substitution effect as non-gun-related homicides were not affected. Two pieces of evidence suggest that the mechanism explaining this result is a decrease in the number of people carrying guns in response to the legislation: first, the number of cases of illegal gun carrying decreased and second, only gun-related homicides taking place outside the residence were reduced. Analysis of the subsequent voter referendum, which was defeated by a wide margin, shows stronger support for the complete firearms ban in the areas more affected by gun violence.