objective: to describe the temporal patterns of mortality by homicide in brazil. methods: a series of homicides in brazil from 2000 to 2009 were studied. the explanatory variables were race/skin color, gender and education. the death statistics were obtained from the mortality information system. a trend analysis was performed by means of a polynomial regression for a historic time series (p < 0.05, 95% confidence interval). results: the black population represented 69% of the homicide victims in 2009. the homicide rate increased in the black population, while it decreased in the white population in the period studied. the homicide rate increased in groups with both higher and lower education among blacks; among whites, the rate decreased for those with the lowest level of schooling and remained stable in the group with higher educational levels. in 2009, blacks had a higher risk of death than whites from homicide, regardless of education level. between 2004 and 2009, the homicide rate decreased in the white population, while it increased in the black population. conclusions: the relative risk of falling victim to homicide increased in the black population, suggesting an increase in inequality. the effect of the anti-gun measures implemented in brazil in 2004 was positive in the white population and less pronounced in the black population. overall, race/skin color predicted the occurrence of homicide.