This study analyzes the consistency of the Brazilian national information systems on mortality (SIM) and live births (SINASC) as data sources for evaluating health inequalities according to race/color. Infant mortality rates (IMRs) were obtained according to race/color from death and live birth certificates for the country as a whole and its regions, for the years 1999-2002. The IMR was also estimated according to race/ color, based on the incorporation of deaths and live births with race/color not reported by two criteria. The study compared the IMRs obtained in the study with those estimated by indirect methods. The IMR ratios were also calculated between race/color categories. A substantial reduction was observed during the period in the number of deaths and live births with race/color not recorded. In 2002, infant mortality in black children was 30.0% to 80.0% higher than that of white children and 40.0% to 80.0% higher than that of brown or mixed-race children (pardas); infant mortality in indigenous children was 40.0% to 90.0% higher than that of white or brown children. It is hoped that improved recording in the SIM and SINASC databases will allow a more in-depth discussion of health inequalities according to race, color, and ethnicity in Brazil.