Despite declining smoking rates in the U.S., a substantial fraction of adolescents still smoke. In addition, there are notable racial differences in adolescent smoking. We use Add Health data and apply a nonlinear decomposition method to determine the extent to which racial differences in observable characteristics account for (i) the racial smoking gaps in adolescent smoking (ages 12–18) and (ii) racial gaps in the probability of becoming a smoker in young adulthood (ages 18–24), conditional on being a non-smoker in adolescence. The model includes a host of explanatory factors, including individual, family socioeconomics, smoke exposure, school characteristics, and county crime rate. Of the 19 (9) percentage-point gap in white-black (white-Hispanic) smoking in adolescence, these factors together account for 22–28% (39–77%) of the smoking gap; and of the 18 (13) percentage-point gap in white-black (white-Hispanic) smoking up-take in young adulthood, these factors together account for 26–50% (48–100%) of the gap, depending on which set of coefficients are used for the decomposition. The biggest drivers of racial smoking gaps in adolescence are differences in friends’ smoking and school peer smoking, while only school peer smoking contributes to the explained portion of racial gaps in smoking up-take in young adulthood.