Decomposing race and gender differences in underweight and obesity in South Africa.

Using data from the National Income Dynamics Study, we document differentials in both underweight and obesity across race and gender in post-Apartheid South Africa. Using a nonlinear decomposition method, we decompose these differences across gender within race and then across race within gender. Less than one third of the differences in obesity and underweight across gender are explained by differences in covariates. In contrast, at least 70% of the obesity differences across race are explained by differences in covariates. Behavioral variables such as smoking and exercise explain the largest part of the bodyweight differentials across gender. For bodyweight differentials across race within gender, however, socioeconomic status and background variables have the largest explanatory power for obesity differentials, while background variables play the key role in explaining the underweight differentials. These results indicate that eradicating obesity and underweight differentials will require targeting policies to specific groups.