Explaining gender-specific racial differences in obesity using biased self-reports of food intake

Policymakers have an interest in identifying the differences in behavior patterns - namely, habitual caloric intake and physical activity levels - that contribute to demographic variation in body mass index (BMI) and obesity risk. While disparities in mean BMI and obesity rates between whites (non-Hispanic) and African-Americans (non-Hispanic) are well-documented, the behavioral differences that underlie these gaps have not been carefully identified. Moreover, the female-specificity of the black-white obesity gap has received relatively little attention. In the National Health and Nutrition Examination Surveys (NHANES) data, we initially observe a very weak relationship between self-reported measures of caloric intake and physical activity and either BMI or obesity risk, and these behaviors appear to explain only a small fraction of the black-white BMI gap (or obesity gap) among women. These unadjusted estimates echo previous findings from large survey datasets such as the NHANES. Using an innovative method to mitigate the widely recognized problem of measurement error in self-reported behaviors - proxying for measurement errors using the ratio of reported caloric intake to estimated true caloric needs - we obtain much stronger relationships between behaviors and BMI (or obesity risk). Behaviors can in fact account for a significant share of the BMI gap (and the obesity gap) between black women and white women and are consistent with the presence of much smaller gaps between black men and white men. The analysis also shows that the effects smoking has on BMI and obesity risk are small-to-negligible when measurement error is properly controlled.