Data from the fifth- and eighth-grade waves of the Early Childhood Longitudinal Study—Kindergarten Class were analyzed. Unadjusted linear regression models of BMI percentile that included only indicators for child's race/ethnicity were estimated first followed by adjusted models that iteratively added sets of child, family, and ultimately school covariates. Separate models were estimated by grade and gender. School covariates included detailed indicators for the school socioeconomic characteristics, and the food and physical activity environments.
For Hispanic boys and girls and for black boys, substantial shares of the disparities in BMI were explained by differences in birth weight, BMI at school entry, and current child and family characteristics. Substantial disparities in BMI remained among black girls relative to white girls. Characteristics of the child's school during fifth and eighth grade—specifically, the schools' socioeconomic characteristics as well as measures of the food and physical activity environment—did not explain the disparities for any of the demographic groups.
Differences in the school environment had little additional explanatory power suggesting that interventions seeking to reduce BMI disparities should focus on early school years and even before school entry.