Body mass index in families: spousal correlation, endogeneity, and intergenerational transmission

Previous studies have documented spousal and intergenerational correlations in body mass index (BMI) but few have considered familial weight data augmented with socioeconomic and behavioral control variables. This article considers a U.S. dataset that contains such information on husbands, wives, and grown children. Although certain variables (like education, race, and smoking status) are helpful in explaining an individual's BMI, the BMI of one's spouse (or parents) remains the most significant predictor of BMI. To help distinguish between correlation and causality in the married-adult BMI regressions, we consider two alternative approaches for dealing with possible endogeneity (due to omitted variables): (1) including spousal variables to proxy for omitted variables and (2) modeling spousal BMI in a hierarchical framework to explicitly allow for a "couple" effect. The results suggest endogeneity of educational attainment, but not smoking status, and support prior research that finds different associations of BMI with income for husbands and wives. For grown children, parental BMI and smoking status are identified as significant predictors.