111
The Medical Care Costs of Childhood Obesity
Adam I. Biener - PhD Student, Lehigh University
Chad Meyerhoefer - Associate Professor of Economics, Lehigh University
John Cawley - Professor of Economics, Cornell University
From 2007 to 2010, an estimated 16.9% of US children and adolescents were obese, with significantly higher prevalence among non-Hispanic black and Hispanic children and adolescents. (Ogden et al., 2012) Further, about 33% of boys and 30.4% of girls are considered to be overweight. This paper estimates the marginal effects of increased Body Mass Index (BMI) in US children on their health expenditures. We sample children with biological parents aged 11 to 17 from the year 2000 through 2010 Medical Expenditure Panel Survey (MEPS), a comprehensive, nationally representative survey of the U.S. civilian non-institutionalize population.
Previous research has assessed the cost effectiveness of interventions for childhood obesity. Finkelstein and Trogden (2008) find that, on average, obese children and adolescents incur $220 more in medical spending than normal weight children and that overweight incur $180 more. Monheit et al. (2009) estimate a two-part model of medical expenditures and find that adolescent girls who become obese spend $790 more per year than normal weight girls. They find no significant effect for boys.
These previous estimates of the cost of childhood obesity do not explicitly account for possible endogeneity of weight or measurement error in weight. Child BMI may be endogenous in a medical expenditure model if, for example, children with limited opportunities to engage in physical activity and high body weight, also have poor access to health care. In addition, height and weight in the MEPS is parent-reported and may be contain measurement error.
To address these empirical problems, as well as account for the inherent skewness in the distribution of expenditures, we follow the method in Cawley and Meyerhoefer (2011) and estimate the marginal effect using instrumental variables generalized linear models (IV-GLM). The advantage of IV-GLM is that coefficient estimates will not be biased in the presence of measurement error or endogeneity in child BMI. Our identifying instrument is the BMI of the child’s biological parents. A large and growing body of literature supports the correlation of parent and child’s weight as being solely due to genetics or biological linkage, and not as a result of environmental factors. This validates our choice of instrument as the parent’s weight only affects the child’s medical spending through its effect on the child’s weight.
We estimate two-part models for sub-populations across race, gender, and insurance status and find that, on average, being overweight or obese increases a child’s annual medical expenditures by $554.26, and that this effect for girls may be as high as $1,475.15. We find no significant effects for boys. The size of these effects is larger than those found in previous research that does not account for measurement error or endogeneity.