The Income Gradient and Distribution-Sensitive Measures of Overweight in the U.S.

September 2007

Dean Jolliffe.

Download 'working_paper07-27.pdf'.

Abstract

This paper uses quantile regression and alternate measures of overweight to examine the relationship between income and the body-mass index (BMI). In contrast to the prevalence measure of overweight, the alternate measures considered are sensitive to changes in the body-mass index (BMI) distribution and continuous in BMI at the overweight threshold. The standard prevalence measure indicates that there has been no association between poverty and overweight during the last 30 years of the 20th century. The most recent data from 2004 indicate that the prevalence of overweight for poor people is more than 5 percentage points lower than for the nonpoor. The alternative measures indicate though that the BMI distribution of the poor has been much more positively skewed over this time period, indicating a more severe overweight problem. Quantile regression analysis indicates that the strongest relationship between income and BMI is observed at the tails of the (conditional) distribution. There is a strong negative income gradient in BMI at the obesity threshold and some evidence of a positive gradient at the underweight threshold.

Keywords:
Health, Health Insurance, and Health Care, Research Methods