Access Type

Open Access Dissertation

Date of Award


Degree Type


Degree Name




First Advisor

Allen C. Goodman


This study develops multidimensional measures of obesity based on six body measures, namely, weight, maximal calf circumference, thigh circumference, subscapular skinfold, waist circumference and percent body fat. Previous studies have used body mass index (BMI) as the measure of obesity. BMI does not fully consider body composition and is even more limited when applied to a population of subjects that are heterogeneous in muscularity, age, or bodyweight. Moreover, the use of BMI to classify people as obese and non-obese may result in misclassification problems. More recent studies have chosen some alternative measures of obesity, but there is no consensus on which one is the most accurate one.

This study employs the NHANES 1999-2004 database. To account for the complex survey design of the database, SAS-callable SUDAAN version 10 is used to conduct the analysis. The study sample includes respondents aged 20 and over, non-Hispanic persons who are examined, and women who are not pregnant. There are 2,736 women and 3,181 men who represent respectively, 52.06 million (men) and 55.04 million (women) in the overall population in the study.

In coping with the multicollinearity concerns over the correlated body measures, factor analysis is chosen here. In this study, the baseline model is OLS. Due to the potential endogeneity of obesity, Hausman specification tests are performed, and their results show that 2SLS is preferred over OLS for both gender.

The findings from body measures analysis show that the body compositions of men are more complicated than women. For men, besides the body weight and its related body measures, there are deviations of body fat and its distribution measures. The hypothesis tests for both OLS and 2SLS models demonstrate that for men the new multidimensional measures of obesity are better than the older single indicator in modeling income. For women, however the results are inconsistent.

In this study, obesity defined using the overall elasticity of six body measures to income presents a very different picture of Americans, especially men. Relative to the new criterion, BMI identifies fewer obese men who are otherwise classified as obese. Waist circumference identifies more obese people than BMI when obesity is defined by the new criterion.

In sum, the findings from this dissertation suggest that more body measures should be included in measuring obesity, at least, for men. Relative to most single indicators, multidimensional measures identify more obese men who were otherwise classified as non-obese. The study suggests that policy makers should pay more attention of the possible underestimation of the population of obese men.

Included in

Economics Commons