Explaining Trends in Body Mass Index Using Demographic Counterfactuals

Abstract

The United States is experiencing a major public health problem relating to increasing levels of excess body fat. This paper is about the relationship in the United States between trends in the distribution of body mass index (BMI), including trends in overweight and obesity, and demographic change. We provide estimates of the counterfactual distribution of BMI that would have been observed in 2003–2008 had demographics remained fixed at 1980 values, roughly the beginning of the period of increasing overweight and obesity. We find that changes in demographics are partly responsible for the changes in the population distribution of BMI and are capable of explaining about 8.6% of the increase in the combined rate of overweight and obesity among women and about 7.2% of the increase among men. We also use demographic projections to predict a BMI distribution and corresponding rates of overweight and obesity for 2050.

Keywords:

• BMI
• Counterfactual distribution
• Demographics
• Overweight and obesity

JEL Classification:

• C11
• I10
• J11

ACKNOWLEDGMENTS

We thank seminar participants at Purdue University and the University of Texas at Austin for helpful comments and discussions. Also, we thank two anonymous referees and an editor for helpful comments. Any errors are ours.

Notes

Note that the definitions of overweight and obese have changed over time and vary across papers. In some cases overweight is defined as BMI ≥25 while obese is defined as BMI ≥30, so that it is possible to be obese and overweight. In other cases overweight is defined as 25 ≤ BMI <30. This paper adopts the second approach, but consistent with much of the literature this paper reports the combined rate of overweight and obesity, and the rate of obesity alone. This makes interpreting the results easier: an increase in the combined rate of overweight and obesity is unambiguously a change to “worse” health outcomes while an increase in the rate of overweight alone could be a change to a “better” health outcome if it is the result of a lower rate of obesity alone.

Among the health effects of high BMI are hypertension, coronary artery disease and strokes, diabetes, certain cancers, and more broadly decreased life expectancy (e.g., Haslam and James, 2005). Berrington de Gonzalez et al. (2010) find that there is a U-shaped relationship between BMI and all-cause mortality, with the lowest mortality in the range of 20–25, and increased mortality in the underweight, overweight, and obese ranges of BMI. The “public” economic effects will be discussed, in the context of the results of this paper, in the conclusions. There are also “private” economic costs, including evidence that overweight and/or obesity reduces wages, as in Cawley (2004), among many others.

This is important and motivates our use of the NHANES even though other larger datasets are available. For example, according to self-reported BMI in the Behavioral Risk Factor Surveillance System survey only about 25.6% of the population was obese in 2007. According to the NHANES, that figure was 34%. See http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5728a1.htm or Yun et al. (2006). There is evidence that measurement error in BMI is not classical as, for example, it varies by true BMI (e.g., Rowland, 1990). There is also reason to suspect that measurement error in BMI depends on the survey design, and in particular on whether subjects will eventually be measured by a medical professional, which makes correcting the self-reported BMI more difficult. Given these difficulties, and the need to have an accurate measure of BMI, we use the NHANES data rather than the alternatives. See also Cawley and Burkhauser (2006), the working paper version of Burkhauser and Cawley (2008), for more on these issues and strategies for prediction/imputation of height, weight, and measures of body fatness from self-reported data. Burkhauser et al. (2009) illustrate the empirical importance of the possible different measures of excess body fat.

They find that the combined rate of overweight and obesity in 2007–2008 was 64.1 (72.3) and the rate of obesity was 35.5 (32.2) for women (men). Note that they age-adjust their estimates to the distribution of age in 2000, so their results are not exactly comparable to our results. Also note they consider only 2007–2008 while we consider 2003-2008 in order to increase sample size. Nevertheless, the results are basically consistent with the results from our model, as displayed in Table 2.

Also note that there is no reason that these partial counterfactuals need to “add up” to the complete counterfactual. In an extreme case, suppose that all demographics were simply relabellings of the same variable. Then each partial counterfactual would explain the same amount of the change as the complete counterfactual, and so the partial counterfactuals would “add up” to explaining a multiple of the amount explained by the complete counterfactual.

We calculate this as follows. The incremental direct medical cost due to weight (i.e., versus normal weight) in the population is (size of population) × (P(overweight)(266) + P(obese)(1723)). Therefore the difference for different rates of overweight and obesity is (size of population) × (ΔP(overweight)(266) + ΔP(obese)(1723)). In 2008 the adult population was 221, 057, 612 according to the U.S. Census Bureau. Tsai et al. (2011) also report another estimate based on more studies and find that the incremental direct medical costs of overweight is equal to $498 and of obesity is equal to $1,662 per person per year. Under such figures, we find that demographic change is responsible for roughly 4.4 billion dollars in direct medical costs of overweight and obesity per year.