Measuring and tracking obesity inequality in the United States: Evidence from NHANES, 1971-2014

Background: Because people care about their weight relative to peers and society, obesity inequality plays a role in explaining obesity incidence and the impacts of being obese on subjective well-being. While the increase in obesity prevalence and mean body mass index (BMI) is well documented, the measurement of distributional changes and corresponding obesity inequality is yet to be fully explored. Methods: The present study analyzed BMI data for adults aged 20 to 74 from the National Health and Nutritional Examination Survey (NHANES) I (1971-1974), II (1976-1980), III (1988-1994), and continuous NHANES (1999-2014). We applied tools developed to measure income inequality to analyze the inter-temporal variation in the BMI distribution among US adults. Using stochastic dominance tests, we construct partial orderings on cumulative BMI distributions during the study period. Shapley decompositions and inequality indices are employed to quantify the source and extent of temporal variation and decompose the inequality into within and between-group components considering age, gender, and race. Results: The BMI distribution of each NHANES study first-order stochastically dominated the BMI distribution of the previous wave from 1971-1974 to 2003-2006, whereas more recent comparisons failed to reject the null hypothesis of non-dominance. The Shapley decomposition analysis revealed that horizontal shifts of BMI distributions accounted for a majority of the increase in obesity prevalence since 1988-1991. Especially in recent years when the rate of obesity growth has slowed down, the contribution of the redistribution component dropped significantly and even became negative between 2007-2010 and 2011-2014. The inequality indexes consistently show a worsening of obesity inequality from the mid-1970s to the mid-2000s regardless of population subgroups, and this disproportionate shift of the BMI distribution is unlikely to be a result of a changing ethnic composition of the US population. Conclusion: Our findings demonstrate that seemingly similar increases in obesity prevalence can be accompanied by very different patterns of distribution change. We find that the early phase of the obesity epidemic in the US was largely driven by increasing skewness, whereas more recent growth is a population-wide experience, regardless of demographic characteristics. Increasing morbid obesity certainly played an important role in the initial phase of the epidemic, but more recently the BMI distribution has largely horizontally shifted to the right.