Evidências empíricas da obesidade para o Brasil

Abstract

The present thesis comprises three chapters that analyze obesity and BMI of Brazilian adult population from different perspectives. For this, data from PNS 2013, made by IBGE, were used. The first chapter focuses on social inequality in obesity, measured by adapting the Absolute Poverty Indexes proposed by Foster et al. (1984). Social inequality was measured by concentration index (CI) and horizontal inequality index (HI). Finally, the IC was decomposed using Wagstaff et al. (2001) method. The results show a higher prevalence of obesity among women, being the most intense and severe phenomenon in this population group. Furthermore, obesity measures are concentrated among the richest men and the poorest women. The second chapter analyzes the BMI differential between urban and rural areas. Initially, the method of relative distribution proposed by Handcock e Morris (1998) was implemented. In addition, we use the decomposition based on regressions of RIF as presented by Firpo et al. (2007), which enables an analysis at different points in the distribution of probability of BMI. The results show that urban areas suffer more from obesity problems. In general, the decomposition reveals that, among men, the differences in terms of the level of schooling, and the type of work were fundamental to explain the differential in BMI. For women, there are discrepancies in the type of work, physical activity, and healthy eating. Finally, in Chapter 3, a differential assessment of BMI between men and women is made. The Lasso method proposed by Tibshirani (1996) was implemented in the estimations. In addition, the extended Oaxaca Blinder decomposition is employed, which incorporates a non-paired effect resulting from those covariates inherent in each group. Education affects men and women differently, being a risk factor among the former and a factor of protection between the latter. The decomposition results show that the structure effect has great weight in the BMI gap between the two groups. Multidimensional poverty, age and practice of physical activity are the variables with the greatest positive influence on this effect. In addition, differences in the gender-specific variables contribute to the reduction of this differential.