INEQUALITY OF OPPORTUNITY IN BRAZIL: A MACRO-REGIONAL PERSPECTIVE WITH CONDITIONAL INFERENCE TREES AND RANDOM FORESTS
Opportunity Inequality; Conditional Inference Trees and Forests; Regional Heterogeneity; Intergenerational Mobility.
This dissertation investigated inequality of opportunity in Brazil and its regions, analyzing how circumstances beyond individuals’ control, such as race, sex, and parental education, influence their labor market income and how this relationship varies across regions. To this end, conditional inference trees and forests were employed, which are machine learning methods that reduce the researcher’s subjective intervention in the selection of variables and their interactions, thereby mitigating estimation bias. The analysis used supplementary microdata from the 2014 National Household Sample Survey (PNAD), the last year with information on parental education. The results indicate that these circumstances beyond individual control explain a substantial share of income inequality. The Shapley decomposition of inequality of opportunity Gini coefficients made it possible to qualify this estimate by revealing the relative contribution of each circumstance to explained inequality at both the national and regional levels, highlighting the central role of parental education and the persistence of race as structural factors shaping individuals’ opportunities in the Brazilian labor market. Regional heterogeneity was observed both in the most relevant circumstances and in the magnitude of inequality of opportunity, demonstrating that the opportunity structure is not uniform across the country.