An analysis of pay and occupational differences by gender and race in Brazil - 1987 to 2006

This thesis investigates the magnitude and evolution of gender and racial occupational segregation and wage gaps in Brazil from 1987 to 2006. First, we provide the construction of a new harmonized and temporally consistent re-classification of the occupational codes using the Brazilian household survey, the PNADs. This new occupational classification permits an examination of the evolution of the Brazilian occupational structure over a protracted period of time. Second, we examine the occupational structure in Brazil assessing both the extent and trends in gender and racial based occupational segregation. We use several wellknown indices of segregation (Duncan and Duncan, 1955; Moir and Selby-Smith, 1979; Karmel and Maclachlan, 1988; Silber, 1989) and focus on the evolution over time of the occupational segregation across formal and non-formal labour markets. An attempt is made to assess the main forces driving changes in occupational segregation over time by employing a decomposition of the segregation measures developed by Deutsch, Flueckiger and Silber (2009). Third, we investigate the magnitude and evolution of gender and racial pay gaps in Brazil by employing several decomposition techniques. Together with the standard Oaxaca-Blinder decomposition, we apply the Brown, Moon and Zoloth (1980) decomposition technique, which allows us to account for the impact of occupational segregation on the wage gap. We explore the impact of the selection process on our decomposition results by employing different parametric corrections (the Heckman (1979) and Lee (1983) corrections). Several sensitivity checks are also implemented and alternative correction methods investigated such as the non-parametric imputation method by Olivetti and Petrongolo (2008) and the local wage gap estimation by Machado (2011). Fourth, we attempt to provide a comprehensive portrait of gender and racial wage gaps across the entire wage distribution while exploring the impact of gender and racial occupational segregation on wage determination in the Brazilian labour market. Our analysis particularly focuses on the evolution of the impact of female and non-white occupational intensity on wage outcomes and disparities. We employ quantile regression analysis in order to investigate the role of female and non-white occupational intensity at different points along the conditional wage distribution. We then apply two different decomposition techniques, proposed by Machado and Mata (2005) and Melly (2006), and by Firpo, Fortin and Lemieux (2009), to investigate the determinants of wage disparities at these different points in the wage distribution and to understand how these determinants vary across the wage distribution. Finally, we offer some concluding remarks, discuss the limitation of the research and provide an agenda for future research on the themes investigated in this thesis.