Unpacking Gender and Race Segregation along Occupational Skills and Socioeconomic Status in Brazil

<b>Abstract</b>: The occupational specialization of social groups is closely tied to gender, racial, and class identities, segmenting the labor market into perceived White/Black and male/female roles and skills sets. Using data from 100 million formal workers in Brazil (2003–2019), we examine patterns of occupational segmentation across 426 occupations, identifying distinct skill demands and socioeconomic statuses linked to race/skin color and gender. Classifications of “male” or “female” occupations are shaped by required skills, while distinctions between “White” and “Black” occupations reflect socioeconomic status and historical inequalities. Women and men are segmented by gender-associated skill sets, such as engineering versus caregiving skills. Within these skill sets, strong hierarchical segregation persists, with Black individuals disproportionately concentrated in lower socioeconomic status positions. Despite recent socioeconomic changes, occupational specialization patterns have remained stable. Our findings highlight that the strong association between race and lower-status occupations must be addressed for a more inclusive society<b>Funding</b>: We acknowledge the financial support of the National Council for Scientific and Technological Development (CNPq), in particular processes 406943/2021-4 and 315441/2021-6.In case of any questions related with the content of this repository, please contact<b>:</b><br>Ben-Hur Cardoso ([e-mail])Laís Fernanda S. Souza ([e-mail])<br>Flavio L. Pinheiro ([e-mail])<br>Dominik Hartmann ([e-mail])ContentsThis repository contains the following contents:In the <b>Regressions</b> folder, we share the original regression tables supporting the robustness results shown in the Supplementary Material.The <b>Dataset</b> folder contains the minimum data necessary to reproduce all the results in the main manuscript and supplementary information.The <b>Code</b> folder contains two documents with the necessary code to reproduce all the results and visualizations in the main manuscript and supplementary information<b>Dataset</b> Folder DescriptionThe core datasets used in this study are:- <b>CENSUS_data_by_occupation_socialgroup_year.csv</b>: The Relative Specialization (RS) of each ISCO-08 occupation code in relation to its social group in each year, using Brazilian Census Data.<br>- <b>RAIS_data_by_occupation_socialgroup_year.csv</b>: The Relative Specialization (RS) of each ISCO-08 occupation code within social group each year, using RAIS Data.<br>-<b> RAIS_data_by_region_college_age_occupation_socialgroup_year.csv</b>: The Relative Specialization (RS) of each ISCO-08 occupation code with social group in each year, region, college, and age group, using RAIS Data.<br>-<b> RAIS_data_by_age_occupation_socialgroup_year.csv</b>: The Relative Specialization (RS) of each ISCO-08 occupation code within social group in each year and age group, using RAIS Data.<br>-<b> data_by_occupation.csv</b>: for each ISCO-08 occupation code we have<br>-- isco08_label_en: english label of occupation<br>-- phi_SX: the intensity of skill X<br>-- theta_SX: the specialization of skill X<br>-- isei: The ISEI of occupation<br>-- ISEIa: The regressed Adjusted ISEI<br>Skills X correspond to a single-digit (from 1 to 8) encoding that refers tocommunication, collaboration, and creativityinformation skillsassisting and caringmanagement skillsworking with computershandling and movingconstructingworking with machinery and specialized equipmentAdditional data files include:<br>- <b>isco08_data.csv</b>: extends the data_by_occupations.csv dataset with the RS by gender/race of each occupation<br>- <b>isco08_skill_similarity.csv</b>, <b>netskill.csv</b>, and <b>node_meta.csv</b> provide information on the skill similarity structure between occupations and meta information at the node level (occupation), compiled from the other datasets mentioned above.<br>- <b>Network_layout.gdf</b> encodes the network layout used to draw the networks.<b>Code</b> Folder DescriptionThis folder is composed of two primary documents:A Jupyter Notebook that contains all the code to generate the main visualizations of the manuscript and regression analysis.A Wolfram Mathematica notebook in which we perform the PCA analysis and generate the Graph/Network visualizations.These two notebooks read and process the shared datasets.