The 2010 Brazilian Census contains a wealth of information that could enable research and inform policies in health, education, the economy, and other sectors. The census provides publicly available information in two forms. Firstly, contingency tables are available at the municipal level, for strata defined by race, gender, and education. Secondly, microdata with personal information. To preserve individual anonymity in the data, the census collapsed some variables into broader categories and removed personally identifiable data. The data composition strategies of the contingency tables and the microdata are different and, when comparing samples of both data, we find that the race variable in the microdata ignores the presence of minorities in some municipalities. This suggests that synthetic populations based on the 2010 Census should be created using the contingency tables. Our evaluation shows that the so created synthetic population maintains the values and proportions of the contingency tables and presents totals close to those of the microdata.
Tavares et al. (Thu,) studied this question.