Texto para Discussão, Instituto de Pesquisa Econômica Aplicada (IPEA) 2229
This paper presents a method for classifying the ancestry of Brazilian surnames based on historical sources. The information obtained forms the basis for applying fuzzy matching and machine learning classification algorithms to more than 46 million workers in five categories: Iberian, Italian, Japanese, German and East European. The vast majority (96.4%) of the single surnames were identified using a fuzzy matching and the rest using a method proposed by Cavnar and Trenkle (1994). A comparison of the results of the procedures with data on foreigners in the 1920 Census and with the geographic distribution of non-Iberian surnames underscores the accuracy of the procedure.