Abstract
This paper introduces two methods to derive internationally comparable skills and occupational distance based on machine learning and natural language processing techniques. We apply these measures to produce descriptive facts about employment transitions and workers’ wage distribution in Brazil using the universe of formal labour contracts covering the period from 2003 to 2018. Our findings show that workers who use non-routine cognitive skills intensively are better off in terms of employment, wages, and occupational switching. Overall, we observe signs of a routine-biased technological change and employment polarisation, however, this trend has only emerged since the Brazilian economic crisis of 2014.