Brazil and the USA: Similarities in Workforce Dynamics

 

Wage inequality decreased in 15 of the 16 Latin American countries during the initial decade of the 20th century

Brazil serves as an illustration: the Gini index decreased from 0.48 to 0.44 between 2002 and 2015 (Messina and Silva 2021; Almeida, Ehrl, and Moreira 2021). However, the diminished urban, gender, and racial wage gaps were predominantly the result of one-time level effects that were difficult to replicate (Ferreira, Firpo, and Messina 2017; Firpo and Portella 2019). Moreover, income redistribution programs that were more straightforward to implement and enjoyed widespread societal support had already been established during a period of robust economic growth. Due to this, Holland and Schneider (2017) have contended that the "easy" period of income redistribution in Latin America may have concluded. If this is the case, the political and labor market obstacles to reducing inequality will be even more formidable. The current outlook for Brazil is anything but optimistic. Initially, the nation experienced a historic economic recession prior to the Covid-19 pandemic and has yet to regain its 2013 GDP level. Jaimovich and Siu (2020), Hershbein and Kahn (2018), and Graetz and Michaels (2017) have all demonstrated that a recovery may not inevitably result in the restoration of lost jobs, despite the uncertainty surrounding the timing of the Brazilian economy's return to growth. These papers suggest that firms restructure during a decline, substituting routine occupations with machines or digital technologies, in accordance with the hypothesis that technological change is skill-biased.

The employment and wage fluctuations in the formal Brazilian labor market from 2003 to 2018 are the subject of this paper. We are particularly intrigued by the correlation between the skills of laborers and the process of transitioning into a new job. Our objective is to enhance our comprehension of the labor market dynamics in a middle-income country with respect to secular (technology-related) changes and economic crises. 1 The current paper will investigate whether the patterns of job transitions, skill-biased technological change, and polarization that are observed in rich countries are also prevalent in Brazil, as these phenomena are primarily studied in developed countries. Aggregated statistics concerning employment transitions and changes are presented as an initial measure. Consequently, worker-level regressions demonstrate the correlation between skills and changes in occupation, sector, wage, and labor market exit risk. In order to address these inquiries, we utilize the RAIS (Relação Anual de Informações Sociais), an administrative registry that contains all formal employment contracts between 2002 and 2019. Upon selecting a 10% random sample from this database, we are left with more than 50 million observations of 6.4 million men and women from all sectors.

We present two innovative approaches to quantifying occupational distance and skills that are derived from Natural Language Processing (NLP) and Machine Learning.

The initial method produces skill intensity scores that are: (a) highly comparable across countries; (b) require minimal data; (c) do not depend on cross-country occupation matching; and (d) replicate an international standard. Autor, Levy, and Murnane (2003) introduced the skill categories of non-routine cognitive (NRC), routine cognitive (RC), non-routine manual (NRM), and routine manual (RM), which we employ frequently. In contrast, the existing approaches can be categorized into three groups: those that are derived from a combination of the tasks originating from the United States and an occupational crosswalk (Goos, Manning and Salomons 2014; Arellano-Bover 2020; Ehrl and Monasterio 2021); a discrete classification based on an ad hoc classification of a few occupations (Bachmann, Cim and Green, 2019); or a continuous measure derived from the ad hoc classification of tasks as shares in occupations in the spirit of Spitz-Oener (2006). The primary issue with our study is that the methodologies are not suitable for low- and middle-income countries, despite the fact that they each have their own strengths and weaknesses. Our approach offers a continuous skill assessment that is fundamentally predicated on the task descriptions of occupations in the country in question, specifically the Brazilian Classification of Occupations (Classificação Brasileira de Ocupações, CBO). Our method replicates existing skill measure definitions, rather than renaming them, which is a distinctive feature. Consequently, this straightforward procedure is easily transferable to any other nation.

The second methodological contribution is the calculation of occupational distance, which is the degree of divergence between any two occupations in terms of the content of their tasks. This variable is subsequently employed to assess occupational changes in the context of productivity losses and human capital, as was previously done by Gathmann and Schönberg (2010). Once more, this method generates country-specific measures and can be implemented whenever a national occupational dictionary with task descriptions is accessible. As anticipated, cognitive skill scores increase as the wage distribution progresses, while manual skill scores decrease. Additionally, additional metrics indicate that our skill measures are consistent with evidence from other countries. In general, the findings verify that employment transitions are associated with the capabilities of workers. The utilization of moderator variables in our regression analysis demonstrates a significant degree of heterogeneity in the skills of workers in relation to their age, tenure, and firm size. In particular, we have observed that the likelihood of occupational and sectoral switching is comparatively low among workers who employ NRC skills extensively. The risk of exiting the labor market is relatively low for NRC workers, and they are more likely to secure nearby, higher-paying positions in the event of an occupational change. In terms of these attributes, it is evident that RM and NRM employees are in a worse position.

In general, there have been indications of employment polarization and routine-biased technological change (RBTC) since the 2014 Brazilian economic crisis.

The manufacturing sector's middle-skilled occupations experienced the most significant relative employment losses during the most recent recession, which spanned from 2013 to 2018. Additionally, employment gains are observed at the lower and upper ends of the distribution when occupations are ranked by their mean wage, while occupations in the middle of the wage distribution experience employment losses. Additionally, there is a discernible trend among middle-wage employees to employ an increasing number of NRC skills. Additionally, RM specialized workers appear to be more susceptible than NRM workers, which is consistent with the RBTC and automation hypothesis. For instance, when working in a different sector or occupation, RM employees are inclined to relocate to more distant occupations. RM workers also exhibit the lowest transition rates to occupations that rely on distinct, higher-paid skill types and the lowest direct job-to-retirement transition probability.