Principal investigator: Bruno Benevit
Original title: Assortative Matching or Exclusionary Hiring? The Impact of Employment and Pay Policies on Racial Wage Differences in Brazil
Authors: François Gerard, Lorenzo Lagos, Edson Severnini, David Card
Location of the Intervention: Brazil
Sample Size: 22,62 million workers
Sector: Jobs
Primary Variable of Interest: Salary
Type of Intervention: Company policy
Methodology: OLS, IV
Summary
In economic literature, various mechanisms seek to explain the existence of income disparities. Among them, the way in which companies' hiring and salary policies impact pay inequalities has been gaining attention from researchers. In this sense, this study investigated how such policies explain the wage differences between whites and non-whites in Brazil. Additionally, the authors also investigated how workers' observable skills explain the observed distribution for high-paying jobs. The results indicated that non-white workers have a lower propensity to work in higher-paying companies, receive lower salary premiums, and that one-third of the selection for high-paying jobs is not explained by selective matching of workers' skills.
- Policy Problem
Race is one of the main variables that explain the wage level of workers in various countries, even when considering educational aspects and other observed variables. According to the theoretical framework proposed by Becker (1957), each worker faces a wage determined by the market.
Recently, several authors have sought to associate specific company differences with the occurrence of wage disparities. According to these authors, this dynamic could be explained by the market power of firms, given by a monopsonistic relationship in setting workers' wages. Thus, firms in sectors with higher demand for skilled workers set higher wages to attract them.
In this sense, wage inequality will depend on two factors: (i) the extent to which higher-wage firms employ whites and non-whites differently – a classification effect between firms – and (ii) the relative size of the wage premiums offered by a given firm to different racial groups – a relative wage setting effect.
Several studies present evidence of racial bias during the worker selection process, for both minorities and non-minorities (ÅSLUND; HENSVIK; SKANS, 2014; GIULIANO; LEVINE; LEONARD, 2009). Similarly, the literature indicates a significant underrepresentation of the non-white population in high-paying positions in the United States and Latin American countries (GERARD). et al..
- Policy Implementation Context
Brazil has a colonial history marked by African slavery, with three recognized racial groups: whites, mixed-race people, and blacks. The abolition of slavery in Brazil occurred late, in 1888, and the subsequent period was marked by the absence of legal segregation, unlike what occurred in the United States and South Africa (GERARD). et al..
Despite the absence of racial segregation policies and considerable racial mixing, socioeconomic inequality between whites and non-whites has always been a prominent feature of Brazilian society. Efforts by the Brazilian state to change this scenario began with the 1988 Constitution and the enactment of anti-racial discrimination laws between the late 80s and early 90s.
Even with the adoption of affirmative action policies and the implementation of the Racial Equality Statute at the beginning of this century, which brought greater attention to the problem of racial inequality in Brazil, the disparities between whites and non-whites in the country are still noticeable in various segments and comparable to those of other Latin American countries.
- Evaluation and Policy Details
According to data from the National Household Sample Survey (PNAD), the working population aged 25 to 54 is composed of 50% white, 42% mixed-race, and 8% black. In the Southeast, the most developed and populous region of the country, the proportion of white and mixed-race individuals is approximately 58% and 33%, respectively. In terms of employment rates, approximately 43% of men and 25% of women are employed in the non-agricultural private sector in Brazil. In the Southeast, this rate varies between 47% (white) and 51% (black) for men and between 23% (mixed-race) and 27% (white) for women. Regarding education, the racial difference between white and non-white men is approximately 1,6 years of schooling, and 1,25 years of schooling among women.
PNAD data also indicate that the wage gap between whites and non-whites ranges from 27% to 33% when considering only state and year effects, and from 11% to 13% when the experience and education of workers are also considered. This difference is higher in the Southeast region, and also when considering only the most educated workers.
The main analysis in this article used data from the Annual Social Information Report (RAIS), which provides longitudinal data on the characteristics of formal workers in Brazil. For the main analysis, data on workers aged 25 to 54 in southeastern Brazil, covering the period from 2002 to 2014, were considered.
The information in RAIS is collected from annual reports submitted by companies to the Ministry of Labor about all employees who were on the payroll in the previous year, including their hiring and termination dates, average monthly earnings during the year, monthly earnings in December, hours worked, age, gender, education, and race. The structure of the RAIS data is similar to that presented in the PNAD (National Household Sample Survey).
- Method
To assess the effect of specific firm hiring and wage-setting policies on wage inequality between whites and non-whites in Brazil, this study estimated two-way fixed-effects ordinary least squares (OLS) models to capture each firm's wage premium. To do this, a counterfactual exercise was performed assuming the absence of racial discrimination in hiring and wage setting (within the same company). The method presupposes the plausibility of the exogenous mobility condition.
The study's baseline model considers the theoretical framework proposed by Abowd, Kramarz, and Margolis (1999) to estimate the logarithm of the hourly wage of workers from a specific race-gender group and period. The model considers the invariant personal characteristics of the workers, the specific wage premium of each firm-period for a given race-gender group, and variable characteristics (e.g., fixed effects of time and experience). For the establishment of the bidirectional fixed-effects model, the sample used considers only observations in firms with workers from both groups.
To correct for potential bias arising from differences in network Among white and non-white workers, the authors estimate the fixed effect of workers in isolation, considering the effects of firms on this variable using OLS and instrumental variable (IV) methods. For IV, the firm effect was used as an instrument, considering the same gender and the opposite race group.
The study also presents a series of analyses decomposing the two-way fixed-effects model to identify specific sources and effects. The decomposition approaches seek to identify the impacts of each individual and firm's effects on wage inequality, comparing the results of the baseline model to the usual Mincerian wage determination model and verifying possible selective matching effects. Skills-based matching effects were also verified using the same counterfactual approach described above, and the size of these effects was checked for different quantities of the distribution of workers' personal effects.
Finally, the study also presents a series of robustness analyses verifying the adequacy of the models, altering their specifications, and verifying the effects found considering the subsample for the Northeast region of Brazil.
- Main results
The results of the main analysis indicated that the difference in wage premiums paid within the same firms explains approximately 20% of the variation in hourly wages for the four race-gender groups. In other words, the income associated with average firm premiums explains a higher percentage compared to non-whites, occurring for both men and women. Furthermore, it was identified that 5% to 6% of the racial wage gap is explained by firm-specific effects, i.e., wage differences practiced within the same company.
When assessing the reasons for the distribution of worker groups among firms, evidence was also found of a strong impact of selective skill matching for workers in all groups. Firms that pay 10% higher wages have workers who earn 5-8% more in any given workplace, explaining approximately 18% of the wage gap between whites and non-whites.
The results of the estimates considering the distributions of workers' skills indicate that about two-thirds of the effect of the distribution of groups among firms is explained by selective matching based on skills (12% of total inequality). The other third (6-8%) represents the residual portion not explained by skills, incorporating discriminatory hiring and retention policies. Additionally, non-white workers with higher skills experienced a greater penalty from the portion not explained by skills.
- Lessons in Public Policy
In this article, the authors examined how firms' wage and hiring policies influence wage inequality between whites and non-whites in Brazil. To this end, the authors examined how workers from different racial and gender groups are distributed among firms with different levels of wage premiums and whether firms establish their selection and compensation criteria using neutral criteria (education, experience, and personal skills).
The results of this article indicate that most of the wage gap between whites and non-whites in Brazil is explained by the educational level and personal skills of workers. However, one-third of the distribution of non-white workers among firms is not explained by these factors. Additionally, it was identified that the effect not explained by observable factors is greater for non-white workers with higher qualifications. This evidence highlights that a significant portion of wage inequality in Brazil can be associated with discriminatory practices by firms, reinforcing the need for policies aimed at reducing racial inequalities in the country.
References
ABOWD, JM; KRAMARZ, F.; MARGOLIS, DN High Wage Workers and High Wage Firms. Econometrics, v. 67, no. 2, p. 251–333, Mar. 1999.
ÅSLUND, O.; HENSVIK, L.; SKANS, ON Seeking Similarity: How Immigrants and Natives Manage in the Labor Market. Journal of Labor Economics, v. 32, no. 3, p. 405–441, Jul. 2014.
GERARD, F.; LAGOS, L.; SEVERNINI, E.; CARD, D. Assortative Matching or Exclusionary Hiring? The Impact of Employment and Pay Policies on Racial Wage Differences in Brazil. American Economic Review, v. 111, no. 10, p. 3418–3457, 1 Oct. 2021.
GIULIANO, L.; LEVINE, DI; LEONARD, J. Manager Race and the Race of New Hires. Journal of Labor Economics, v. 27, no. 4, p. 589–631, Oct. 2009.
