Gender and Salaries of the Qualified
Workforce in Brazil and Mexico
Maria Cristina Cacciamali and Fábio Tatei

The databases used to develop this research include the National Household Sample Survey (pnad), produced by the Brazilian Geography and Statistics Institute, and the National Occupation and Employment Survey (enoe), created by the Mexican National Statistics and Geography Institute (inegi). Both obtain information using procedures recognized by the international statistics system and allow for the construction of comparable analytical categories to estimate the effects of discrimination on the qualified labor market. Using micro-data, this study built the category of “employed” and subjected it to three filters: private urban sector not including domestic services, age group above 20 years and complete higher education. Then the data was subjected to an additional filter for those with complete higher education and employed in a job that requires university training, such as doctors, dentists, lawyers and others. The criteria used to select the employed population with higher education included the number of years of schooling: 15 for Brazil and 16 for Mexico.3 The sample selection of employed persons with higher education amounted to 6,058,234 people in Brazil, of which 3,032,082 were men and 3,026,152 were women. In Mexico the sample size was 4,877,758 people, with 2,896,832 men and 1,980,926 women. The sample for executive employment, and higher education professionals and technicians was made up of 4,006,079 observations in Brazil – 2,006,493 men and 1,999,586 women – and 1,811,963 observations in Mexico, with 1,120,640 men and 691,323 women. The discrimination analysis follows the assumptions of conventional economic theory; in other words, it is possible to estimate a person’s income based on his individual and economic characteristics. In this way, we used Mincer’s salary estimate equation (1974):

, where: W is a vector that represents income per hour of an individual’s work, β is the coefficient vector, X is the vector for the individual’s characteristics and ε is the error. In our model, the vector X is made up of the following variables:

  • personal characteristics : years of study, approximate experience by age and age squared, gender.
  • regional insertion: geographical macro-regions; 5 regions considered for Brazil – North, Northeast, Southeast, South and Mid-West – and 8 for Mexico – Northeast, Northwest, West, East, Central-North, Central-South, Southeast and Southwest. The category “Southeast” was the comparison for the two countries.
  • economic characteristics: For Brazil, the following was used: executives, professionals of the sciences and arts, mid-level technicians, administrative service workers, service employees, commercial service providers and salespeople and workers in the production of goods and services and repair and maintenance. For Mexico:professionals, technicians and workers in the arts, educators, officials and executives, office workers, industrial workers, artisan workers and assistants, commercial workers, transport operators, personal services employees and protection and surveillance workers. The category “professionals of the arts and sciences” was the base of comparison for the two countries.

The study breaking down income was carried out using the model proposed by Ronald Oaxaca (1973), and similarly set forth by Alan Blinder (1973). Initially, discrimination is given by the following relationship:


where the term represents the relationship between observed salary for women and men and is the same relationship in the absence of any discrimination. When the term is unknown, the estimate of discrimination is the same estimate of said term. In this way, in the absence of discrimination, men and women face the same salary structure, and discrimination thus produces a situation where the human capital earnings of the non-discriminated person are overvalued, or those of the discriminated person are undervalued. Using the least squares method (lsm), the salary estimate is given by:


3 For Brazil, the cut was made using specific variables from the pnad (v4745), which indicates the highest level of education for a person in the study. For Mexico, this direct information does not exist, so a method was created to build a compatible variable.