The gender wage gap in Colombia and the role of education and other job characteristics

Gender Economics

During the period 1994-2010 the average observed gender wage gap in Colombia is significant and persistent: 13% in 1994 which slightly decreases to 12% by 20101. But the gap is not constant along the wage distribution, as Figure 1 depicts.

Figure 1: Observed Gender Wage Gap across the Wage Distribution of Workers

Observed Gender Wage Gap across the Wage Distribution of Workers

The figure reveals that for the two years, the gender wage gap follows a non-monotonic trend: it is very high at the bottom of the wage distribution, decreases around the middle and increases slightly for the highest wage percentiles. More precisely, the wage differential is around 54% (44%) among women and men who gain the 5% of the lowest wages in 1994 (2010), decreases at around 10% for middle earners and particularly for 2010 raises to 20% for highest wage earners.

With respect to differences in the gender wage gap among the two years, we can see that it is lower at the bottom of the wage distribution for 2010 with respect to 1994, although it decreases more slowly, whereas in 2010 the gap is sharper at the right end of the wage distribution. When accounting of the determinants of these differences, we must take into account that the Colombian labor market has gone through very significant compositional changes. The most relevant one is the change in the educational attainment of workers. In 1994 the percentage of college-educated workers was just 11%; however, by 2010 this figure had quadrupled. This increase has been more pronounced for female workers, whose proportion of college-educated workers was 8 percentage points higher than that of their male counterparts by 2010.

Together with changes in the educational level, some job characteristics have also experienced an impressive change during this period: the number of employees with indefinite contracts fell by half, mainly in the private sector. And the proportion of public sector workers also dropped sharply, from 13% in 1994 to 7% in 2010. Both changes have affected more women than men.

As these changes have not affected men and women equally, they undoubtedly have had an impact on the distribution of wages and on the gender wage gap. Therefore, we relate changes in the observed gender wage gap between 1994 and 2010 with observed changes in education, type of employment (public/private) and type of contract (indefinite/other type of contract) in Colombia. This may help predict future trends in the gender wage gap in Colombia, and therefore may help to design public policies aimed at decreasing the wage gap between men and women.

Using the methodology developed by DiNardo, Fortin, Lemieux (DFL) (1996), which is explained in a recent paper, we compute what the distribution of wages would have been in 2010 had the educational attainment, the proportion of public-sector workers, and the proportion of indefinite contracts remained at 1994 levels. The counterfactual distribution reveals the extent to which the gender wage gap has been affected by changes observed in each of the three characteristics and by all three of them together2.

Our main results:

Educational attainment has contributed to reduce the gender wage gap, on average in 4 percentage points and the greater impact is observed at the top of the distribution of wages, that is, among female and male workers with the highest wages. This means that if educational attainment had not occurred, especially for women, gender wage gap would be on average 16% instead of 12%; and 18% in percentile 75th instead of 4%.

With regards to labor characteristics, the decrease in the number of public employees has contributed to increase the gap, mainly at the top of the wage distribution. Therefore, if reduction in the size of state had not happened, on average gap would have disappeared and would favor especially women who gain higher wages.

On the other hand, the decrease in the number of workers with permanent contract has helped increased the gap in around 7 percentage points and in this case the greater impact is observed at the bottom of the distribution of wages, that is, affecting mainly women who gain the lowest wages.

Finally, given that these three changes have occurred during the period of reference simultaneously, it would be interesting to learn to what extent the three together have changed the female and male wage distributions, and hence the gender wage gap. Findings reveal that the effect of job characteristics has helped counter the positive effect of education.

So far this analysis has been performed for the observed gender wage gap, not taking into account the whole set of differences in observable personal and job characteristics of workers. In order to compare wages of men and women with similar characteristics, we also have computed the adjusted gender wage gap, which is the gender wage gap that we would observe among men and women who have the same demographic and job characteristics. Figure 2 presents the distribution of the Adjusted Gender Wage Gap for 1994 and 2010.

Figure 2: Adjusted Gender Wage Gap across the Wage Distribution of Workers

Adjusted Gender Wage Gap across the Wage Distribution of Workers

Figure 2 reveals that when we compare similar men and women, on average the gap falls to 8% and 9% in 1994 and 2010, which suggests that observable characteristics explain an important part of the observed gap. Moreover, figure 2 shows that unlike the observed gap depicted in Figure 1, the adjusted gap is monotonically increasing throughout the distribution of wages, revealing the so-called ceiling pattern (term introduced by Albrecht, Björklund and Vroman (2003)), which means that women’s wages fall further behind similar men´s wages at the top of the distribution than at the middle or at the bottom.

Additionally, we have computed the adjusted wage gap, controlling for sample selection of workers in sample, which means controlling for the fact that labor force participation has increased during the analysed period, especially for women in around 11 percentage points. Failure to account for this might bias the estimated adjusted wage gap.

In the line of previous studies, our results confirm that sample selection is an important issue to take into account when estimating wages in Colombia and the glass ceiling effect observed without control for non-randomness in the sample is stronger after controlling for sample selection.

1. The observed gender wage gap is computed as the difference between the average logarithms of hourly wages of male and female workers.

2. The data to perform this work were drawn from the Colombian Household Survey for the second quarters of 1994 and 2010, carried out by the Colombian National Statistical Department (DANE).


Authors contact information:

  • Luz Karime Abadía Alvarado: Departamento de Economía, Pontificia Universidad Javeriana, Bogotá, Colombia.
  • Sara de la Rica: Department of Foundations of Economic Analysis II, University of the Basque Country UPV/EHU, Bilbao, Spain.


Share this