Decomposition of Gender Income Differentials

The aim of this article is to explain the decomposition technique of income differentials proposed by Oaxaca and Blinder (1973), the methodology commonly used in Economics to analyze the causes of income differentials by gender, ethnicity, race, etc. The article is organized as follows:
 * Review of causes of gender income differentials.
 * Presentation of methodology of decomposition of income differentials proposed by Oaxaca and Blinder.
 * Limitations of Oaxaca-Blinder methodology.

Review of causes of gender income differentials
Empirically, the existence of gender income differentials has different reasons. Some of them are briefly described below. First, education and experience in the labor market are important determinants of productivity and, therefore, of labor income. This is the theory of human capital. In the case of gender income differentials, women are paid less than male workers because women tend to accumulate less education and work experience, in comparison with men. Since women often take more responsibility for family and household work, women’s labour force participation is often interrupted and women more often work part time. On the other hand, income differences can be explained by occupational segregation, in other words, women and men are employed in different sectors, occupations and jobs. In most cases women are concentrated in lowest paying jobs. Moreover, the theory of discrimination is a main determinant of gender income differentials, which means that women are paid less than equally productive men are. In the decomposition of gender income differentials, discrimination is the component that is not explained by observable characteristics.

Decomposition of income differentials proposed by Oaxaca and Blinder
The Oaxaca-Blinder decomposition technique is commonly used to identify and quantify the separate contributions of group differences in observable characteristics, such as education, experience, and in unobservable characteristics, such as discrimination. The initial step is to formulate a mincerian equation for the natural logarithm of income of men and women:
 * lnwi=βiχi+μi (1)

The matrix X contains the variables that affect the productivity of workers and thus their income. The vector μ is a random error term assumed normally distributed with zero mean and variance σ2.


 * lnwi=β(2)

The income differential is obtained by decomposing the natural logarithm of hourly income (w) for men (i = m) and women (i = f) estimating a mincerian equation for each group and subtracting one from the other:


 * (lnwm-lnwf) = (βmχm + βfχf) + (μm-μf) (2)

Then, a counterfactual term is constructed to indicate what the income that women would get is if they have the same remuneration as men (βmχm), and then this term is added and is subtracted on the right side of equation 2. Thus, in the model, the income differential attributed to the difference in observable characteristics (χm-χf)βm(called characteristic effect)and the income differential attributed to the difference in remunerations by gender χf(βm-βf, called coefficient effect, are obtained:


 * (lnwm-lnwf) = χm-χf+χf(βm-βf (3)

Limitations of Oaxaca-Blinder methodology
This decomposition technique has the following limitations:
 * The equations produced in the model evaluate the gender income differentials at the mean male and mean female income. The problem is that these equations cannot be used to analyze different discrimination levels in the different percentiles in the income distribution.
 * The choice of reference group in the model affects the results produced by the decomposition.
 * The technique uses a linear model (log-linear).