A critique of the regression-based decomposition method: an illustration using the contribution of gender to income inequality
Date
2009-10
Authors
Anglim, Allyanna
Kua, Kesterson
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This paper quantifies the contribution of sex to income inequality by employing the recently-developed regression-based decomposition approach. In the process, we show that this decomposition procedure is flawed and could produce biased and incorrect results because of its inherent inability to completely account for the effects of other explanatory variables. This, coupled with defects in data, results in inconsistencies and the unreliability of the results of the decomposition. We argue, however, that general inferences about the relative importance and contribution trend of each explanatory variable can still be made. Other variables are also shown to contribute to income inequality, the most notable of which are infrastructure and education. We show that taken together, human capital variables explain a less than expected amount in the variation of income logarithms, but the importance of which are not negligible. Regional location is also shown to contribute a considerable amount to income inequality. We conclude that should better data be available for this particular study, and should the decomposition procedure be properly carried out in accordance with its theoretical purpose, the effect of sex on inequality is expected to be bigger and positive, which is in line with standard literature concerning the explanatory power of sex in the logarithms of income. We call for improvements in the methodology, especially in the ability to account for other explanatory variables in order to get the true, unbiased and absolute effect of each independent variable on the dependent variable.
Description
Keywords
Income equality, Gender inequality, Gender wage gap, Regression based decomposition, Labor market, Labor force