Simulating bias in the estimator of labor market discrimination

In this study, the statistical bias in the most widely used estimator of labor market discrimination is investigated. An algebraic expression for the bias is derived and then simulated for a selected set of explanatory variables and model parameters. The results for a large national dataset seem to indicate that when the variables years of schooling and labor market experience are used in earnings functions, the estimator tends to underestimate the actual or true amount of labor market discrimination.