Robust decomposition analysis of wage differentials

Article type: Research Article

Authors: Keith, Kristen^{a; *} | LeSage, James P.^a

Affiliations: [a] Department of Economics, University of Toledo, 2801 W. Bancroft Street, Toledo, OH 43606, USA. Tel.: +1 419 530 4754; Fax: +1 419 530 7844; E-mail: [email protected]

Correspondence: [*] Corresponding author

Abstract: We propose a robust Bayesian approach to the analysis of wage differentials based on the standard wage decomposition model. Although this model has been widely used to study wage discrimination based on race or gender, little attention has been given to the impact of aberrant observations (outliers) or heteroscedasticity on the resulting inferences regarding the discrimination effect. Results from Monte Carlo experiments demonstrate the advantages of our approach in the presence of outliers and heteroscedasticity. Simulation examples provide a comparison of estimates and inferences from the robust Bayesian approach with those from more traditional least-squares methods of wage decomposition analysis.

DOI: 10.3233/JEM-2004-0238

Journal: Journal of Economic and Social Measurement, vol. 29, no. 4, pp. 487-505, 2004