Heteroscedasticity-Robust Quantile Regression Decomposition with Endogenous Structural Breaks: A Bayesian Semiparametric Framework for Wage Gap Estimation in Transitional Economies

Abstract

This study addresses the persistent methodological limitations in wage gap decomposition analysis across transitional economies by proposing a Bayesian semiparametric quantile regression framework that simultaneously accounts for heteroscedasticity, endogenous structural breaks, and latent distributional heterogeneity. Employing panel microdata from six Central and Eastern European labor markets (2010–2023), the model integrates Markov Chain Monte Carlo sampling with Dirichlet process priors to recover counterfactual wage distributions at granular quantile intervals. Empirical results demonstrate that conventional Oaxaca–Blinder decompositions systematically underestimate the unexplained gender wage gap by 8.3–14.7 percentage points at upper-tail quantiles, primarily attributable to endogeneity bias in occupational sorting variables. The proposed framework yields superior out-of-sample predictive accuracy (CRPS reduction of 19.2%) and offers tractable posterior inference for policy-relevant decomposition components.