Marcia M Schafgans and Morton Stelcnery

Published December 2006

---

In this paper we revisit the gender decomposition of wages in the presence of selection bias. We show that when labor market participation decisions of couples are not independent, the sample selection corrections used in the literature have been incomplete (incorrect). We derive the appropriate sample selection corrections, based on a reduced form model for the joint participation decisions of both spouses. The influence that husbands’ participation decision has on the female participation decision also highlights the importance of using data on both spouses for the analysis of the gender wage gap. Taking account of these issues might influence the outcome of the decomposition analysis and affect the evidence of discrimination. We analyze its potential impact by analyzing the gender earnings differential using Canadian census data.