This paper is concerned with (1) whether "unobserved" firms' effects affect gender disparity in earnings, (2) if they do, what is their mechanism of operation, (3) what are the components of gender disparity in earnings when we decompose it into the "within-firm gender disparity" and "disparity due to t the effects of the selection into and out of employing firms" by taking into account gender differences in human-capital as a mediating variable, and how large the effects of those components are, and (4) whether regularly employed women's selection of their employers through the decision to retain/leave a job indicates a tendency in favor of firms with more advantageous conditions regarding earning prospects.
This paper has obtained statistical evidence that women tend to select themselves out of firms which are less advantageous in earning prospects, mainly through their selective job retention/departure at the time of marriage and/or childrearing, through an analysis of firms' unique effects on individual earnings and their association with the gender of regular employees.. It has also shown that this selection effect becomes invisible due to its cancellation by an indirect effect that reflects men's relative advantage in human capital which leads to their employment in advantageous firms. By elimination of this indirect effect, this paper has shown that without the selection of their employers, gender disparity in earnings would have increased by about 5%.
In order to obtain those insights, this paper also made an innovative use of the DFL decomposition method of inequality by introducing a new method to handle firm heterogeneity.