Regression-based decompositions have been used in the literature to try to identify the sources of wage gaps between men and women. These decompositions allow assessing how much of the gap is explained by observed gender differences in terms of individual productive characteristics, the remaining unexplained portion being ascribed to differences in unobserved characteristics and/or asymmetries in labour demand (see OECD, 2008a). Educational attainment and labour market experience typically explains only a small or even negligible portion of the gender wage gap. By contrast, labour market segmentation by occupation, type of contract, industry as well as firms and establishments typically explain a far larger share (see e.g. Altonji and Blank, 1999; Reilly and Wirjanto, 1999; Datta Gupta and Rothstein, 2005; Heinze and Wolf, 2006). However, evidence based on large-scale matched employer-employee data shows that even taking into account a fine disaggregation of occupations, industries and establishments, more than 50% of the wage gap remains unexplained (e.g. Bayard et al., 2003). More important, the gender distribution of jobs is itself the outcome of the equilibrium in the labour market. It provides therefore some indication of the channels through which a gender wage gap arises, but sheds no light on the ultimate causes of the gap.
Sunday, April 29, 2012
Are the Republicans Denying Gender Pay Disparity?
Meet the Press featured a debate between Rachel Maddow and Alex Castellanos. When Ms. Maddow noted a recent finding by the Institute for Women’s Policy Research noting that women earn 77 cents to every dollar earned by men, Mr. Castellanos first denied there was any disparity and then argued factors such as the choices when make about how much to work or what professions to enter explain all of any alleged difference. Really? What evidence did Mr. Castellanos offer for this claim? None apparently - even though there has been substantial research on this issue. This 2008 OECD chapter summarizes much of this research: