Working Paper: NBER ID: w10892
Authors: Casey B. Mulligan; Yona Rubinstein
Abstract: Rising wage inequality within-gender since 1975 has created the illusion of rising wage equality between genders. In the 1970's, women were relatively equal (to each other) in terms of their earnings potential, so that nonwage factors may have dominated female labor supply decisions and nonworking women actually had more earnings potential than working women. By 1990, wages had become unequal enough that they dominated nonwage factors, so that nonworking women tended to be the ones with less earnings potential, and the wage gap between workers and nonworkers was large. Accounting for the growing selection bias using both parametric and semi-parametric versions of the Roy model, we show how the earning power of the median woman has not caught up to the earning power of a median man, even while the earning power of the median working woman has. As an illustration, we give some attention to wives with advanced degrees -- they have high and stable labor force participation rates -- and show how their measured wages have grown at about the same rate as those of men with advanced degrees.
Keywords: No keywords provided
JEL Codes: J21; J31; J16; C31
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| rising wage inequality within genders (J31) | selection bias in the labor market (J79) |
| nonworking women in the 1970s had higher earnings potential than working women (J16) | shift in earnings potential by 1990 (J31) |
| wage inequality increases (J31) | wage distribution of working women diverges from that of nonworking women (J31) |
| average married woman's earnings potential has not increased relative to average married man's earnings potential (J31) | misleading closing of gender wage gap (J79) |
| wage growth of married women with advanced degrees mirrors that of male counterparts (J31) | does not reflect overall closing of gender gap (J16) |
| Heckman selection model (C24) | numerical estimates of selection bias growth over time (C51) |