Gender Income Gap among White-Collar Regular Workers: A clarification of the determinants that explain an 80% of the gap, and its mechanism

A decomposition of the gender gap in earned income among white-collar regular employees in Japan was conducted using the DiNardo-Fortin-Lemieux method based on propensity-score weighting. The results show that gender differences in three human-capital variables, namely, education, age, and the employment duration, explain 35% of the gender gap, and gender differences in occupation, hours of work, and positional rank explain an additional 43%, thereby explaining 78% of the gender gap by the six variables. The variable that had the greatest unique explanatory power was gender difference in positional rank. The study also conducted diagnoses about the validity of propensity-score estimates and also demonstrated that the alternative estimate for the unexplained gender gap based on the linear regression model had considerable bias. By considering a counterfactual situation where women had the same distribution of human-capital variables as men, and another counterfactual situation that women's distribution of positional ranks as well as human-capital variables became the same as men's, and analyzing how the unexplained gender gap in earned income in each of those counterfactual situations varies with variables, this study obtained the following findings. 1) The tendency for the gender gap to increase with age can be explained almost completely as a result of a widening gender gap in positional rank with age. 2) Compared with other women, female college graduates can expect a higher rate of reduction in the gender gap by reducing the rate of job separations at the time of childrearing and thereby increasing employment duration. 3) Among clerical workers, an equalization of human capital and positional rank between men and women will not lead to as much a reduction in the gender gap as that among people with other white-collar occupations. 4) Gender equality of opportunity in earned income is realized considerably better among those with middle-rank managerial/administrative positions or above than among people with lower positions.