Estimation of the Effects of Statistical Discrimination on the Gender Wage Gap

How much of the gender wage gap can be attributed to statistical discrimination? Applying an employer learning model and Instrumental Variable (IV) estimation strategy to Japanese panel data, I examine how women's generally weak labor force attachment affects wages when employers cannot easily observe an individual's labor force intentions. To overcome endogeneity issues, I use survey information on individual workers' intentions to continue working after having children and Japanese panel data with exogenous variation in average quit rates for female workers. I find that the extent of statistical discrimination is greatest for young age cohorts, ages 24 to 35, and that it diminishes for older cohorts. I also find that if employers could observe an individual's labor force intentions, the gender wage gap could be reduced from 17% to 5% for workers aged 24 to 29.