Search
Program Calendar
Browse By Day
Browse By Time
Browse By Person
Browse By Session Type
Personal Schedule
Sign In
Access for All
Exhibit Hall
Hotels
WiFi
Search Tips
This paper examines how gendered sorting and heterogeneous major-specific returns jointly shape gender differences in STEM wage premium in the United States over the past two decades. Using the 2003, 2013, and 2023 waves of the National Survey of College Graduates, I estimate major-specific potential wages with a double machine-learning framework and decompose gender gaps in the STEM wage premium using a novel heterogeneity decomposition method. The approach separates gender differences in within-major STEM returns from gendered assignment across STEM subfields. The results show that women consistently have, on average, higher wage returns to STEM than men conditional on STEM major, indicating that the STEM premium is, in principle, larger for women. However, the internal wage structure of STEM becomes increasingly stratified over time, with computer science and engineering majors pulling away from biological, environmental, and agricultural fields. As this hierarchy sharpens, gendered sorting across STEM majors becomes the dominant mechanism generating gender differences in the STEM premium. Women’s growing concentration in the lowest-return STEM fields reverses their initial advantage: the female-male difference in the STEM effect on wages is positive in 2003, near zero in 2013, and negative by 2023, despite women’s stronger within-major returns. These findings reconceptualize STEM as an evolving, stratified opportunity structure in which gender inequality arises not primarily from unequal pay within fields, but from unequal access to rapidly appreciating ones. As technological change reshapes the economic hierarchy of majors, stable patterns of gender segregation can generate widening inequality—even without increasing discrimination.