Published:
Directly comparing postgraduate earnings of workers with bachelor’s degrees and those with only a high school diploma overstates the earnings gains of the college graduates. This bias, called selection bias, stems from the self-selection of high school graduates into college. This study explores the earnings premiums of postsecondary awards and degrees in Washington state, but corrects for this selection bias using a statistical method called propensity score matching. This method compares the earnings of those who received postsecondary awards and degrees with high school graduates who received no degree, but who were comparable in many other measurable ways.
Among other things, we discovered:
- While completing a postsecondary credential or degree generally lead to higher annual real earnings, females consistently earned less than men, regardless of educational achievement.
- The increase in earnings associated with each postsecondary degree type differs between males and females (with respect to their comparison groups).
- While the earnings gender deficit seems to grow with educational attainment, female earnings as a percentage of male earnings remains fairly static.
- The female earnings premium exceeds the male earnings premium for short- and long-term certificates, but was smaller than the male premium for the associate, bachelor’s and graduate degrees.
- For most degree types, the female earnings premium decreased over time (with respect to the comparison group) while the male earnings premium did not.
- The gender deficit for workers who earned higher degrees grew over time, while the gender deficit for comparable high school graduates (who earned no degree) decreased.
- The hours worked by male and female bachelor’s degree holders became more similar over the six years, even as their earnings diverged.
- For bachelor’s degrees, the major categories in which females earned the most had a smallest percentage of females; the major categories with the highest percentage of females earned comparatively less.
