Self-efficacy and interest in STEM subjects as predictors of the STEM gender gap in the US : the role of unobserved heterogeneity

Abstract

We used HSLS:09 data to investigate how self-efficacy perceptions and interest in STEM subjects in high school predict subsequent STEM field participation. We focused on the role of unobserved heterogeneity in shaping gender differences in the predictive power of STEM-specific self-beliefs and interest. We used an endogenous treatment effects approach. We found that: (a) unobserved attributes play an important role, and the findings are distinctly different between genders; (b) boys’ field of study choices are driven by interest in STEM subjects, while for girls it is early development of confidence in their math abilities; (c) interventions aimed at closing the STEM gender gap can work only for girls with above average confidence in their ability to do well in STEM subjects.