The gender diversity gap in mathematics

Abstract

This paper documents gender gaps in mathematics performance and confidence among Chilean 10th graders, providing new evidence on transgender and non-binary students using longitudinal data that follows students from 4th to 10th grade. Using rich administrative data, we estimate within-school models with lagged outcomes and propensity-score weights to ensure robust comparisons. The longitudinal structure allows us to control for baseline performance and track the evolution of outcomes, covariates, and school contexts over time. Relative to cisgender men, we find sizable test-score gaps for cisgender women and most gender-minority students. These gaps are highly distribution-dependent: they widen in the upper tail for students assigned female at birth, while for transgender women they concentrate in the lower tail and vanish among top performers. For non-binary students assigned male at birth, gaps are near zero across the distribution. More robustly, confidence gaps are large and persist even after conditioning on own test scores. Gender-minority students report much higher rates of discrimination and aggression; these exposures explain little of the score gap but account for up to one-third of the confidence gap for certain groups, operating mainly via differences in exposure rather than returns. Finally, school context seems to matter, as schools with formal religious curricula are associated with wider achievement gaps.