Gender gaps in educational participation can mask inequality in what women and men study, with consequences for later education and employment. We study whether criminal political representation contributes to such sorting at India's higher-secondary level, where science is the principal route to tertiary STEM education. Using nationally representative surveys, candidate affidavits, election results, and school-availability data, we instrument criminal political representation using close contests between candidates who differ in accusation status. We find that criminal political representation widens female disadvantage in science choice but has no discernible effect on the enrolment gender gap. A one-standard-deviation increase in the share of criminally accused politicians widens the science gap by 5--7 percentage points, with girls shifting toward arts/humanities. The adverse effect is larger where fewer schools offer science. Criminal political representation also increases perceived harassment of unmarried girls, while prior harassment is associated with a wider female disadvantage in science choice. Political selection can thus alter the composition of female human capital even when educational participation is unchanged.