We exploit the random allocation to a semi-external assessment in Math (SEAM) at the end of high school in Denmark to test the effect of SEAM on subsequent enrollment and graduation in post-secondary education. We find that SEAM in high school reduces the gender gap in graduation from post-secondary STEM degrees, and we discuss possible mechanisms. Our results show that cancelling external assessments, as was temporarily implemented in many regions during the COVID-19 pandemic, may impact gender differences in human capital accumulation in the long run.