When good education is free, college admissions are transparent and performance based, one might expect there to be little gender bias in placement at the university level. Yet, the college major choice decisions of students vary considerably by gender. Using Turkish data, we examine what lies behind these differences. Two channels seem to dominate: performance differences by gender and differences in preferences across majors. We then estimate a model of preferences and run counter-factual simulations to evaluate the role of these two channels on the placement gender gap. Finally we show that policy measures, such as giving women preference in STEM subjects, will not work as well as expected and show that more directed policies are needed.