Using a large-scale novel panel dataset (2005–14) on schools from the Indian state of Assam, we test for the impact of violent conflict on female students' enrollment rates. We find that a doubling of average killings in a district-year leads to a 13 per cent drop in girls' enrollment rate with school fixed effects. Additionally, results remain similar when using an alternative definition of conflict from a different dataset. Gender differential responses are more negative for lower grades, rural schools, poorer districts, and for schools run by local and private unaided bodies.