We study the effect of preferences for boys on the performance in mathematics of girls, using evidence from two different data sources. In our first set of results, we identify families with a preference for boys by using fertility stopping rules in a large population of households whose children attend public schools in Florida. Girls growing up in a boy-biased family score on average 3 percentage points lower on math exams when compared to girls raised in other types of families. In our second set of results we find similar strong effects when we study the correlations between girls' performance in mathematics and maternal gender role attitudes, using evidence from the National Longitudinal Survey of Youth. We conclude that socialization at home can explain a non-trivial part of the observed gender disparities in mathematics performance and document that maternal gender attitudes correlate with those of their children. [The Florida Department of Education and Health provided the linked population-level administrative data that permitted this analysis to take place.]