Using census data from Nepal we examine how the partial derivatives of predicted household welfare vary with parental education. We focus on fertility, child survival, schooling, and child labor. Female education is not as strongly associated with beneficial outcomes as is often assumed. Male education often matters more, and part of the association between female education and welfare is driven by marriage market matching with more educated men. Controlling for the average education of parental cohorts does not change this finding. But when we use educational rank to proxy for unobserved ability and family background, the positive association between female education and beneficial outcomes becomes weaker or is reversed. For women, the association between educational rank and outcomes is strong: women who obtain more schooling than their peers in school have fewer children and educate them better. In contrast, for men the statistical association between education and household welfare remains strong, even after we control for educational rank within their birth cohort.