Parents preferring sons tend to go on having more children until a boy is born and to concentrate investment in boys for a given number of children (sibsize). Thus, having a brother may affect a child’s education in two ways: an indirect effect by keeping sibsize lower and a direct rivalry effect where sibsize remains constant. We estimate the direct and indirect effects of a next brother on the first child’s education conditional on potential sibsize. We address endogenous sibsize using twins. We find new evidence of sibling rivalry and gender bias that cannot be detected by conventional methods.