We investigate whether omitted family background variables are responsible for high returns to schooling estimated in Brazil. Returns to schooling fall by about one-third when parental schooling is added to wage equations. Surprisingly, the schooling of fathers-in-law has larger effects on wages than the schooling of fathers. On the basis of a model of assortative mating, we interpret this as evidence that parental characteristics represent unobservable worker attributes rather than nepotism in the labor market. We conclude that the 'family background bias' in returns to schooling is modest and need not imply returns to family connections.