This study uses a discordant sibling-based design to assess the extent to which education is proxying for pre-adult experiences and predispositions rooted in the family. It draws on a unique data set that combines official voting records with Census data on siblings and their parents. The results show that the association between education and voting is considerably reduced when parental education, parental voting and unobserved characteristics that are shared by siblings within the same family are taken into account. This finding is confirmed by a variety of robustness checks. We end with a discussion of the benefits and limitations of sibling-based designs for testing causal hypotheses.