This paper examines the impact of family background on wages and returns to education. Using data from the GSOEP, we show that family background matters in the determination of wages. Moreover, returns to education appear to be heterogeneous with family background accounting for part of this heterogeneity. Estimated returns to education are higher for individuals from families with lower education levels. However, a simple regression analysis suffers from serious endogeneity problems. The construction of a sibling sample allows to control for unobserved characteristics that are shared by family members. Using a family fixed-effects estimation we attempt to reduce the endogeneity bias. Our results suggest that the conventional estimates overstate the returns to education. Moreover, family background accounts or a large part of the variation in wages in the sibling sample.