'Using information on family background, we estimate returns to education, allowing for the heterogeneity of returns. In order to control for the unobserved heterogeneity shared by family members, we construct a siblings sample and employ family fixed-effects and family correlated random-effects models. Our main result is that family background still matters despite the substantial political efforts to equalize educational opportunities in Germany. Persons with less-educated parents earn lower wages, but have higher returns to education. This supports the view that persons from less-educated backgrounds still face higher marginal costs in the educational system. The same interplay between the wage level and marginal returns is found for the...