Why educated mothers don't make educated children?

More educated parents are observed to have better educated children. From a policy point of view, however, it is important to distinguish between causation and selection. Researchers trying to control for unobserved ability have found conflicting results: in most cases, they have found a strong positive paternal effect but a negligible maternal effect. In this paper, I evaluate the impact on the robustness of the estimates of the characteristics of the samples commonly used in this strand of research: samples of small size, with low variability in parental education, not randomly selected from the population. The part of the educational distribution involved in any identification strategy seems to be a key aspect to take into account to reconcile previous results from the literature.