This paper presents two optimizing models of educational choice, discusses
issues of identification, estimates earnings equations in the context of th
ese models, and presents conditions under which he can test one against the
other. The estimates indicate that education is endogenous for young peopl
e's earnings, creating a downward bias in estimated returns from education
that assume exogeneity. Identification and estimation relies on family back
ground information from a special sample from the British Household Panel S
tudy 1991-95, which matches mothers and their young adult children. Our est
imates favor a family model over an individual model, and they suggest that
parents allocate resources to education to compensate for differences in t
heir children's earnings endowments.