Identical twins have been used to control for "ability" in efforts to obtai
n unbiased estimates of the earnings impact of schooling and of biases in e
stimates that do not control for earnings endowments. This study (1) presen
ts new estimates of schooling returns and of "ability" bias using a new twi
ns sample, (2) develops and applies a test of the significance of that bias
, and (3) demonstrates that there may be "ability" bias even if the genetic
ally-endowed component of ability does not affect schooling decisions direc
tly if this component of ability is correlated with other family characteri
stics such as income that do affect schooling. It is thus not possible to i
dentify the separate contributions of family constraints and individual abi
lity to "ability" bias. The basic empirical result is that, net of measurem
ent error, upward "ability" bias is statistically significant in OLS estima
tes, causing an overestimate of the schooling impact of 12%. [ JEL 121, J24
] (C) 1999 Elsevier Science Ltd. All rights reserved.