The marginal and average returns to schooling in the UK

We explore the possibility that different instruments may affect different margins of the schooling distribution, and consistent with recent evidence we find a large, significant downward bias in the least squares estimate of the returns to schooling and that the estimates are stable with respect to the choice of instruments. However when we remove the assumption that log wages are linear in years of schooling we find sensitivity to the choice of instrument. (C) 1999 Elsevier Science B.V. All rights reserved. JEL classi fication: J31; I21; J24.