The authors model schooling careers in the Dutch educational systems (including the decision to drop out) as a discrete sequential choice, in which expected earnings play an important role. Selectivity bias from correlation between the errors in the decision function and the earnings function appears empirically relevant and they correct for it by applying a proper maximum likelihood estimation procedure. The results indicate that drop-outs behave differently from graduates. In the decision functions on schooling careers, expected earnings (based on estimated earnings functions in mid-career) are never significant. Yet, the educational choices appear ex post consistent with choices according to comparative (earnings) advantage.