The impossibility of using highest achieved level of education to estimate the probability of passing a transition unconditional on having passed previous transitions

Abstract

A common model for estimating the inequality of educational opportunity estimates for each transition the effect of social background on passing the transition given that one is at risk. The resulting estimates are conditional on being at risk. Sometimes one wants the unconditional effects. In many models this can be achieved by controlling for all observed and unobserved variables. The purpose of this note is to show that even if one could estimate such models one would still not get the effects of social background unconditional on being at risk.