We consider the estimation of the treatment effect of education on earnings by IV, while maintaining the hypothesis that the data are generated by a dynamic life-cycle skill accumulation model. We provide an answer to the following question: would the econometrician prefer an instrument generated by the implementation of compulsory schooling regulation (a form of "dictatorial policy intervention"), or one generated by an educations subsidy (an "incentive-based" policy reform). Because the identifying moment condition can be decomposed in two mutually exclusive ex-post conditions that are based on the complying status; namely Post-Treatment Choice Orthogonality and Post-Treatment Choice Opposite Selection, we can introduce both the identity of those affected (and unaffected) and the distinctive nature of a policy into the analysis. The policy design exercise requires two elements; to adjust the identity of those affected (the complier/non-complier heterogeneity gap) to the con icting effects of schooling on post-schooling skill formation and on its opportunity cost, and to keep track of the statistical strength of the instrument. Because of its relative degree of exibility, education subsidies are capable of generating a reliable instruments. However, compulsory schooling regulations, which basically annihilate comparative advantages for a non-random subset of the population, offer practically no design possibilities. Our calibrated model sheds light on the incidence of the low (and possibly negative) IV estimates that have recently been reported in the recently emerging literature on compulsory schooling. |