Previous studies of the effect of school entry age on educational attainment may be severely biased because they violate the monotonicity assumption needed for LATE. We propose an instrument not subject to this bias and show no effect on children born in the fourth quarter of moving from a December 31 to an earlier cutoff. To determine whether the IV estimates diverge because of bias or because they estimate different LATEsm we estimate a structural model of optimal entry age that reconciles the different IV estimates. Our estimates imply that one standard instrument is badly biased but that the other diverges from ours because it estimates a different LATE. We also find that an early entry age cutoff that is applied loosely (as in the 1950s) is beneficial but one that is strictly enforced is not.
Joint paper with Rashmi Barua (Singapore Management University), October 2008.