Dr. Dick Startz (Economics Department, UCSB)

Tittle: Instrumental Variable Estimation Under Weak Identification

Abstract: The instrumental variable (IV) estimator is the most important tool—after regression—used by econometricians. For some 20 years now, economists have realized that the standard distribution theory for IV estimation is extraordinarily misleading in weakly identified models. In this talk I review the standard approach to IV and some of the earlier work by colleagues and myself to identify the problem under weak identification. I then introduce work-in-progress on the use of Bayesian mixture models to produce estimators which are robust under both good and weak identification. Finally, I touch on several other estimation problems with mathematically similar difficulties..