An Alternative Way of ComputingEfficient Instrumental VariableEstimators

A new way of constructing efficient semiparametric instrumental variableestimators is proposed. The method involves the combination of a large number ofpossibly inefficient estimators rather than combining the instruments into anoptimal instrument function. The consistency and asymptotic normality isestablished for a class of estimators that are linear combinations of a set ofv?? ??consistent estimators whose cardinality increases with sample size. It is shown thatthe semiparametrically efficient estimator lies in this class. The proofs do not relyon smoothness of underlying criterion functions. Potential use of the estimator canovercome the undersized sample problem. in simultaneous equation systemestimation.