Specification versus data fitting: SEM prediction and the Q-class estimator

We propose a new class of limited information estimators built upon an expl icit trade-off between data fitting and a priori model specification. The e stimators offer the researcher a continuum of estimators that range from an extreme emphasis on data fitting and robust reduced-form estimation to the other extreme of exact model specification and efficient estimation. The a pproach used to generate the estimators illustrates why ULS often outperfor ms 2SLS-PRRF even in the context of a correctly specified model, provides a new interpretation of 2SLS, and integrates Wonnacott and Wonnacott's (1970 ) least weighted variance estimators with other techniques. We apply the ne w class of estimators to Klein's Model I and generate forecasts. We find fo r this example that an emphasis on specification (as opposed to data fittin g) produces better out-of-sample predictions. Copyright (C) 1999 John Wiley & Sons, Ltd.