Automating the Selection of Model-Implied Instrumental Variables

Abstract

Recently, interest has grown in the use of instrumental variables (IVs) in estimating factor analysis and latent variable models such as structural equations models. Bollen (1996) suggested a two-stage least squares (2SLS) technique that makes use of model-implied IVs in estimating the measurement and latent variable models. Model-implied instrumental variables are the observed variables in the model that can serve as instrumental variables in a given equation. One difficulty inhibiting the practical use of the 2SLS estimator is identifying the model-implied IVs. The authors provide a simple procedure that identifies the model-implied IVs and a computer algorithm that can easily be implemented to automate the selection of IVs for simultaneous equations, factor analysis, and latent variable models.