Identification information and instruments in linear econometric models with rational expectations

Abstract

This paper establishes a convenient set of criteria for the identification of linear simultaneous econometric models with rational expectations. We consider the model with one-period-ahead expectations conditional upon current exogenous or predetermined variables or their forecasts if current values are unavailable. A collection of instrument counting rules are developed which turn out to be identical with those for the standard textbook model, provided that we treat the expectational variables in the light of endogenous variates. This suggests that rationally formed expectational variables are best treated as a second set of endogenous variables. A general inverse relationship holds between information and identification: the less agents know about current values of exogenous variables the greater are the possibilities for statistical identification of the system describing their behavior.