Equilibrium and Identification in Linear Panel Models

Abstract

In this article we investigate an approach for simultaneously estimating cross-lagged and cross-instantaneous causal influences in two-variable, multiwave, linear panel models introduced by Greenberg et al. This approach requires observations over at least three time points and achieves identification by making assumptions about the consistency of parameter values across the adjacent time intervals 1-2 and 2-3. Our analysis demonstrates that multiwave models can be identified by imposing consistency constraints of this sort. However, this is a useful method only under a very restrictive set of empirical conditions.