Learning in a Misspecified Multivariate Self-referential Linear Stochastic Model

This paper introduces a general method to study stability (under learning) of equilibria resulting from agents with misspecified perceptions of the law of motion of the economy. This is done by transforming the actual and perceived laws of motion into the form of seemingly unrelated regressions and then linearly projecting the actual law of motion into the same class as the perceived law of motion. I study the New Keynesian IS-LM model with inertia under all possible classes of restricted perceptions. It turns out that the results found in Bullard and Mitra (2002, 2003) are robust under misspeci.ed expectations.