On learning and rational expectations in an overlapping generations model (Q1823124)

This paper presents an overlapping generations model with first money, with which to study whether learning actually leads to rational expectations. In this model, expectations are formed according to a family of expectations functions indexed by a single parameter. The parameter is updated each period through learning. The paper shows that unless agents have rational expectations (perfect foresight) to begin with, the expectations scheme will almost never converge to the rational expectations equilibrium. According to the authors, the reason for the lack of convergence to the rational expectations model is that the expectations formation function is estimated using only a fixed and finite number of observations at a time.