Adaptive Learning and Business Cycles

This paper analyzes the quantitative importance of adaptive learning in business cycle fluctuations. We first introduce adaptive learning in a real business cycle model and a New Keynesian model, using specifications drawn from the literature which assume that agents learn about the equilibrium laws of motion. We consider a variety of learning rules, and find that in both environments learning has very minor effects on the volatility and the persistence of the key economic variables. However we discuss some potential theoretical drawbacks to this formulation of learning, and consider an alternative formulation in which agents learn about the structural features of the economy. In some simplified settings, we show that structural learning has much greater effects. We also illustrate how learning with misspecified beliefs can lead to fluctuations of a different kind, as agents “escape” from an equilibrium. Overall, our results show that the importance of learning depends greatly on the specification of beliefs.