Two-sided Learning in New Keynesian Models: Dynamics, (Lack of) Convergence and the Value of Information

This paper investigates the role of learning by both private agents and the central bank (two-sided learning) in a New Keynesian framework populated by private agents and a central bank that have asymmetric imperfect knowledge about the true data generating process. We assume that all agents employ the data they observe (which can be different for different sets of agents) to form beliefs about the aspects of the true model of the economy that they do not know, use these beliefs to decide on actions, and revise beliefs through a statistical learning algorithm as new information becomes available. We study the short-run dynamics of the model and policy recommendations coming out of our model, in particular concerning central bank communication. Two-sided learning can generate large increases in volatility and persistence, and can alter the behavior of the variables in the model in a significant way. We also show that our model does not converge to a symmetric rational expectations equilibrium and highlight one source that disables the convergence results of Marcet & Sargent (1989). Finally, we identify a novel aspect of central bank communication in models of learning: communication can be harmful if the central bank’ model is substantially misspecified.