Learning in a Generalized Dornbusch Model of Exchange Rate Dynamics

In this paper we propose a framework for studying possible causes of excess exchange rate volatility. The framework consists of a generalized Dornbusch model of exchange rate dynamics, involving imperfect substitutability between assets, lagged nonlinear protfolio adjustment and les than perfectly rational expectations. As the model involves non-linear portfolio adjustment, it remains globally bounded even when the steady state is locally unstable. This economic environment is populated by a group of sophisticated agents who employ a maximum likelihood learning algorithm tolearnce about the "true" model. We use simulations to study the convergence of the learning scheme and its effect on exchange rate dynamics. Our analysis suggests that learning of speed of adjustment type parameters can be a source of exchange rate bubbles because of their effect on the local stability of the steady state.