Learning Benevolent Leadership in a Heterogenous Agents Economy

This paper studies the potential commitment value of cheap talkinflation announcements in an agent-based dynamic extension of theKydland-Prescott model. In every period, the policy maker makesa non-binding inflation announcement before setting the actualinflation rate. It updates its decisions using individual evolutionarylearning. The private agents can choose between two differentforecasting strategies: They can either set their forecast equal tothe announcement or compute it, at a cost, using an adaptive learningscheme. They switch between these two strategies as a function ofinformation about the associated payoffs they obtain throughword-of-mouth, choosing always the currently most favorable one.Weshow that the policy maker is able to sustain a situation with apositive but fluctuating fraction of believers. This equilibrium isPareto superior to the outcome predicted by standard theory. Theinfluence of changes in key parameters and the impact of transmissionof information among nonbelievers on the dynamics are studied.