Managing Expectations without Rational Expectations

Should a policymaker manage expectations by offering forward guidance in terms of the likely value of a future policy instrument or a target for an equilibrium outcome such as aggregate output? We study how the optimal approach depends on plausible bounds on agents' depth of knowledge and rationality. Agents make mistakes in predicting, or reasoning about, the behavior of others and the GE effects of policy. The optimal communication strategy minimizes the bite of such mistakes on implementability and welfare. This goal is achieved by offering guidance in terms of an outcome target rather than a policy value if and only if the GE feedback is strong enough. Our results suggest that central banks should stop talking about interest rates and start talking about unemployment when faced with a steep Keynesian cross or a prolonged liquidity trap