On robustness of average inflation targeting

Abstract

This paper considers the performance of average inflation targeting (AIT) policy in a New Keynesian model with adaptive learning agents. Our analysis raises concerns regarding robustness of AIT when agents have imperfect knowledge. In particular, the target steady state can be locally unstable under learning if details about the policy are not publicly available. Near the low steady state with interest rates at the zero lower bound, AIT does not necessarily outperform a standard inflation targeting policy. Policymakers can improve outcomes under AIT by (i) targeting a discounted average of inflation, or (ii) communicating the data window for the target.