On Robustness of Average Inflation Targeting

This paper considers 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 may not be robustly stable under learning if the length of the averaging window is not public knowledge. Near the low steady state with interest rates at the zero lower bound, AIT does not necessarily outperform standard inflation targeting policy. Policymakers can improve outcomes under AIT by communicating the averaging window, or using an asymmetric rule that responds more aggressively to below-target average inflation.