What to target? Insights from a lab experiment.

• History-dependent monetary policy (AIT) lacks empirical evidence. • In a forecasting lab experiment, history-dependent alternatives are compared. • The performances of average inflation-targeting depend on its exact specification. • Longer histories are destabilizing, stronger reaction coefficients stabilizing. This paper compares alternative monetary policy regimes within a controlled lab environment, where groups of participants are tasked with repeatedly forecasting inflation in a simple macroeconomic model featuring only the dynamics of interest rates, inflation and inflation expectations. Average-inflation targeting can replicate the price path observed under price-level targeting in the presence of disinflationary shocks and enable subjects to coordinate on simple heuristics that reflect the concern of the central bank for past inflation gaps. However, this depends on the exact specification of the policy rule. In particular, if the central bank considers more than two lags, subjects fail to form expectations that are consistent with the monetary policy rule, which results in greater inflation volatility. Reinforcing communication around the target helps somewhat anchor long-run inflation expectations. [ABSTRACT FROM AUTHOR]