2024-08-04-欧洲央行-平均通胀目标制_在多大程度上审视过去和未来_（英）_48页_1mb

Summary

This paper analyzes the optimal length of the averaging window in an average inflation targeting (AIT) rule using a New Keynesian model incorporating behavioral and heterogeneous agent features. The central bank faces occasional binding effective lower bounds (ELB) on interest rates and persistent supply shocks, and can use quantitative easing (QE).

Key findings:

1. Optimal Window Length: Under a conventional degree of cognitive discounting (limitation) from Gabaix (2020), the optimal averaging window is infinite, meaning price-level targeting (PLT) dominates both AIT and inflation targeting (IT). As cognitive discounting increases, the optimal window length becomes finite but remains long. However, shortening the window does not bring it close to IT unless the central bank responds very quickly to deviations from the target path.

2. Policy Comparison: AIT consistently outperforms IT in terms of welfare loss reduction across different degrees of cognitive discounting. The superiority of PLT is less pronounced under higher myopia (stronger cognitive discounting).

3. Role of Speed of Return: The speed at which the central bank returns average inflation to its target path significantly impacts the optimal window length. A faster return shifts the optimal window toward shorter lengths, potentially making IT preferable under realistic reaction parameters.

4. Extensions: Introducing backward-looking inflation expectations or allowing QE does not fundamentally alter the main results. The benefits of history dependence in AIT remain robust even with these adjustments.

The model resolves the "forward guidance puzzle" and captures monetary policy amplification through indirect general equilibrium effects. The global solution approach highlights welfare effects from uncertainty about future ELB binding risks, which the local solution method fails to capture.