Marleau Lecture Series on Economic and Monetary Policy

If a monetary policy path is chosen optimally, any perturbation to that path should have no first-order effect on welfare. Drawing on this insight, we show that the impulse responses to monetary shocks are sufficient statistics to evaluate the optimality of a given policy, and we propose a ``sufficient-statistic targeting rule'' that preserves all the benefits of targeting rules ---simplicity, transparency and immunity to time-consistency problems---, while remedying their one major limitation: non-robustness to model variations. The sufficient-statistic targeting rule encompasses earlier targeting rules from the literature and generalizes the Brainard conservatism principle to a model-free setting. We test the optimality of the Fed monetary policy over the recent period, and we reject that the Fed policy was/is optimal