Real-time model uncertainty in the United States: 'Robust' policies put to the test

I study 46 vintages of FRB/US, the principal macro model used by Federal Reserve Board staff for forecasting and policy analysis, as measures of real-time model uncertainty. I also study the implications of model uncertainty for the robustness of commonly applied, simple monetary policy rules. I first document that model uncertainty poses substantial challenges for policymakers in that key model properties differ in important ways across model vintages. Then I show that the parameterization of optimized simple policy rule--rules that are intended to be robust with respect to model uncertainty--also differ substantially across model vintages. Included in the set of rules are rules that eschew feedback on the output gap, rules that target nominal income growth, and rules that allow for time variation in the equilibrium real interest rate. I find that many rules, which previous research has shown to be robust in artificial economies, would have failed to provide adequate stabilization in the real-time, real-world environment seen by the Fed staff. However, I do identify certain policy rules that would have performed relatively well, and I characterize the key features of those rules to draw more general lessons about the design of monetary policy under model uncertainty.