Conditional Welfare Comparisons of Monetary Policy Rules

Various approaches to optimal monetary policy have been used to select time-invariant policy rules, including the timeless perspective approach by Woodford (1999) and the unconditional expected utility criterion of McCallum (2000). In this paper, we argue instead that policy rules should be evaluated in terms of the household"s conditional welfare at the time of adoption, integrated over the ergodic distribution of initial conditions generated by the previous policy. Thus, the performance of time-invariant rules can be compared against the benchmark value achieved by the Ramsey solution, while ensuring that the coefficients of the rule are not fine-tuned to specific values of the initial conditions. Finally, using the benchmark new Keynesian model, we consider the case when the initial conditions are generated by optimal policy under discretion, and show that the optimal time-invariant rule yields conditional welfare superior to that of the rules previously proposed in the literature