The Performance of Forward-Looking Monetary Policy Rules under Model Uncertainty

Recently, increasing attention is being devoted to interest rate rules that respond directly to economic forecasts rather than relying on current and past observations. Empirical studies suggest this 'forward-looking' rule provides a reasonable description of recent monetary policy in several industrial countries. Such rules are also advocated on analytical grounds including simplicity, transparency, and efficiency. But, such rules can fail to generate unique rational expectations equilibrium under various combinations of structural and policy parameters, raising concerns whether forward-looking rules are robust to model uncertainty. Here, we analyze the efficiency and robustness of forward-looking rules using four structural macroeconometric models of the U.S. economy: the Fuhrer-Moore model, Taylor's Multi-Country Model, the MSR model of Orphanides and Wieland, and the FRB staff model. All four incorporate assumptions of rational expectations, short-run nominal inertia, and long-run monetary neutrality, but differ in other respects such as real expenditures and the dynamics of prices. We assume a policy objective of minimizing a weighted sum of the unconditional variances of the inflation rate, the output gap, and federal funds rate changes. For given model and particular class of policy rules, we determine the region of the parameter space for which simple forward-looking and backward-looking rules generate unique equilibria. Within this region, we determine the performance of rules on the policy frontier -- the best obtainable outcomes for output, inflation, and funds-rate volatilities. Finally, we evaluate robustness to model uncertainty, taking rules that perform well in one model and assessing their performance in each of the other three. Our analysis yields three significant conclusions. First, the indeterminacy problem is not one of practical concern. All four models exhibit a relatively high degree of nominal and real inertia consistent with U.S. data. Second, in each model, we find that forward-looking rules provide negligible stabilization benefits compared with well-designed backward-looking rules. Finally, the forward-looking rules that perform well in one model often perform very poorly in the other three. By contrast, simple backward-looking rules taken from the policy frontier of one model are generally very close to the frontier in each of the other three. Thus, while small improvements in output and inflation variability can sometimes be obtained using forward-looking rules, we find them to be much less robust to model uncertainty than simple backward-looking rules.