Robust Macroprudential Policy Rules under Model Uncertainty

Against the backdrop of elevated model uncertainty in DSGE models with a detailed modeling of financial intermediaries, we investigate the performance of optimized macroprudential policy rules within and across models. Using three canonical banking DSGE models as a representative sample, we show that model-specific optimized macroprudential policy rules are highly heterogeneous across models and not robust to model uncertainty, implying large losses in other models. This is particularly the case for a perfect-coordination regime between monetary and macroprudential policy. A Stackelberg regime with the central bank as leader operating according to the rule by Orphanides and Wieland (2013) implies smaller potential costs due to model uncertainty. An even more effective approach for policymakers to insure against model uncertainty is to design Bayesian model-averaged optimized macroprudential rules. These prove to be more robust to model uncertainty by performing better across models than model-specific optimized rules, regardless of the regime of interaction between the two authorities.