It has been noted that empirical monetary policy reaction functions show smaller impact coefficients and more persistence than reaction functions that are computed from policy optimization experiments using conventional macroeconomic models. Papers by Rudebusch, Smets and others attempt to explain this discrepancy as arising from a mixture of data and model uncertainty. This paper offers an alternative and simpler explanation. In a complex economy optimal policy make look weak and slow only because it is appropriately responding a wide variety of shocks. Using the Board of Governors' large-scale rational expectations model of the U.S. economy, FRB/US, we exploit a new optimal control tool, developed by Finan and Tetlow, that is particularly useful for computing rules of large-scale models. We show that the optimal rule--which actually contains some 350 arguments--is observationally very similar to empirical reaction functions. Thus it is possible that policy conduct appears as it does because the Fed has more information than researchers not less.
