This paper derives loss functions for analyses of optimal monetary policy that are grounded in the welfare of private agents, in the case of explicit optimizing models of private-sector behavior in which the real effects of monetary policy result from nominal price rigidity. A quadratic approximation to the utility-based welfare criterion is developed that allows comparison between this criterion and the ad hoc quadratic loss functions typically assumed in the literature on monetary policy evaluation. It is shown that the goal of inflation stabilization, generally presumed to be an important (and perhaps the preeminent) goal of monetary policy, can in fact be justified in such a framework, insofar as variable inflation results in real distortions when prices are not adjusted throughout the economy in a perfectly synchronized fashion. The exact sense in which inflation variability matters for welfare, however, depends upon the details of price-setting behavior.

Conditions are described under which optimal policy involves complete stabilization of the price level. It is shown that this may be optimal even in the presence of "supply shocks" of several possible sorts (including technology shocks and exogenous variation in preferences regarding labor supply), and even in the presence of distortions that imply that the optimal output gap is positive (and despite existence of a non-vertical long-run Phillips curve). At the same time, a variety of reasons are discussed why complete price-level stabilization is not optimal in more complicated (and probably more realistic) settings.