Abstract
Macroeconomic models are important ingredients in the monetary policy process, and, in the Norwegian case, projecting a forward interest rate path. In this paper we argue that when deciding on a model strategy, it is crucial to consider the purpose of models. If the purpose is to understand basic mechanisms in the economy and implications of economic policy, we need a set of models that highlight these features. If the purpose is to forecast short-term developments, a different set of models may be required. Given the complexity of the real world, we argue that it is better to provide the policymakers with a good characterization of uncertainty instead of only providing point forecasts, i.e. it is better to be "roughly right" than "exactly wrong". A robust strategy for handling uncertainty should be an inherent part of the preferred system of models.