Model selection for monetary policy analysis: How important is empirical validity?

We investigate the economic significance of trading off empirical validity of models against other desirable model properties, and the potential loss from ’overestimating’ model uncertainty and basing monetary policy on a relatively robust model, or on a suite of models. We find that differences in model specification and even differences in estimates of key parameters across comparable models may entail widely different monetary policy and macroeconomic performance. Our results therefore caution against compromising the empirical validity of models when selecting a model for policy analysis. We also find that potential costs from basing monetary policies on the relatively robust model or on a suite of models, even when it contains the valid model by assumption, can be quite substantial. This suggests huge gains from efficient exploitation of available information sources to avoid overestimation of model uncertainty. Our investigation is based on three alternative econometric systems of wage and price inflation for Norway.