Review of macro modelling for policy purposes at Norges Bank
- Fabio Canova, Francesco Furlanetto, Frank Smets and Volker Wieland
- Occasional Papers
The evaluation team was given a mandate to assess whether the current set of models employed for monetary policy analysis and forecasting purposes (i) constitute best practice in the policy arena and in relation to recent academic developments; (ii) are appropriate to answer relevant policy questions for an inflation-targeting central bank in a small open economy; (iii) and can be further developed in the years to come and in what way.
Our general conclusions are that i) Norges Bank is at the high end of the spectrum of best-practice central banks; ii) that the modelling strategy used to support the policy process is adequate and iii) that Norges Bank is conscious of recent developments at the frontier of macroeconomic modelling and is actively working in some of these areas. Nevertheless, we indicate a number of changes to the structure of the core model (NEMO) and, more generally, to the set of models used to support the policy decision process. We also present some strategic considerations on model developments for the years to come.
We believe that the policy analyses and the forecasting process are credibly and consistently conducted and well understood by staff and management. Norges Bank uses a New Keynesian DSGE model as the core model for policy analysis, forecasting, risk management and communication that is well developed, consistent, and state of the art. The model has evolved over time and is used effectively to address a variety of questions regarding policy and transmission channels. Core staff and management at Norges Bank seem confident that the current set of models and the way they have been used in the recent past are appropriate to answer policy questions. We perceived a sense of common ownership of the model used to guide the production of the interest rate path.
While we endorse the strategy of using one core model, we recommend supplementing the analysis with simple monetary policy rules and some smaller models for cross-checks, in order to avoid fine tuning policy to the core model. Moreover, we provide a list of detailed suggestions to improve the current modelling framework. NEMO should explicitly incorporate trends to allow forecasts of the level of the variables. Correlations among endogenous variables should be checked. The relative importance of various shocks should be examined and out-of-sample forecast evaluation of staff projections regularly performed. Satellite models could also be used to build credible international scenarios to analyse risks, fiscal policy and other issues that, if added, could considerably complicate the NEMO analysis.
Model development at Norges Bank seems to be organised appropriately, but we recommend more cooperation among the different units and some outsourcing of model maintenance/upgrading. We advise building a network with modelling teams from other resource-rich, small open economies and we stress the importance of documenting model development (and policy work more generally) in the form of polished and publishable policy research.
In terms of future developments, we believe that heterogeneities, non-linearities and alternative expectation formation mechanisms will be areas of research that central banks should take advantage of in the next 5-10 years. Norges Bank seems conscious of these developments and is actively working in these areas.
ISSN 0802-7188 (printed), 1504-0577 (online)