Discussion Paper

No. 2011-6 | April 11, 2011
The Accuracy of a Forecast Targeting Central Bank

Abstract

This paper evaluates inflation forecasts made by Norges Bank which is recognized as a successful forecast targeting central bank. It is reasonable to expect that Norges Bank produces inflation forecasts that are on average better than other forecasts, both ‘naïve’ forecasts, and forecasts from econometric models outside the central bank. The authors find that the superiority of the Bank’s forecast cannot be asserted, when compared with genuine ex-ante real time forecasts from an independent econometric model. The 1-step Monetary Policy Report forecasts are preferable to the 1-step forecasts from the outside model, but for the policy relevant horizons (4 to 9 quarters ahead), the forecasts from the outsider model are preferred with a wider margin. An explanation in terms of too high speed of adjustment to the inflation target is supported by the evidence. Norges Bank’s forecasts are convincingly better than ‘naïve’ forecasts over the second half of our sample, but not over the whole sample, which includes a change in the mean of inflation.

Data Set

Data sets for articles published in "Economics" are available at Dataverse. Please have a look at our repository.

The data set for this article can be found at: http://hdl.handle.net/1902.1/15840

JEL Classification

C32 C53 E37 E44 E47 E52 E58 E65

Cite As

Nina Skrove Falch and Ragnar Nymoen (2011). The Accuracy of a Forecast Targeting Central Bank. Economics Discussion Papers, No 2011-6, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2011-6

Assessment



Comments and Questions


Anonymous - Forecast evaluation
April 12, 2011 - 09:25

The paper provides an interesting analysis on the forecasts made by Norges Bank. However, the forecast evaluation is based on the simple comparison of point values of different loss functions (mean forecast error and mean squared forecast error). From a statistical point of view, the differences noted across different models ...[more]

... could be just apparent, given that the loss functions could be statistically equivalent. To infer some possible different behavior across the models compared in this study, you should use more formal tests for evaluating forecasts. Some examples are given by the Diebold and Mariano test, and by the Model Confidence Set of Hansen et al. (Oxford Bulletin of Economics and Statistics, 2003, and Econometrica, forthcoming).


Ragnar Nymoen - uncertainty and significance
April 15, 2011 - 11:42

The issue about significance of forecast differences is both relevant and important. In our case, where we study published (official) forecasts, the number of observed forecast errors is limited, and increase only slowly with calendar time. We therefore chose to present simple MSFE and MFE graphs. However, we will ...[more]

... follow your advice and seek ways of assessing the uncertainty, including formal tests, although they may need to be interpreted with care given the still relatively small sample sizes, in particular for the longer horizons.


Anonymous - Referee Report 1
April 29, 2011 - 09:59

See attached file


Ragnar Nymoen - Authors' reply to Referee I report
May 10, 2011 - 11:57

Pdf with reply.


Anonymous - Referee Report 2
June 10, 2011 - 09:28

It is very rare that I read a paper that is very well written, I so wholeheartedly agree with, is well documented, and answers an interesting question that extends the literature in a direction that is relevant.

The paper compares the forecasting performance of an ‘inflation-targeting’ central bank (Norges ...[more]

... Bank) against forecasts from an economic model and a simple univariate forecasting model and the role that structural breaks play in forecast errors.

In the process the authors:

(i) Clearly sets the analysis within the correct literature on forecasting and the role that structural breaks plays in forecast errors;

(ii) Clearly sets out the link between the theory of forecasting with structural breaks and the practicalities of forecasting for the purpose of monetary policy;

(iii) Carefully carries out the theoretical and applied analysis and documents the process to the point where the reader can replicate the analysis (not something that most econometricians attempt routinely); and

(iv) Models the statistical process of inflation in a way consistent with the actual statistical process of inflation. This may seem strange praise for an econometric analysis of inflation but most empirical papers on inflation proceed under the assumption that inflation is either stationary (with a constant mean) or integrated. Both assumptions are very difficult to sustain in an empirical sense.