Discussion Paper

No. 2014-25 | June 20, 2014
How Might a Central Bank Report Uncertainty?

Abstract

An important question for central banks is how they should report the uncertainty of their forecasts. This paper discusses a way in which a central bank could report the uncertainty of its forecasts in a world in which it used a single macroeconometric model to make its forecasts and guide its policies. Suggestions are then made as to what might be feasible for a central bank to report given that it is unlikely to be willing to commit to a single model. A particular model is used as an illustration.

JEL Classification:

E50

Assessment

  • Downloads: 2558

Links

Cite As

Ray C. Fair (2014). How Might a Central Bank Report Uncertainty? Economics Discussion Papers, No 2014-25, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2014-25


Comments and Questions


Anonymous - Referee report
August 04, 2014 - 16:13

See attached file


Ray C. Fair - Reply to referee report
August 06, 2014 - 09:32

The referee was generally positive, and the following are my remarks about his/her general comments.

I could add discussion of what the Bank of England does. There is some discussion in the paper about procedures at the Fed and the Bank of Norway, and it would be easy ...[more]

... to add the Bank of England.

On DSGE models, the procedures discussed in the paper do not cover these models. The general model (page 3) can be dynamic, nonlinear, have autoregressive errors, and have rational (model consistent) expectations, but this is not the DSGE setup. The FRB/US model discussed in the paper and the 221 + 171 models used by the Bank of Norway follow the specification of the general model and are not DSGE models. (The forecasts of the Bank of Norway models are fed into a DSGE model at the end for policy analysis.) So this is the group of models that the paper is aimed at. The FRB/US model follows exactly the methodology assumed in the paper and the methodology behind the MC model that is used as an example, so the paper has practical relevance.

On parameter uncertainty, I have done stochastic-simulation experiments (not discussed in the paper) drawing first only structural error terms and then both structural error terms and coefficient estimates, and the uncertainty from the coefficients is much smaller than that from the structural error terms. I can discuss this if you want. I have published papers on this, and it is in the MM document on my website. The procedures in the paper could be modified to incorporate parameter uncertainty, but this would be tedious and computationally expensive. The gain is likely to be small, and my guess that it would never be done in practice by central banks.

Finally, the referee makes a good point about model uncertainty. Since models are not generally nested, my view is that it is best to do the procedures in my paper for a number of models and simply look at the differences to gauge how sensitive the results are to the use of different models. Trying to rigorously combine density forecasts is problematic for the kind of models considered in this paper.