Forecasting revisions of German industrial production

ABSTRACT

Macroeconomic variables, such as industrial production or GDP, are regularly and sometimes substantially revised by the official statistical offices. Nevertheless, there are only few attempts in the previous literature to investigate whether it is possible to forecast these revisions systematically. In this article, it is illustrated how revisions of German industrial production can be forecasted with respect to both the direction and the level of the revision. We are the first to use a large data set for this purpose.

KEYWORDS:

• Industrial production
• revisions
• forecasting
• large data sets
• forecast combination

JEL CLASSFICATION:

• C53
• E37
• E66

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ There are some peculiarities with the industrial production release data set worth to be mentioned here. First, industrial production figures released before 22 November 2005 were derived manually and not by automatic systems. It does not surprise us that the revision frequency and intensity differ during this period appraising to 21 release dates in total. Second, releases belonging to the release dates in the transition phase to automation (from 22 November 2005 to 24 January 2006) have been published twice per month with occasional differences between them. The authors assume that the statistical agency regards the second release per month as the definite one, hence the first one has been ruled out from the data set.

² The aspect of benchmark revisions is intensively discussed in Knetsch and Reimers (2009).

³ See Timmermann (2006) for an overview and further details.

⁴ The full list of results is listed in the working paper version of this article in Bührig and Wohlrabe (2015).