THE ECONOMETRICS OF MACROECONOMIC MODELLING

Revisions of output data: a case for real-time variables?

A first version of the quarterly national accounts (QNA) data is published by Statistics Norway shortly after the end of each quarter, based on a limited information set. As more information accrues, the data are revised and the final figures appear with a 18-months lag. Often there are substantial discrep­ancies between the first and the final quarterly data. The Norwegian QNA show that on average for the period 1995-99 growth in GDP for Mainland Norway was revised up by almost 1 percentage point per year, and, for example, the output growth for 1999 was adjusted from 1.1% to 2.7%. In Figure 10.1(a) we plot the growth rates for output according to the two sources together. The graphs reveal substantial revisions of output growth in the Norwegian mainland economy. The estimated change in interest rates accord­ing to the standard Taylor rule in Table 10.1 (FLX) is shown in Figure 10.1(b). Since the data revisions alone may induce up to 50 basis points change in the interest rate, there is a clear case for using interest rate rules with real-time variables.

Figure 10.1. Old and revised data for output in the mainland economy and
corresponding Taylor-rates, 1990(1)-2000(4). (a) Old and revised data for
output growth in the Norwegian mainland economy, five quarters centred
moving average. (b) Standard Taylor-rule interest rates from using five
quarters centred moving averages of old and revised data for output growth