Palgrave Handbook of Econometrics: Applied Econometrics

(Grace) #1
Gunnar Bårdsen and Ragnar Nymoen 901

2005

0.5

1.0

1.5

2.0

2.5

July 2005

Actual

Normetrics

Norges Bank

Average of other forecasts
(than Norges Bank’s)

2006 July 2006

Figure 17.9 Forecasts for annual core inflation in 2006, published at different times. Percent.
Monthly figures, January 2005 to December 2006
Source: Norges Bank,Monetary Policy ReportJanuary 2007, and http://folk.uio.no/rnymoen/Normetrics.


any worse than the average forecasts, and most vintages of Normetrics forecasts
are considerably better. All Normetrics forecasts produced in 2005 also improve on
Norges Bank’s forecasts (June 2005 is the exception). In the period November 2005
to February 2006 the gap between Normetrics and Norges Bank actually widens – as
Norges Bank’s forecasts were adjusted upward, away from what eventually became
the actual inflation rate of 0.8%. However, the forecast inMonetary Policy Report
February 2006 is accurate, while the Normetrics forecast stays at 1% until September
2006.
Since equilibrium-correction is ubiquitous in macroeconometric models, one
can safely assume that some of the “other forecasts” are based on EqCMs. Specifi-
cally, Norges Bank’s forecast are based on the rational expectations solution of an
equilibrium-correction model with leads in variables. Hence, all forecasts in Figure
17.9 may have been damaged by any location shifts that took place in 2006 – they
wereafter-forecaststructural breaks. In particular, the forecast that was published
(early) in 2005 had a large exposure to forecast failure. For the same reason, the
forecast errors are reduced as more “2006 information” is conditioned upon. In that
perspective we can interpret the figure as evidence that the Normetrics forecasts
are more adaptive than the other equilibrium-correction forecasting mechanisms
covered by the graph.
The accuracy of the model-based forecasts in Figure 17.9 are less impressive when
compared to forecasts from a simple DDD, though. For example, at the start of 2005
a forecast based on the double difference of the log of the CPI-AET (consumer price

Free download pdf