Palgrave Handbook of Econometrics: Applied Econometrics

Gunnar Bårdsen and Ragnar Nymoen 905

2003

–2

0

2

4

6

2004 2005 2006 2007 2003

2.5

2.0

1.5

3.5

3.0

4.0

4.5

2004 2005 2006 2007

2003

–2

–5

–10

0

2

4

10

6

8

2004 2005 2006 2007 –4 2003

–2

0

2

8

4

6

2004 2005 2006 2007

2003

1

2

3

4

7

5

6

2004 2005 2006 2007 2003

0

4

8

16

12

(^20042005200620072003)
4
8
16
12
2004 2005 2006 2007
2003
3
2
1
5
4
6
7
2004 2005 2006 2007
2003
88
84
96
92
100
104
2004 2005 2006 2007
(a) Consumer price inflation (b) Unemployment rate (c) Wage growth
(d) GDP growth
(g) Money market interest rate (h) Real credit growth (i) Currency depreciation rate
(e) Import price inflation (f) Real exchange rate
Figure 17.12 Dynamic NAM forecasts 2005(1)–2007(3), with end-of-sample for estimation
of parameters in 2004(4)
Actual values are shown by solid lines, and forecasts by dashed lines. The distance between the two dotted
lines represents 70% prediction intervals.
Figure 17.13 focuses on the last three “problem variables” in terms of forecast fail-
ure. The rate of unemployment is now much better forecasted in 2007(1)–2007(3),
which is a sign of adaptation to a location shift which is of the before-forecast
category, where 2006(4) is in the information set. For GDP and real credit growth
the forecast failures persist. In the present version of NAM, the high growth rates
of 2007 can, in part, be explained by the effects of very high oil prices on demand.
In fact, that modeling device was used in Figure 17.6, showing the goodness-of-fit,
but clearly would not be known or of any help to a forecaster preparing a forecast
for 2007 late in 2006.
As mentioned above, using differencing (DDD) to forecast provides a more robust
forecast when non-stationarities are due to location-shifts. As discussed by Hendry
(2006), a differenced version of the EqCM may be interpreted as an augmented
DDD forecasting rule. We therefore consider forecasts from the differenced NAM