Palgrave Handbook of Econometrics: Applied Econometrics

David F. Hendry 53

1940 1960 1980 2000

–0.10

–0.05

0.00

0.05

0.10 Δ^ef

Δef

1940 1960 1980 2000

–2

–1

0

1

residuals forecast errors

1990 1995 2000

–0.025

0.000

0.025

0.050

~Δe

f

Δef

–3 –2 –1 0 1 2 3

0.2

0.4

0.6

residuals

N(0,1)

Figure 1.10 Old model revised data fitted and actual values, residuals and forecasts foref,t

Next, automatic remodeling at 1% on the revised data up to 1989 (with impulse
saturation to remove the outliers) led to:

c 1 =ef+8.49−0.35e+0.21(pf−p), (1.45)

with a much simpler final equation being selected:

ef,t= 0.35

(0.032)

st− 0.25

(0.02)

c1,t− 1 + 0.65

(0.05)

et− 0.29

(0.04)

(pf−p)t

− 0.05

(0.01)

I 30 − 0.08

(0.01)

I 31 − 0.08

(0.01)

I 32 + 0.03

(0.01)

I 70

(

R∗

) 2

=0.90FM(8, 51)=60.4∗∗̂σ=0.012Far(2, 50)=0.65

χ^2 ( 2 )=2.09Farch(1, 50)=0.13Freset(1, 51)=0.11

Fhet(18, 33)=0.89FChow(11, 52)=0.68( 1990 − 2000 ). (1.46)

Nevertheless, despite the revisions, the model in (1.46) has many features in com-
mon with both its predecessors, and is constant over the next 11 years as Figure
1.11 reports, andFChow(11, 52)confirms. The short-run elasticities still exceed their
long-run counterparts, but by less than previously.