Palgrave Handbook of Econometrics: Applied Econometrics

52 Methodology of Empirical Econometric Modeling

1920 1940 1960 1980 2000

–0.02

0.00

0.02

ef,1 – ef,1 ef,1 – ef,1

1920 1940 1960 1980 2000

0.00

0.05

0.10

0.15

1920 1940 1960 1980 2000

–0.03

–0.02

–0.01

0.00

0.01 s 1 – s 0

1920 1940 1960 1980 2000

–0.2

0.0

0.2

c 1

c 0

c 0 *

Figure 1.9 Deviations between old and revised data on food expenditure time series

food is unaltered – yet family size is unrecognizably different. The impacts on the
equilibrium-correction terms,c 0 in (1.39), that calculated for the revised datac∗ 0 ,
andc 1 in (1.45) below, are also shown (see Hendry, 1994; Cook, 2008, on possible
approaches for cross-data-vintage encompassing).
First, enforcing the identical specification to (1.40) but on the revised data over
1930–89, testing on the 11 new years led to:

ef,t= 0.34

(0.04)

st− 1 − 0.27

(0.02)

c0,t− 1 + 0.57

(0.06)

et+ 0.09

(0.05)

et− 1

− 0.40

(0.04)

(pf−p)t− 0.10

(0.01)

I 31 − 0.12

(0.01)

I 32 + 0.04

(0.01)

I 34

+ 0.02

(0.01)

I 41 + 0.05

(0.01)

I 42 + 0.03

(0.01)

I 51 + 0.02

(0.01)

I 52 + 0.04

(0.01)

I 70

(

R∗

) 2

=0.93FM(13, 45)=94.9∗∗̂σ=0.011Far(2, 44)=5.74∗∗

χ^2 ( 2 )=2.40Farch(1, 44)=2.79Freset(1, 45)=0.01

Fhet(18, 27)=0.82FChow(11, 46)=1.42. (1.44)

The revisions have altered the coefficients to some extent, the fit is poorer and there
is significant residual autocorrelation, but (without “correcting” the standard errors
for that problem), the Chow test does not reject, although as Figure 1.10 reveals,
the forecast errors are clearly autocorrelated.