Palgrave Handbook of Econometrics: Applied Econometrics

Katarina Juselius 363

Beta2’*x(t)

Beta2’*R1(t)

1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997

1975

2.4

5 4 3 2 1 0

–1

–2

–3

–4

1.2

0.0

–1.2

–2.4

–3.6

1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997

Figure 8.4 The graphs of the second cointegration relation (β′xtin the upper panel,β′R1,t
in the lower panel)

“more” pulling than pushing, which is an interesting observation as one would
expect the opposite during a currency float.

8.5.3 Estimating the long-run structure

Table 8.3 reports the estimates ofα,β, 1 andfor the choice ofr=1. The esti-
matedβrelation suggests thatp1,tandp2,tare almost homogeneously related.

Testing the hypothesis gives a test statisticχ^2 ( 1 )=0.56[0.46]and, thus, price
homogeneity ofβ′xtseems acceptable^5 when allowing for a broken trend. The
presence of a broken linear trend might seem difficult to interpret but is probably a
proxy for omitted variables effects, such as the effect of productivity differentials on
relative prices, the so-called Balassa–Samuelson effect (Balassa, 1964; Samuelson,
1964). The change in the trend slope at reunification supports this interpretation.
What is more surprising, however, is that the sign of the nominal exchange rate is
opposite to the expected one. Based on Figure 8.1, it is easy to see why: over the
sample period, relative prices and nominal exchange rates have frequently moved
in opposite directions for extended periods of time. For this reason, the data do
not support theppprestrictions (1,−1,−1) onβ.