medium-run steady-state relations. As we are not yet able to impose and test overi-
dentifying restrictions on the estimated vectors, interpreting the unrestricted estimates
does not make much sense and will not be done here.
9.2 The dynamics of the short-run adjustment
Table 5 reports the estimates of the short-run adjustment coefficients associated with
the polynomially cointegrating relations, a^ = 1,2, 3, and the coefficients ζ i = 1,..., 5
associated with the changes in the five equilibrium errors, τ'i∆xt. The number of
estimated coefficients is large, making it difficult to summarize the main results in a
simple way. We shall not make detailed comments on the results, but instead give a
cursory description of the basic adjustment mechanisms in this system.
The estimated a1 shows that all variables, except the German long-term bond
rate, react very significantly on the equilibrium error from the IKE relation, pointing
to its importance for the international transmission mechanisms. The estimated a2
shows that relative prices, nominal exchange rates, and US inflation rate and the two
German interest rates (whereas not the US rates) react significantly to a deviation
between expected inflation and its determinants. The estimated a3 is consistent with
the interpretation of the third cointegration relation as a relation for German inflation
rate, as it is essentially prices which are reacting on an equilibrium error, albeit the
short-term interest rates show some small effects.
Given that the nominal exchange rate was found to exhibit no long-run feed-back
effects in Section 8, it is somewhat surprising that there are two significant a coefficients
in the exchange rate equation. However, the test for a zero row in a in Section 8 was
for β relations between the levels of variables, whereas the estimated a coefficients
in 9.2 correspond to polynomial cointegration relations containing variables in levels
and differences. Furthermore, the combined relation (a1,2β 11+ a2,2^2 t) suggests that
the two significant cointegration relations almost neutralize each other, nonetheless
with some small but significant evidence of the first relation being important for the
nominal exchange rates. Thus, the result suggests that the ppp may act as an anchor
for exchange rates, even though we do not expect its relationship to be completely
stable over time.
One of the important questions in international macro is why prices and exchange
rates adjust so sluggishly to the ppp. The answer provided by FGJ is that, with imper-
fect knowledge, equilibrium in the goods markets is no longer characterized by PPP,
but by a cointegrating relationship between ppp, b1 — b2 , and ∆p1 — ∆p2. FGJ show
that relative goods prices adjust to this equilibrium relation extremely fast, which can
be seen from the estimates in Table 5 which are also reported in FGJ. The estimated
adjustment coefficient a 11 = 0.39 shows that relative goods prices (and US inflation
rate with al3 = —0.59) adjust very fast to the first cointegration relation. By contrast,
the adjustment of relative prices to ppp is very slow -(0.39x0.01). Thus, ppp acts as an
anchor for prices, but the chain is very long indeed.
The estimated coefficients of ζi may not be highly interesting for the following
21