Katarina Juselius 381the importance of first translating the basic assumptions of the theory model
into testable assumptions on the CVAR model. As an illustration, the chapter
showed how to translate the assumption of a stationary PPP and long-run price
homogeneity, together with the assumption that prices are pushing and the
exchange rate is pulling, into testable hypotheses in the CVAR model. This the-
ory consistent scenario showed, among others, that a stationary real exchange
rate is inherently associated withonestochastic trend having generated prices
and nominal exchange rates. The finding of two (rather than one) stochastic
trends was particularly important, as it suggested the existence of an additional
source of permanent shocks that have contributed to the persistent behavior in
the data. This additional shock seemed to be related to speculative behavior in
the market for foreign exchange and pointed to the importance of addressing
the long swings puzzle jointly with another puzzle in international finance, the
“forward premium puzzle.” Similar to the long swings puzzle, the forward pre-
mium puzzle also has to do with persistent movements in the data, now in the
forward premium: (R1,t−R2,t−Ets12,t+m), whereRi,tis an interest yield of
maturitym.
Thus the two puzzles are connected, in that both stem from the determination
of the nominal exchange rate in the foreign exchange market. Based on a coin-
tegrated VAR analysis of German and US prices, the exchange rate, and interest
rates, Juselius and MacDonald (2007) find thatpppand the real interest rate spread
are cointegrating though individuallyI(1), or even nearI(2). This is strong evi-
dence against the implications of the Dornbusch (1976) sticky-price model with
RE that PPP and real interest parity each should hold as equilibrium cointegrating
relationships. A theoretical justification for this strong feature in the data is, how-
ever, provided by Frydmanet al. (2008), who were able to show in a two-country
monetary model with IKE that goods prices and exchange rates adjust to a long-
run equilibrium relation, being a combination of thepppand the real interest rate
spreads.
Furthermore, Johansenet al. (2008) report results that point to the importance
of inflationary expectations measured by the term spread which was found to be
empiricallyI( 1 ). The latter finding again points to the importance of allowing for
not just one, but at least two, stochastic trends in the term structure of interest
rates (Giese, 2008), and thus to a reconsideration of the monetary policy interest
rate channel.
This illustrates how the VAR approach can be used constructively. Starting with
the basic information set, carefully structuring the information in the data, and
adding more information if needed, might at an early stage suggest how to modify
either the empirical or the economic model, or both.
The following passage from Hoover (2006) pinpoints the fundamental difference
between an approach based ona prioritheory and the general-to-specific approach
to empirical economics:
The Walrasian approach is totalizing. Theory comes first. Empirical reality must
be theoretically articulated before it can be empirically observed. There is a sense
that the Walrasian attitude is that to know anything, one must know everything.