Gunnar Bårdsen and Ragnar Nymoen 863Even though there now exists a literature on identification of cointegration
vectors, it is worthwhile to reiterate that identification of cointegrating vectors
cannot be databased. Identifying restrictions have to be imposeda priori. It is there-
fore of crucial importance to have a specification of the economic model and its
derived steady-state before estimation. Otherwise we will not know what model
and hypotheses we are testing and, in particular, we could not be certain that it
was identifiable from the available dataset.
17.2.5.3 Third step: the dynamic SEM
The final step is to identify the dynamic structure:
A 0 yt=A 0 c+A 0 αβ′
yt− 1 +∑ki= 1A 0 t−iyt−i+A 0 ut,by testing and imposing overidentifying restrictions on the dynamic part –
including the forward-looking part – of the statistical system.
The estimated parameters and, therefore, the interpretation of the model dynam-
ics are dependent upon the dating of the steady-state solution. However, the steady-
state multipliers are not. The economic interpretations of the derived paths of
adjustment are not invariant to the identification of the dynamic part of the model,
whereas the steady-state parts of the model are (see Bårdsen and Fisher, 1993, 1999).
17.2.6 Example: the supply side of a medium-term macroeconomic model
One main focus of an empirical macro-model is always going to be the supply
side. We end the section on methodology by giving an extended example of the
theoretical and econometric specification of a labor market model that we later
include in a macro-model, intended for medium-term analysis and forecasting.
The first step is the specification of the relevant economic theory to test. We next
develop the theoretical relationships into hypotheses about cointegration that can
be tested in a statistical model and identified as steady-state relationships, Steps
1 and 2 above. We also go through Step 3 in detail. Throughout the rest of the
chapter we let lower-case letters denote natural logarithms of the corresponding
upper-case variable names, soxt≡ln
(
Xt)
.17.2.6.1 Economic theory
A main advance in the modeling of labor markets rests on the perception that firms
and their workers are engaged in a partly cooperative and partly conflicting sharing
of the rents generated by the operation of the firm. In line with this assumption,
nominal wagesare modeled in a game theoretic framework which fits the compar-
atively high level of centralization and coordination in Norwegian wage-setting
(see, e.g., Nymoen and Rødseth, 2003; Barkbu, Nymoen and Røed, 2003, for a
discussion of the degree of coordination).
The modeling of nominal wage-setting in a game theoretic framework is a theo-
retical advance with several implications. Linked with an assumption of monopo-
listically competitive firms, it represents an incomplete competition model of the