THE ECONOMETRICS OF MACROECONOMIC MODELLING

Causality

The main-course model specifies the following three hypothesis about causation:

where * denotes one-way causation. Causation may be contemporaneous or of the Granger-causation type. In any case the defining characteristic of the main-course model is that there is no feedback between, for example, domestic cost of living (pt) and the wage level in the exposed sector. In his 1977 paper, Aukrust sees the causation part of the theory (H4mc-H6mc) as just as important as the long-term ‘controlling mechanism’ (H1mc-H3mc). If anything, Aukrust seems to put extra emphasis on the causation part. For example, he argues that exchange rates must be controlled and not floating, otherwise pft (foreign prices denoted in domestic currency) is not a pure causal factor of the domestic wage level in equation (3.2), but may itself reflect deviations from the main course, thus

In a way, ..., the basic idea of the Norwegian model is the ‘purchasing power doctrine’ in reverse: whereas the purchasing power doctrine assumes floating exchange rates and explains exchange rates in terms of relative price trends at home and abroad, this model assumes controlled exchange rates and international prices to explain trends in the national price level. If exchange rates are floating, the Norwegian model does not apply. (Aukrust 1977, p. 114)

From a modern viewpoint this seems to be unduly restrictive since the cointeg­ration part of the model can be valid even if Aukrust’s one-way causality is untenable. Consider for example, H1mc, the main-course proposition for the exposed sector, which in modern econometric methodology implies rank reduc­tion in the system made up of we, t, qe, t, and ae, t, but not necessarily one-way causation. Today, we would regard it as both meaningful and significant if an econometric study showed that H1mc (or more realistically H1flmc) constituted a single cointegrating vector between the three I(1) variables {we, t,qe, t,ae, t}, even if qe, t and ae t, not only we, t, contribute to the correction of deviations from the main course. Clearly, we would no longer have a ‘wage model’ if wt was found to be weakly exogenous with respect to the parameters of the cointegrating vector, but that is a very special case, just as H4mc is a very strict hypothesis. Between these polar points there are many constellations with two-way causation that make sense in a dynamic wage-price model.

In sum, although care must be taken when we attempt to estimate a long-run wage equation with data from different exchange rate regimes, it seems unduly restrictive a priori to restrict the relevance of Aukrust’s model to a fixed exchange rate regime.

equations between growth rates

A'we, t = gf + gae, (3.10)

Aws, t = Awe, t, (3.11)

Aqs, t = Aws, t - gas, (3.12)

Apt = Фgf + (1- 4‘)Aqs, t. (3.13)

Most economist are familiar with this ‘growth rate’ version of the model, often referred to as the ‘Scandinavian model of inflation’. The model can be solved for the domestic rate of inflation:

Apt = gf + (1 - Ф)(дас - gas),

implying a famous result of the Scandinavian model, namely that a higher productivity growth in the exposed sector ceteris paribus implies increased domestic inflation.