THE ECONOMETRICS OF MACROECONOMIC MODELLING
Empirical wage equations
Empirical models of Nordic manufacturing wage formation are reviewed and updated in Nymoen and Rpdseth (2003). Their results for Denmark, Finland, Norway, and Sweden strongly reject the Phillips curve specification. The evidence against the Phillips curve hypothesis, 9w = 0, is not confined to the Nordic countries; see, for example, Grubb (1986) and Dreze and Bean (1990) who analyse manufacturing wages for a number of European economies.
Turning to the bargaining model, the main idea is that the NAIRU can be derived from the long-run real wage and price equations. If there is no wedge term in the wage equation, the NAIRU is independent of the real exchange rate. However, the above analysis shows that only subject to specific restrictions does the wage curve NAIRU correspond to the steady state of the system. The Nordic study by Nymoen and Rpdseth (2003), while supporting that ш = 0, implies strong rejection of the NAIRU restrictions on the dynamics. Results for other European countries give the same impression: for example, 6 out of 10 country-studies surveyed by Dreze and Bean (1990) do not imply a wage curve NAIRU, since they are not genuine product real wage equations: either there is a wedge effect in the levels part of the equation (ш > 0), or the authors fail to impose фwq = 1, фыр = 0.[45]
For the United Kingdom, there are several individual studies to choose from, some of which include a significant wedge effect, that is, ш > 0 (see, for example, Carruth and Oswald 1989 and Cromb 1993). In a comprehensive econometric study of United Kingdom inflation, Rowlatt (1992) is able to impose dynamic homogeneity, фwp+фwq = 1 in wage formation, but the NAIRU restriction фwq = 1 is not supported by the data.[46] The work of Davies and Schptt-Jensen (1994) contains similar evidence for several EU countries. For the majority of the data sets, consumer price growth is found to be important alongside producer prices, and as we have shown this is sufficient to question the logical validity of the claims made in the same study, namely that a steady-state unemployment equilibrium is implied by the estimated real-wage equations.
OECD (19976, table 1.A.1) contains detailed wage equation results for 21 countries. For 14 countries the reported specification is of the wage-curve type but the necessary restrictions derived above on the short-run dynamics are rejected. Phillips curve specifications are reported for the other seven countries, notably for the United States, which corroborate evidence in other studies; see Blanchard and Katz (1997) for a discussion.
This brief overview confirms the impression that the evidence from European data supports a wage curve rather than a Phillips curve specification. However, in the light of the model framework of this section, the estimated wage curves do not support the identification of the implied NAIRUs with the equilibrium level of unemployment.
