Aggregate Wage Flexibility in Selected New EU Member States 5
varying estimations of wage flexibility (using the Kalman filter methodology), and analyzing the
effect of steps towards EMU membership on wage flexibility in a panel and cointegration
framework.
3. The Relationships Tested
In this paper, the issue of wage flexibility is addressed using three alternative methods:
(i) classical time series and panel estimates; (ii) the Bayesian approach; and (iii) cointegration and
error correction. The main hypothesis is stated as follows: How different is wage
flexibility/aggregate labor market adjustment in the following three groups of countries: (a) those
with autonomous exchange rate policy (the CE-4); (b) the ERM-II participants (three of them
having “hard pegs”); and (c) peripheral members of the euro area (the EMU-3)?
Notice that the causality between the exchange rate regime and wage flexibility (labor market
flexibility in general) can go in both directions. One way to address this issue is to base the
empirical results on a solid theoretical framework, which implies one direction of causality. For
example, wage-setting (WS) models could be used to study the impact of the exchange rate
change on wage adjustment. Next, the direction of causality can be tested empirically (for
example, in the sense of the Granger causality). Alternatively, there are methods (e.g. the third
method in the above list) which do not impose any a priori assumptions on the direction of
causality. Even if the precise assessment of causality is disputable, the estimation of aggregate
wage/labor market adjustment may be still informative. For example, a lack of adjustment may
motivate a need for deeper institutional reforms.
Each of the three methods has its own pros and cons. The first two methods focus on short-run
wage adjustment by estimating the basic Phillips curve specification. The second method
explicitly addresses the issue of structural changes. Using the same variables as in the first
method, the second method relaxes the assumption that the model’s coefficients are constant. This
is achieved by applying the Kalman filtering technique. So, institutional changes are detected.
Both methods, however, work with variables in differences (to render the series stationary).
Hence, the long-run dynamics are neglected. Alternatively, the third method is designed to assess
the long-term relationships between the variables in levels and also the short-run adjustment (the
error correction term). There is a risk, however, of there being no long-term significant and stable
relationship for some of the countries. In such a case, this suggests a potential problem on the
labor market (long-lasting shocks, absence of equilibrium).
(i) The classical estimation framework relies on the assumption that the regression parameters
are unknown constants. Following Alogoskoufis and Smith (1991), we estimate the basic Phillips
curve specification and test the stability of the coefficient on unemployment under fixed versus
floating exchange rate arrangements2.
2 The wage-unemployment trade-off can also be modeled within the broader framework of the open economy, as
described, for example, in Layard et al. (1991, p. 389). In such a model, which includes wage setting, price
setting, trade balance, and output gap-unemployment equations, the nominal exchange rate affects wages and
prices via import prices. In other words, price-setting behavior in the open economy depends on international
competitiveness (Carlin and Soskice, 1990, p. 255, and Layard et al., 1991, p. 385). Our choice of the
parsimonious Phillips curve (1) is driven by the data availability for Eastern European countries.