EMU's Decentralized System of Fiscal Policy

the Maastricht Treaty and then was put in charge of restoring competitiveness because
exchange rate realignments are no longer possible, we expect to find θ^b < 0 and θ^a > 0 .

Another hypothesis, mentioned above, is that the use of fiscal policy in the face of a
deterioration of competitiveness depends on the cyclical position of the economy. For
instance, it could be procyclical during upswings and acyclical or countercyclical during
downswings. This hypothesis can be tested by adding the following terms:

^λb+Et-1^yt^++λb-Et-1^yt^-+λa+Et-1^yt^++λa-Et-1^yt^-

whereλ^b+ andλ^b- refer to periods before 1992 when external competitiveness is week and
when, respectively, the output gap is positive and negative, with a similar definition for
λ^a+ andλ^a- the coefficients corresponding to the post-1992 period. Including these terms
generally deteriorates all the estimates, and yields non-significant values for these terms. This
remains the case if we eliminate the pre-Maastricht years - i.e. we impose λ⁺ = λ^- = 0 -
with λ^a+ > 0 , andλ^a- < 0 but still not significant.

In the end, therefore, we estimate:

(5)     dt=c+β^bEt-1yt+β^aEt-1yt+β^a-Et-1yt^-+θ^bXt+θ^aXt+γbt-1+ρdt-1+ut

To measure competitiveness, we use the real exchange rate, defined as relative unit labor
costs, constructed such that an increase represents in external competitiveness. Since some
real exchange rates are trended, we use the Hodrick-Prescott filter with a high degree of
smoothing to preserve low frequency fluctuations.⁸ We instrument Xt with its lag and the real
exchange rates of Japan and US and their own lags.

Table 1 presents our results.⁹ The deficit dt is cyclically-adjusted and net of interest payments
on the debt. We present first country by country estimates and then the results from pooling
all Euro Area countries together. The country by country estimates suffer from the short
sample; the longest time series are only available for 1971-2006, and much less so some
countries. The pooled estimates assume that the coefficients are the same for all countries,
which may not to be the case, but they rely on a larger sample (the panel is unbalanced with a
total of 270 observations).

Table 1. The budget deficit reaction function

Source: see Appendix

⁸ We set the smoothing parameter to 1000.

⁹ Due to missing data, Greece, Luxembourg and Portugal are not included in the sample.

<<   <   6   7   8   9   10   11   12   >   >>