positively to the level of government debt. According to the results presented by those
authors, there seems to be a statistically significant positive relation between public debt
and the primary budget surplus, being therefore impossible to conclude that
governments do not take into account their respective intertemporal budget constrainst.
In other words, fiscal policy might have been implemented according to a Ricardian
regime and therefore, these empirical results could not validate the FTPL hypothesis.
An approach similar to the one implemented by Mélitz (2000) is also adopted by Creel
and Sterdyniak (2000), who mention that fiscal policy could be characterised by a
Ricardian regime in Germany and in the US, and by a non-Ricardian regime in France.
Additionally, another possible reading of the results presented by these two authors
might be the conclusion that fiscal policy may have been, in the past, sustainable in
Germany and not sustainable in France.15
4.2. Some evidence for the EU-15
The idea of implementing causality tests between real government debt and the real
primary surplus, implied in the VAR models mentioned above, is not an easy approach
to assess the possibility of the FTPL. In fact, both these variables are part of the present
value borrowing constraint, a constraint that in the end holds true in any fiscal regime,
either Ricardian or non-Ricardian. Since we are concerned with the EU-15 countries, a
possible strategy might be to pool the data and use panel models along with some
plausible testable assumptions.
In order to assess the possibility of the FTPL for the EU-15, panel data are used for the
primary budget surplus, as a percentage of GDP, and for the debt-to-GDP ratio, between
1970 and 2001.16 I use therefore 32 years of annual observations for 15 countries.
The existence of differences between the several countries is taken into account, by
allowing that the autonomous term changes from country to country, in each cross-
15 This is in fact one of the conclusions presented by Afonso (2000).
16 The statistical sources are mentioned in the Annex.
22