allowed deficit (α3, β2 < 0). In other words, the more binding the fiscal rule the higher
the incentives to resort to fiscal gimmickry.
Since x, z, and p (and therefore d) are simultaneously determined, we report results
obtained using two stage least squares. We present two sets of estimates: one uses the
most recent data releases, the other one considers the values first reported by Greece,
Italy, and Portugal over the period 1998-2004 (that is data published before the statistical
revisions discussed in Section 6.3). We also introduce a dummy variable to test for
structural breaks after 1998.
The sample includes data for fifteen countries over 1994-2004. We also run regressions
on the sub-sample of euro-area members. Deficit-specific SFA are obtained from the Buti
et al. (2007) dataset as the sum of cash-accrual differences and of net acquisitions of
financial assets, excluding privatization, following the discussion in Section 6.2. The
“true deficit” is obtained by summing the deficit-specific SFA to deficit data used in the
context of the excessive deficit procedure (EDP) as reported in the AMECO database.
The “growth effect” is computed using data from the AMECO database. Finally, and
again in line with the discussion in Section 6.2, debt-specific SFA are obtained residually
by subtracting from total SFA the deficit-specific SFA component and the growth effect.
Tables 6.3 and 6.4 report the estimation results, which are in line with expectations from
the model. All coefficients are correctly signed, statistically significant and exerting
quantitatively large effects on the dependent variables. There is no significant difference
between results for the EU15 and those for the euro area. Using the original data releases
for Greece, Italy, and Portugal induces an improvement in the regression fit for the
deficit-specific equation, but not for the debt-specific one, possibly reflecting the limited
extent of revisions to changes in debt compared to revisions in deficits.
The deficit-specific SFA is positively correlated with the “true deficit”. For each one
percent of GDP increase in the true deficit there is an estimated 0.3 percent of GDP
increase in deficit-specific SFA. The deficit-specific SFA is also negatively correlated
with the maximum allowed deficit (H). The increase in deficit-specific SFA associated
with a one percent of GDP increase in the allowed maximum deficit is estimated at 0.2-
0.25 percent of GDP, using the data set which includes recent revisions to data for
Greece, Italy, and Portugal, and to almost 0.5 percent of GDP, excluding such revisions.
The debt-specific SFA is positively correlated with the “true deficit”. For each one
percent of GDP increase in the true deficit, there is an increase by about 0.3 percent of
GDP in debt-specific SFA, enough to offset the corresponding estimated increase in
deficit-specific SFA. The debt-specific SFA is negatively correlated with the “growth
effect”. For each one percent of GDP increase in the growth effect, there is a 0.30-0.35
percent of GDP decrease in debt-specific SFA. Finally, the debt-specific SFA is also
negatively correlated with the maximum allowed change in debt (W). For each one
percent of GDP increase in the allowed maximum change in debt, there is an estimated
0.2 percent of GDP decrease in debt-specific SFA.
175