Our first robustness check on the elasticity parameters illustrates the effects of this
uncertainty. We conduct a grid search on different values for γ and α that Girouard and
André (2005) provide. Table 5.6 shows the wide range of net lending elasticity that is
obtained by varying only the elasticity of wages to output two standard errors below and
above its point estimate.99 For all possible combinations of this revenue elasticity α and
for a given spending elasticityγ, we impose the identification scheme as in (5.5) on the
VAR. For any of the parameter values in Table 5.6, we always find convergence to a
result identical to that obtained with the point estimate of the elasticity.100 The uncertainty
about the elasticity does not seem to play a major role then, and this confirms the findings
of Blanchard and Perotti (2002) or Marcellino (2002).
Table 5.6 Parameters γ and α
|
France________ |
Germany_____ |
Portugal_______ |
Spain_________ | |
|
Net lending |
0.53 [0.46, 0.61] |
0.51 [0.39, 0.61] |
0.46 |
0.44 |
|
Total spending, γ |
-0.11 |
-0.18 |
-0.05 |
-0.15 |
|
Total revenues, α_______ |
[0.51, 0.66] |
[0.46, 0.68] |
[0.44, 0.52] |
[0.43, 0.53] |
Source: Girouard and André (2005).
One of the other interesting scenarios is the one in which we switch off the elasticities. By
setting γ and α equal to zero, we assume that neither spending nor revenues react to the
cycle. This consequently attributes a larger role to discretionary fiscal policies. The effect
on the structural indicator depends however on the relative contribution of changes in
taxes or spending to fiscal shocks. Figure 5.6 contrasts the structural indicator obtained
with the OECD elasticities against the one with zero elasticities. The effect is only
marginal. In most periods, the results are rather similar. This reflects again the prevalence
of the supply relative to the temporary economic shocks. Oftentimes, there are more
prolonged periods of moderate deviations.
99
100
The wage elasticity is used for calculating the elasticity of the income tax. See Girouard and André
(2005) for an extensive discussion and a quantification of this uncertainty.
The results of the impulse response analysis are largely unchanged. Effects are estimated slightly less
precise, and the effects of the business cycle shock in Portugal are not clear.
148