- is the fiscal policy shock and reflects discretionary changes in the fiscal policy stance.81
We take elasticities for government expenditures (γ) and revenues (α) with respect to
output, and impose these values on the relation in A between the reduced form residuals
for output (εy) and spending (εg) respectively revenues (εt).82 The fiscal shock thus
includes discretionary decisions unrelated to the cycle. Moreover, any government policy
that interferes with the workings of automatic stabilisers on a systematic basis is
considered as a fiscal intervention. Unlike other VAR studies, we split an overall change
in fiscal policy into a part that has a short-term economic effect (the fiscal ‘demand’
shock), and into shocks that may have potentially long-term growth effects (the ‘supply’
shock).83
One important limitation of the current version of the model is that we cannot tell apart
the growth effects coming from ‘pure’ technology shocks from those deriving from tax
and spending decisions. Our supply shock is thus a combination of all shocks with long-
term output effects. The negative effects of distortionary taxation or incentive-distorting
spending show up in this shock, as well as the possibly positive effects of government
investment. Instead, we isolate in the fiscal ‘demand’ shock only those changes in the
discretionary budget stance that have temporary effects on output. A full-fledged analysis
of the economic growth effects of fiscal policy would require additional restrictions.84
The current identification is sufficient though for the purpose of deriving a fiscal
indicator. We summarise our assumptions in (5.5) (see also Table 5.2):
•00
C(1) =
and A = • γ α
(5.5)
We can not simply set to zero the elasticity γ of government expenditures.
Unemployment benefits move over the cycle in EU-countries, even if their contribution to
variation in total spending is not large. The parameter γ comes directly from the
elasticities calculated by the OECD that we reported in Table 5.1. Instead of multiplying
each revenue category by its cyclical elasticity and GDP share, we have subtracted the
spending elasticity (row 2 in Table 5.1) - accounting for its share in GDP - from the
elasticity of total net lending (row 7 in Table 5.1) so as to obtain the total elasticity of
revenues α . The coefficients do not sum to zero as the budget is assumed to be
countercyclical. Table 5.3 summarises our parameter assumptions.
81
82
83
84
This is not a replication of the results in Blanchard and Perotti (2002) as they require additional
short-term constraints on the timing of the effects whereas we consider long-term constraints.
We therefore need to impose two different coefficients γ andα which results in one
overidentifying restriction. Blanchard and Perotti (2002) instead net out the cyclically sensitive
transfers from spending, and assume a zero elasticity on other spending categories. As the sensitivity
analysis in section 4.4 demonstrates, this does not seem to affect our results.
For this reason, we do not expect responses to our fiscal shock to be similar to those documented in
the empirical literature. Our distinction is more consistent with the theoretical models of fiscal
policy.
We make some suggestions in the concluding section. We considered the effect of loosening the
long-term constraint on either government spending or revenues. We could not reject longer-term
effects of fiscal shocks, endorsing the hypothesis that supply side effects of fiscal policy decisions
affect the ‘supply’ shock.
131