As yet, there is no universally accepted technique for decomposing the primary balance to
GDP ratio into its cyclical and discretionary components, although representative discussions
of the merits and drawbacks of various approaches can be found in Gramlich [1990],
Bouthevillain and Quinet [1999], Bruni and Tujula [1999] and Chalk [2002].
3.1 The Blanchard fiscal impulse
We take our lead from Alesina and Perotti [1995], Alesina, Perotti and Tevares [1998] and
others in employing a method due to Blanchard [1993]6 to construct fiscal impulses, i.e. the
discretionary changes in fiscal policy from one year to the next. The relative size of this
impulse is then used to identify years in which significant consolidation attempts have been
undertaken. The method is simple, transparent and appears to generate reliable results.
In practice, the Blanchard method extracts cyclical movements in the primary balance using
individual regressions for each component of the primary balance that incorporate the country
specific unemployment rate and a set of deterministic variables (a constant, and a
deterministic trend)7. Predicted values for revenues and transfers are then calculated
conditional on the previous year’s unemployment rate (i.e. by replacing the contemporaneous
unemployment rate with its lagged value in the estimated equations) and this allows us to
construct a predicted primary balance based on an unchanged unemployment rate. The
Blanchard measure of the structural fiscal impulse is then calculated by subtracting the
predicted cyclically adjusted primary balance from its actual value.
3.2 Identifying General Government Fiscal Consolidation Attempts
Having generated a set of discretionary fiscal impulses it is necessary to introduce an
operational definition of a fiscal consolidation attempt. Since there is no 'script' that
governments follow in consolidation attempts, or indeed a comprehensive and accurate record
of when governments instigated changes in their policy stance with the intention of improving
their fiscal balances, the standard approach is to use some kind of subjective criteria (see for
example McDermott and Wescott [1996], Kamps [2001], Giavazzi et al. [2000], OECD
6 A commonly used alternative is to cyclically adjust the primary balance by applying the Hodrick Prescott filter
(see for example Bruni and Tujula [1999] and Roger and Ongena [1999]). Bruni and Tujula [1999] report results
based on both the Blanchard and HP methods and judge the Blanchard method to be superior in identifying periods
of tight fiscal stance that accord with the consensus of commentators.
7 Our general regressions included unemployment, a constant, time and time squared. For each regression, the best
performing specification was chosen on the basis of F and t-tests assessing the individual and joint significance of
the contribution of the time and time2 terms. Note these regressions are not intended to provide a full explanation
of the behaviour of the fiscal balances, the deterministic terms are included purely on statistical grounds and reflect
the requirement that the regressions are reasonably well specified and that the equation residuals are stationary.