papers that investigate the role of nominal versus technology shocks in economic
fluctuations (Nelson and Plosser, 1982; King et al., 1991; GaH, 1992).
Second, the variation in the structural balance is different from that in traditional two-step
methods. This discrepancy owes to the definition of structural balance. This is perhaps
best illustrated with an example. Consider a tax cut, for a given level of government
spending and exogenous output. This would lead to a deficit, ceteris paribus. If fiscal
policy indeed has real economic effects as the empirical literature suggests, then the tax
cut temporarily boosts output. As a consequence, tax revenues will increase and spending
on unemployment benefits decrease, and the budget surplus will rise. The traditional
measure for cyclical adjustment takes out all cyclical variation, also the one induced by
fiscal policy, which leads to an overstatement of the structural balance. In our approach,
we control for this economic effect of the tax cut. The SVAR-model excludes that part of
the variation in GDP due to discretionary fiscal measures whereas the conventional
models take total output variation into account. But our approach goes even one step
further. Imagine that the tax cut also raises potential output in the long term. This widens
the gap between actual and potential output at the moment the fiscal shock occurs.
Structural balance would be improved as the increased tax base (now, and in the future)
makes the fiscal position more sustainable. Similar arguments can be made for the effects
of spending. As a consequence, our indicator of structural balance does not necessarily
display a smaller variation than traditional indicators. This will particularly be true if (a)
the indicator is mainly driven by fiscal or supply shocks; or (b) if the underlying
economic shocks we retrieve are more volatile than what conventional output gap
measures suggest.
Our model-based indicator has some favourable properties in comparison to more
conventional measures. First, the long-term constraints hold the promise of imposing
fewer contentious restrictions on the short-term effects of the fiscal shocks. Any
anticipation effect and the contemporaneous reactions of fiscal balances to economic
conditions are not constrained. Second, the simultaneous determination of a measure of
cyclical output and fiscal balance is internally more coherent. While the method is
definitely more complex, total uncertainty is quantified. We impose a minimal set of
economic restrictions and the validity of these assumptions can be discussed. As the
empirical model is also consistent with recent DSGE models of fiscal policy, these
assumptions can be tested. Sensitivity analysis can make clear the weakness of the model
in some specific direction. Moreover, progress in theoretical models of fiscal policy can
lead to further refinements of the approach. Third, by adopting an economic - and not a
statistical - method, the end-point problem of filters is eliminated. The indicator gives
timely information on changes in the fiscal stance.86 Finally, our indicator is also more
relevant for the assessment of fiscal policy. Our measure indicates better the change in the
stance of fiscal authorities, also with a view to growth effects and long-term
sustainability.
At the same time, the econometric approach suffers from some weaknesses. First,
extensions are difficult as the method is rather data demanding - at least in the time series
dimension. The annual frequency of the data may lead to some difficulties in the
identification of business cycle shocks, for example. Second, the gains of loosening the
constraints of short-run effects of fiscal policy have to be set off against some additional
The inclusion of structural breaks remains problematic, however. But in contrast to statistical
methods, the economic consequences of one-off fiscal events are modeled in our approach.
133