XREAP2007-14
robust test for sustainability can be based on the response of the fiscal surplus st to initial public debt
bt* , as in (1):6
st = ρbt* +μt . (1)
A strictly positive response of the government to debt developments is a sufficient condition for fiscal
policy satisfying the intertemporal budget constraint (Bohn, 1998). The basic intuition is that ρ> 0
in (1) implies that future debt is reduced by factor (1 - ρ)n at horizon n indicating compliance to the
budget constraint.7
Extending the test to different tiers of government, we can infer on a problem of sustainability in the
federal fiscal system. By testing the response of surpluses to debt, we avoid making strong
assumptions on the expectations of bail out, and may disregard all sorts of strategic interactions
between different tiers of government. The fiscal system is thus only a latent variable in our tests.
In particular, in a reaction function like (1), a strictly positive response of the general government
surplus to the debt stock is a sufficient condition for aggregate sustainability of the fiscal system.
Nonetheless, if at least one tier of government takes on the burden of debt of another government
level, overall sustainability can still be consistent with a soft budget problem. Direct bail out or
transfers by one government tier may compensate for the debt build-up of another government. Hence,
we need to disaggregate sustainability by inferring on the debt response for different government
levels. Let us assume there is a single federal government, and N-1 regions in the country. We
henceforth denote variables for the federal government level with subscript ‘1’ and the regions by 2,
... , N. There are then several ways to infer on sustainability for different government levels,
depending on the restrictions we impose on the interaction between several government levels.
What matters for aggregate sustainability is that all governments react to some extent to aggregate
6 Alternative tests of the intertemporal budget constraint have been suggested in the literature. These are usually
based on the time series properties of deficit and debt variables. Tests for the stationarity of debt and surpluses,
or the cointegration between spending and revenues, entail rather strict economic assumptions. There is in fact a
broad class of stochastic processes that violate these time series properties, but nonetheless satisfy the
intertemporal budget constraint (Bohn, 2007).
7 The only necessary assumptions are that the data generating process for fiscal policy is stationary and ergodic.
The residual μt is a composite of other determinants that in the aggregate are assumed to be bounded as a share
of gdp. it can be shown that that ρ> 0 also holds for fiscal policies that react in a non-linear way to debt
(Bohn, 2005). Likewise, it is sufficient that the condition applies infinitely often within sample (Canzoneri et al.,
2001). of course, ρ> 0 is not a necessary condition and hence there exist fiscal policies that violate this
condition but still are sustainable.