estimated relationship, they find evidence that the constraint holds. Ahmed and
Rogers (1995) also indicate that Gt and Tt are difference stationary, even taking
account of Perron (1989) type breaks. Quintos (1995) makes a distinction between
weak and strong forms of sustainability, where weak sustainability is some
relationship between government spending and revenue without evidence of
cointegration, and strong sustainability is a one-to-one relationship and the existence
of a cointegrating relationship. In this context, no relationship between spending and
revenue suggests fiscal policy is unsustainable.6
However, to our knowledge, none of the above-noted studies consider the
interrelationships between global capital markets and fiscal sustainability. If global
shocks matter for fiscal sustainability, tests of fiscal sustainability in a panel setting
that do not fully account for any cross-sectional dependence are likely to yield the
wrong conclusions. Cross-sectional dependency, which if not accounted for, leads to
size distortion in standard panel unit root tests, and therefore presumably
cointegration tests, which bias them towards the alternative hypothesis of stationarity
(see O’Connell, 1998, Andrews, 2005, and Breitung and Pesaran, 2007).
Bohn (2007) recently levied a further criticism of sustainability tests based on
unit root testing and cointegration. Specifically, Bohn (2007) suggested that rejections
of sustainability based on stationarity and cointegration tests are invalid because in an
infinite sample, any order of integration of debt is consistent with a transversality
condition, which in turn implies that intertemporal budget constraint is always
satisfied. Bohn (1998, 2007) argues that more emphasis should be put on the
economics of the IBC and proposes an alternative means of testing the sustainability
of fiscal policy, based on the responsiveness of the primary surplus to the debt-GDP
6 See also Davig and Leeper (2005) and Thams (2007) for empirical evidence examining fiscal rules
using Markov Switching methods.