although these are conditional upon the evidence of cointegration as the absence of
cointegration would imply a spurious relationship.12
< Insert Table 2 here >
Explicitly taking account of the time series properties of the data we estimate Bohn
(1998) type response functions of the primary surplus on government debt, where a
positive and significant coefficient on debt (i.e. ρ>0) in equation (6) is taken as
supportive of fiscal probity. To ensure our results are robust, we report coefficient
estimates for β from equation (11) based on four different panel estimators: Fixed
Effects (FE), Bai and Kao (2005) CupFM and Bai, Kao and Ng (2007) CupFM and
CupBC. The fixed effects results are likely to be biased, as they are neither robust to
panel nonstationarity nor to cross-sectional dependence, which is evident in our data
set (see Table 1). Nevertheless, we report FE for comparison with Mendoza and Ostry
(2007), who support fiscal probity for emerging market and industrial countries.13 The
estimators of Bai and Kao (2005), and Bai, Kao and Ng (2007) are both robust to
panel nonstationarity and cross sectional correlation, but these estimators differ in the
way they control for cross-sectional dependence. Bai and Kao (2005), for example,
rely on a stationary common factor, while the common factor in Bai et al. (2007) can
be nonstationary. As we show below, the latter outcome is the case here. This makes
Bai et al. (2007) the appropriate estimator.
< Insert Table 3 here >
12 If our data were stationarity, our results would be conditional upon omitted stationary variables (see
Bohn, 1998).
13 Mendoza and Ostry (2007) find evidence that both industrial countries and emerging market
economies primary surplus is positively and significantly related to the level of debt, with a coefficient
of approximately 0.04. Overall they find little difference between the conduct of fiscal policy in
industrial countries and emerging market economies.
16