The results generated by the panel regressions point to broadly similar regression
models across the three rating agencies (see Tables 4, 5 and 6). In view of the analytical
considerations above the discussion will focus on the random effects estimations. This
is supported by the Hausmann tests reported at the end of each table pointing to the
acceptability of the random effects approach. Nevertheless, we also report the pooled
OLS and the fixed effects results for completeness and comparison purposes.
[Insert Tables 4, 5 and 6 here]
We report the results of two models for each of the rating agencies, the unrestricted and
the restricted model. While the unrestricted model incorporates all variables discussed
above, the restricted model contains only the variables which were found to have a
statistically significant impact. Although the sequence of excluding individual variables
in moving from the unrestricted to the restricted regression can have an impact on the
final specfication, the restricted models presented in the tables are quite robust to
alternative exclusion procedures. As can be seen from the statistics reported at the end
of each table, the explanatory power of the models is very high with R-square values
around 95 per cent and it remains almost constant moving from the unrestricted to the
restricted versions, while the number of observations increases marginally. In addition,
the variables found to be significant in the unrestricted model generally remain
significant with the same sign in the restricted version.
The restricted models reveal a homogenous set of explanatory variables across agencies.
On the real side, GDP per capita and GDP growth rates turn out significant for all three
companies. In the fiscal area, this applies to the government debt ratio as a difference
from the average and to the government effectiveness indicator. On the external side,
the average external debt ratio and the average level of reserves are found to be
significant across agencies. Default, EU and industrial country dummies are also
significant for all agencies. Moreover, the size of the coefficients is of the same order of
magnitude and they have the expected signs. In particular, the level and growth rate of
real income drive up the rating, government and external debt have a negative impact
and government effectiveness and higher external reserves have a positive impact.
17