In general the ratings attributed by the three agencies are quite similar. As shown in
Table 3 around 50 per cent of all observations have the same pair-wise rating. It is also
interesting to notice that S&P and Fitch have much closer ratings, and Moody’s is more
divergent with a significant number of observations having a distance of two notches
vis-à-vis the other two agencies. This might indicate, for instance, that Moody’s and
S&P give different weights to different indicators or simply reflects the uncertainty in
measuring the default risk.
[Insert Table 3 here]
For the present study we limited the sample to 1995-2005 because of data availability of
explanatory variables. The variables inflation, unemployment, GDP growth, fiscal
balance and current account entered as a 3 years average, reflecting the agencies’
approach to take out the effect of the business cycle when deciding on a sovereign
rating. The external debt variable was taken from the World Bank and is only available
for non-industrial countries, so for industrial countries it was attributed the value 0,
which is equivalent to having a multiplicative dummy. As for the dummy variable for
European Union, we consider that the rating agencies anticipated the EU accession.
Thus we tested the contemporaneous variable as well as up to three leads. We find that
for Moody’s and S&P the variable enters with two leads, while for Fitch we find no
anticipation of EU accession. (See Appendix 1 for a full list of variables used in the
estimations as well as their specification and data sources.)
Regarding the estimation procedure, starting out with the broadest possible set of
variables, we sequentially dropped those that did not reveal any explicative power
(export growth, investment, trade openness, domestic credit growth, interest payments
and also most of the regional dummy variables).
4.2. Linear panel results
4.2.1. Full sample
16