The name is absent

models presented above, rating agencies base their decision to a considerable extent on
projected economic developments. Thus, a full empirical model of the agencies’
approach would need to incorporate the agencies’ expectations regarding the relevant
explanatory variables. However, as the agencies generally do not publish their
projections, any such modelling attempt would remain highly tentative. Still, the
observation that many of the actual rating changes are predicted by the models with a
lag of one or two years appears to support the relevance of this point. Second, ratings
agencies also generally make a clear point that they cover qualitative variables in
addition to quantitative data in the ratings process. While the relative importance of the
qualitative and quantitative factors that enter the ratings decision is uncertain (and might
well vary across countries), rating agencies’ public statements indicate that such factors
can play an important role.⁵ In the models above, by contrast, the only variable
reflecting such considerations is the government effectiveness indicator and it thus
appears likely that in these models the impact of qualitative factors is under-represented.

The most noticeable difference between the models is not the number of corrected
predicted changes but the total number of predicted changes. In fact, the ordered probit
and random effects ordered probit predict significantly more changes than the OLS and
random effects counterparts. For instance, for S&P, while both OLS and random effects
predict around 79 upgrades and 50 downgrades, the ordered probit model predicts 102
upgrades and 64 downgrades.

4.5. Examples of specific country analysis

In terms of the magnitude of the coefficients, the comparison between the ordered probit
and random effects ordered probit is not straightforward because the estimated distances
between the categories are different. But in general, once this is accounted for, by
standardising the coefficients in relation to the average jump, they are both in line with
the linear panel results. An improvement of 2 percentage points in the budget deficit, a
reduction of 5 percentage points in public debt, or a higher GDP growth by 3 percentage
points, all have the same relative impact on the ratings between 0.1 and 0.2 notches. An
increase of 10 per cent in GDP per capita would improve the rating by 0.15 to 0.25

⁵ For example, see Rother (2005) for an analysis of the impact of EMU convergence on country ratings in
eastern Europe.

26

<<   23   24   25   26   27   28   29   >>

More intriguing information