2
2
period: Freedom House, Polity IV, and Vanhanen. The indices have been standardized
to a scale of 0 to 10 (with the latter figure representing the highest democratic value).
The scale is then divided into three intervals according to degree of democracy: 0 –
3.33; 3.34 – 6.67; 6.68 – 10. The score given the countries within each interval
according to the one index (the one italicized) has then been correlated with their
score according to each of the other indices. As can be seen, the correlations here are
generally lower. In some cases they are extremely low, or even non-existent. Even
negative correlations come into view in a few cases.
Looking at the three different intervals, we find the degree of concordance is
generally best in the highest one (r is .53 on average). It is clearly worse in the lowest
interval (r = .41), and worst of all in the middle (r = .29). Significant differences also
appear when different scales are matched against each other. When Freedom House is
matched against Polity, the correlations in the highest interval are on average very
strong (r = .73), while in the lowest one they are weakest (r = .25). When Vanhanen’s
index is matched against the other two, by contrast, a partly different pattern emerges.
When matched against Freedom House, the correlations are strongest in the lowest
interval (r = .47) and weakest in the middle (r = .38). As against Polity, on the other
hand, the correlations are strongest in the highest interval (r = .47) and weakest of all
in the middle (r = .13).
We find as well that it sometimes makes a big difference when the same indices are
matched against each other in reverse order (so that the selection of countries
becomes different). This is a result, in part, of the fact that the different cases are
variably distributed among the three intervals by the different indices. Polity has the
smallest number of cases in the middle interval throughout. As a consequence, the
correlations often become weaker when a selection is made in respect of the countries
found within this interval in Polity’s index. Vanhanen’s index, on the other hand, has
the largest number of cases in the lowest interval; this helps make the correlations,
when tested on the basis of this index, strongest in the scale’s lower section.
The conclusion is that, as a rule, extreme differences among countries will register in
much the same way on the three scales. Countries like Australia and Costa Rica will
find themselves at a similar distance from such countries as North Korea or Saudi
Convention