1
1
Over the course of the last hundred years, great political changes have taken place
across the globe. Many countries have been democratized. In order to document these
developments, scholars have developed a number of indices of democracy. Not
surprisingly, the criteria for democracy underlying the various indices differ in some
respects. Methodological variations are also found. Some indices are based on a large
number of criteria which are then combined; others have only a small number of
criteria. The method for coding varies as well. Some indices are dichotomous,
distinguishing only between democracy and non-democracy. Others are trichotomous,
with a third category coming between democracy and the purer forms of dictatorship.
Still others identify a larger number of forms of government, variously placed along a
continuous and more finely graduated scale.
Notwithstanding these differences, the various indices show a striking degree of
concordance. When they are matched against each other, the correlation co-efficient
(r) tends to lie at the 0.75 – 0.95 level. One could take this to indicate that the choice
of index scarcely matters: all indices do much the same job. However, that would be a
too sanguine conclusion. Elkins (2000) has found that it is only when a graduated
scale for measuring democracy is used that the oft-cited correlation between peace
and democracy is confirmed; dichotomous methods, by contrast, do not yield such a
result. Casper and Tufis (2003) have considered how different measurements of
democracy along a graduated scale co-vary with different explanatory variables. It
proves to be the case that, among other things, the well-known correlations with
socio-economic variables actually fluctuate considerably—depending on how
democracy is measured (for other examples, see Teorell and Hadenius 2004).
It appears to make a difference, then, which index is used. Notwithstanding the strong
correlations among them, the various indices do not do the same job. This may seem
surprising. It bears recalling, however, that the statistical concordance is found only
on a rough and overarching level. By contrast, if we look at how countries are graded
at different segments of the democracy scale, we sometimes find large divergences.
Examining the Concordance
Table 1 shows the relationship between different placements along a scale of
democracy for three continuous indices with international coverage over a substantial
Convention