Several ideas pertinent to the requisites of democracy have been reviewed and subjected to empirical testing in the preceding pages. The chaff has thereby been separated from the wheat. The method of this division took the form of a stepwise regression. The possible explanatory factors which already at the first scrutiny could not display significant association – or which lost this quality in later stages – were then disregarded. However, a problem is inherent in this approach. The risk is that those which fall short of the limit would have done better if they had been included at a later stage of the analysis. Indeed, it usually happens that on controlling for more variables, associations either persist (by and large) throughout or decline to an ever-marked extent. This circumstance motivated the layout chosen here. But there are, of course, exceptions; those which appear ‘dead’ can ‘revive’ when further attributes are included in the regression.

Moreover, it is not self-evident to demand the significance level – at least 0.05 on double-sided (two-tailed) testing – which is here applied. In some similar studies a lower limit was used (e.g., in the form of only a one-sided test at the said level). Thus, it would be worth trying to find out whether the final result is essentially changed if a lower standard of significance is used.

For this reason, we have performed a series of regressions, where from the start all the attributes of any interest were included. Variables such as the percentage employed in service, trade with the EC, fragmentation, island states, colonial background, etc.