For example, how well do ‘voter turnout’ or ‘degree of multi-partyism’ rep-
resent ‘political development’ as an independent variable? How satisfactory
is the total of US loans and credit in a single year as an indicator of the
external influences that might encourage a military take-over? The weak
conceptual foundation of much multivariate analysis can result in unjustifi-
able conclusions being drawn from the statistical results.
Thirdly, measures used as dependent and independent variables should
cover the same time periods, otherwise temporal variation introduces prob-
lems of ‘simultaneity bias’. How, for example, can it reasonably be claimed
that ‘party fractionalization’ in 1975 caused coupsstarting fifteen years
earlier, especially when the military commonly banish political parties and
therefore determine themselves the level of pluralism (or ‘fractionalization’)
in the political system?
Fourthly, there are tautological pitfalls that await the quantitative analyst,
when some part of what is to be explained is contained within the definition
of the explicandum(the explanatory or independent variable). An example is
when ‘increased military expenditure’ is used as an independent variable but
has resulted from military domination of the state, the dependent variable
that is to be explained by expenditure on the military. The same problem is
found in the attempt discussed later to explain military coupsby reference to
the level of the political culture in a society, when political culture is defined
by reference to a lack of consensus, evidence for which is found in the fail-
ure of the civilian system to withstand military supplantment.
The use of doubtful indicators of the variables under consideration and the
unwillingness of researchers to replicate each others’ methods in different
regions undoubtedly account, along with the problems of correlational analy-
sis already mentioned, for the widespread dissatisfaction with this mode of
analysis. However, it would be wrong to be too dismissive, either on the
grounds that there are no identifiable patterns of military intervention – that it
is totally random – or that statistical comparisons and correlations force the use
of indicators of political, social and economic development that are of doubt-
ful validity. Quantitative analysis generate interesting hypotheses. The fact
that so far they have produced nothing more conclusive and leave us still need-
ing a convincing theory – and are probably incapable of meeting these needs –
possibly explains some of the frustration and irritation which they engender.
Forms of intervention
The main focus of academic interest has been supplantment – the act of tak-
ing political control by force and replacing civilian institutions with military
176 Understanding Third World Politics