these to whole societies is disputable. Statistical manipulation of
existing sets of data about human societies may be a partial substitute
for experimental techniques, but it could be argued that few con-
vincing data sets exist. Some attempts at marshalling these include
the World Handbook of Political and Social Indicators(Taylor and
Jodice, 1983), and the Country Indicators for Foreign Policy Project at
Carleton University, Canada (www.carleton.ca/cifp). One very basic
problem for international data sets is that many countries do not have
reliable population figures, for example Nigerian census figures have
been politically contested because of their influence on the ethnic
balance of power. It is also difficult to compare financial values in
different currencies because of artificial exchange rates and diffe-
rences in purchasing power.
Scholars committed to a scientific approach to politics have sought
to overcome this problem by collecting quantitative data about poli-
tical behaviour. Classically this has been done through social surveys
which may be carried out on a large scale by market research firms, or
on a smaller scale by researchers themselves. These tend to be focused
upon voting behaviour and mass attitudes to political systems.
However, legislative voting patterns and the texts of newspapers,
political speeches, expenditures by governments and a wide variety
of other observations can also be treated as quantitative data to be
subjected to statistical analysis.
Modern statistical analysis is a very sophisticated discipline which
enables the researcher to make judgements on the existence, or not,
of significant associations between variables. These are commonly
assessed as being 95 per cent or 99 per cent unlikely to have occurred
by chance. However such questions as what proposition to investi-
gate, what variables to examine, whether to treat variables as depen-
dent or independent require an explicit or implicit theory of what is
happening to be tested. There is a logical gap between a statistical
association and a causal relationship which is what such researchers
generally aspire to (see John, 2002).
On a philosophical level it has been argued that the sort of causal
explanation that would be perfectly satisfactory in physical science
would be unsatisfactory in explaining social phenomena – social
explanations need to explain the motives of the persons involved,
not just predict successfully what will happen (Runciman, 1969).
Additionally, if we accept that human knowledge and motivation are14 POLITICS