WAR AS A SUITABLE SUBJECT FOR STATISTICAL ANALYSIS

One of the crucial tasks for an analysis of warlike behaviour is the formulation of theories of how human beings actually behave in relevant situations. We are trying to discover the causes of war, though this involves finding the causes of a large number of ancillary phenomena also. To analyse such things meaningfully, we make hypotheses about what factors relate to war and other relevant phenomena. These may be ad hoc hypotheses which seem plausible, though we aspire to groups of hypotheses bound together in a deductive framework as described in chapter 2. These hypotheses are generalisations about classes of events and are not concerned only with specific, individual events. If we want to know whether arms races precede wars or not, or how certain sorts of alliance formation relate to war, then we examine instances of these phenomena and endeavour to make appropriate generalisations. That is, we hypothesise some relationships and then test them. This is particularly critical when we come to talk of policy. If we wish to know whether a particular alliance pattern is likely to promote stability and peace or otherwise, we approach it as an instance of a generalisation about alliances and war. Choices are made in the light of expectations about the future: hence we need rational bases for these expectations.