Parallel event sequences in international crises

This study presents a theory of foreign policy decision-making based on event structures that would account for the ability of foreign policy experts to make short-term predictions in the presence of incomplete information and noise. An algorithm is devised that constructs subsequences of events that occur in international crises coded in the Behavioral Correlates of War (BCOW) data set. Four subsets of BCOW crises are considered: nonwar crises, pre-World War I wars, post-World War I wars, and a mixture of nonwar and war crises. About six common subsequences are found in each of the sets; these cover about 35% of the events in the crisis sequences after the sequences have been filtered on the basis of novelty. The subsequences differ substantially between the subsets of crises and exhibit a degree of internal consistency. The subsequences also can be used to differentiate between nonwar and war crises by using the fit of each subsequence to construct a vector characterizing the crisis, then clustering on the basis of those vectors.