The discrete sequential analysis of dynamic international behavior

Abstract

This study reexamines the dynamic patterns of US-Soviet foreign policy behavior using recently developed methods for the sequential analysis of categorical data. The methods proposed here have several advantages over conventional regression based modeling procedures. First, assumptions about the role of time in foreign policy behavior and decisionmaking processes are more realistic in that time is represented as an ordinal sequence on human experience. Second, these methods better approximate the causal structure of actionreaction theories by delineating the temporal order of stimulus and response, and by focusing attention on individual behaviors rather than aggregate behavioral constructs. Third, the probabilistic nature of action-reaction hypotheses are made explicit through the use of logit transformed conditional probabilities. Finally, by accommodating model specifications that are virtually isomorphic to traditional action-reaction modeling efforts, these methods retain the flexibility and interpretability of standard regression approaches. Application of these methods to US-Soviet interaction revealed a marked asymmetry in the superpowers' responsiveness to one another's foreign policy behavior during the 1966–1978 period. According to these analyses, Soviet behavior is more clearly a function of its own prior activities than a response to the United States whereas American behavior is far more dependent on prior Soviet actions than its on own.