Learning, realism, and alliances: An empirical examination of the causes of alliances.

Abstract

This dissertation asks the questions, whether and how do states learn in international relations. It builds a theory of learning in international relations, drawing primarily on principles of social psychology and organization theory. The thrust of the theory is that lessons are drawn infrequently, primarily after formative events such as world wars. The theory is tested on the alliance behavior of minor powers in the twentieth century against the predictions of a leading realist explanation of alliance formation, balance of threat theory. Learning proposes that the post-world war alliance behavior of minor powers will be determined by wartime experiences, while balance of threat proposes that alliance behavior is determined by the current level of international threat. Two learning hypotheses are tested: one assumes that lessons are drawn only from a minor power's own experiences, and the other assumes that lessons are drawn on the basis of the wartime experiences of all relevant minor powers. These theories are submitted to quantitative tests using a logit model, in which the dependent variable is whether or not the minor power in question preferred alliance with a great power during a particular year. The quantitative findings strongly support learning theory. The learning hypothesis assuming that minor powers focus only on their own experiences did particularly well, correctly predicting the dependent variable 87% of the time. The learning hypotheses are both statistically and substantively significant, and the realist variables have at most marginal effects on the dependent variable. The results bear out even in the face of a number of robustness tests in which the specification of the model is slightly altered. A number of cases are examined in depth to assess the validity of the theory's causal explanations, and they also bear out the learning theory as the more powerful explanation. Finally, all cases for which the learning theory did not make the correct prediction are examined in depth, as a means of furthering the theory-building process.