Advances in computing and machine learning have enabled researchers to use many different tools to learn from data. This dissertation is
devoted to using predictive modeling to learn from existing data in international conflict studies with the aim of offering new measures and
insights for applied researchers in international relations. In the first chapter, I explore the expected cost of war, which is a foundational
concept in the study of international conflict. However, the field currently lacks a measure of the expected costs of war, and thereby any
measure of the bargaining range. I develop a proxy for the expected costs of war by focusing on one aspect of war costs - battle deaths. I train
a variety of machine learning algorithms on battle deaths for all countries participating in fatal military disputes and interstate wars between
1816-2007 in order to maximize out of sample predictive performance. The best performing model (random forest) improves performance over that of
a null model by 25% and a linear model with all predictors by 9%. I apply the random forest to all interstate dyads in the Correlates of War
dataverse from 1816-2007 in order to produce an estimate of the expected costs of war for all existing country pairs in the international
system. The resulting measure, which I refer to as Dispute Casualty Expectations (DiCE), can be used to fully explore the implications of the
bargaining model of war, as well as allow applied researchers to develop and test new theories in the study of international relations. In the
second chapter, I use these expected costs of war to explore another foundational concept in international relations: foreign threats.
Researchers commonly theorize about the impact of a state's international security environment - that is the extent to which a state is
threatened by other states - yet the field currently lacks a measure which can effectively proxy for expectations of conflict. In order to
create a new measure of threat, I train a number of machine learning algorithms on fatal militarized disputes over the years 1870-2001. I
aggregate the predictions from these models at the country level to create a new measure of international conflict expectations for all states.
In so doing, I am able to revisit the causes of international conflict via a data-driven approach, as well as provide a new measure of foreign
threat for applied researchers. Finally, in the third chapter, I make use of this new measure to assess how international security affects a
state's human rights behavior. International relations scholars have increasingly relied on domestic institutions to explain international
conflict but less work has focused on reversing the arrow. To this point, political violence scholars have principally relied on domestic
factors to explain the conditions under which leaders use coercive means to maintain power. But, political leaders do not exist in a vacuum;
their decision making is informed by international and domestic factors. Therefore, I rely on both a predictive and inferential approach to
assess whether foreign threats matter for state repression. The measure of foreign threats does emerge as an important variable in predicting
state repression, which suggests that there is a meaningful relationship between international security and human rights behavior. Additionally,
I find some (limited) evidence that the measure is negatively related to human rights behavior: states with high levels of foreign threat are
associated with higher levels of state repression. But this finding is sensitive to model specification and merits further
inspection. / A Dissertation submitted to the Department of Political Science in partial fulfillment of the requirements for
the degree of Doctor of Philosophy. / Fall Semester 2018. / November 16, 2018. / International conflict, Machine learning, Predictive modeling / Includes bibliographical references. / Mark Souva, Professor Directing Dissertation; Jonathan Grant, University Representative; Robert J. Carroll,
Committee Member; Sean Ehrlich, Committee Member.
Identifer | oai:union.ndltd.org:fsu.edu/oai:fsu.digital.flvc.org:fsu_661145 |
Contributors | Henrickson, Philip Edward (author), Souva, Mark A. (professor directing dissertation), Grant, Jonathan A., 1963- (university representative), Carroll, Robert J. (committee member), Ehrlich, Sean D. (committee member), Florida State University (degree granting institution), College of Social Sciences and Public Policy (degree granting college), Department of Political Science (degree granting departmentdgg) |
Publisher | Florida State University |
Source Sets | Florida State University |
Language | English, English |
Detected Language | English |
Type | Text, text, doctoral thesis |
Format | 1 online resource (175 pages), computer, application/pdf |
Page generated in 0.0015 seconds