Abstract
Our paper examines the question of when conflicts expand and what leads particular states to join more quickly than others. Using factors highlighted in the conflict expansion and joining literatures, we derive hypotheses about how these factors make resolve in the face of credible threats more likely and how this either increases or decreases the time to conflict expansion. We also generate expectations about when specific states are likely to join a conflict. We test our expectations using a dataset of the initial belligerents of militarized disputes and all potential joiner states. The results of our analyses suggest that conflict expansion is more often because of initiators' resolve rather than miscalculation since observable signals of likely expansion, such as alliances and power, decrease the time to expansion. Our findings have important implications for research on alliance reliability, balancing and bandwaggoning, and various proposed causes of the democratic peace.
Acknowledgments
We would like to thank three anonymous reviewers and Randolph Siverson, Scott Gartner, and Daniel Kono for their helpful comments and suggestions. Jacob Hogan, Andrea Morrison, and Sam Snideman provided helpful research assistance. The name ordering is reverse alphabetical, and does not denote unequal contributions.
Notes
1According to the MID data, slightly over 15% of conflicts expand beyond the initial disputants.
2We use the term outcome rather than winning because it is not clear that states always join disputes in hopes of one side achieving victory over the other. It may also be the case that third parties join to manage the conflict or bring an end to hostilities regardless of the winner.
3Of course, we recognize that states may aid states through means other than direct participation in a conflict. For example, prior to the U.S. entry into World War II, the U.S. aided the Allies through Lend-Lease and other programs. We hope to examine aid short of formal military commitments in future research.
4We code contiguous as either bordering or separated by no more than 150 miles of water.
5Because the parameter estimates of the Weibull model are almost identical to those of the Cox model we show only Cox results. Results are available from the authors.