Restless dyads: revisiting onset and escalation in militarized interstate disputes

Official URL: http://192.168.1.20/record=b1505249 (Table of Contents)

Modeling escalation of disputes constitutes a major cornerstone in international conflict studies. Many of these studies assume that incidents in a dispute follow a regular path where use of force is preceded by lower levels of hostile action such as threat or display of force. Therefore, these studies treat a dispute as having escalated if force was used. A close examination of the MID IP (Militarized Interstate Disputes Incident Level) dataset suggests inaccuracies may exist in this assumption. The MID IP dataset indicates that parties directly utilize force (i.e. without a preceding threat or display of force) in significant number of disputes. That about 40% disputes between 1993 and 2001 directly start with use of force at the MID Onset level echoes warnings made by Diehl (2006) and Dixon (1993). We identify cases in which the first action of a dispute is use of force ''direct escalations.'' This thesis examines whether this discrepancy poses a threat to the validity of empirical tests of escalation models. To that end, we replicate a comprehensive set of escalation models adopted from Braithwaite and Lemke (2011). More specifically, we compare the models with and without direct escalation cases in the sample of disputes. Results from Heckman selection models indicate that when direct escalations are excluded from the sample, being a joint democracy loses its pacifying effect for dyadic MID onset, although its pacifying effect increases for escalation of these disputes. Therefore, we can say that if democratic dyads try to communicate before involving direct use of force, they might find peaceful resolutions for their problems even at the escalation level. Further analysis suggests that territorial controversy and geographical contiguity significantly increase the likelihood of direct escalation. Finally, results also suggest that direct escalations may partially be an artifact of biased data collection; poorer/peripheral dyads carry a higher likelihood of experiencing a direct escalation.