Territorial Changes and the Treatment of POWs: Evidence from the Korean War

• In Event: Bringing IR Back in: The International in Conflict Processes

Abstract

Under what condition, do warring parties target prisoners of war (POWs)? Although the laws of war, such as the 1949 Geneva Convention, require humane treatment of POWs, abuse and arbitrary killings of POWs still have prevailed in recent conflicts. For instance, during the civil war in Iraq between the ISIS and Iraqi forces, numerous instances were reported of killings of POWs by both parties, such as beheadings, mutilations, and prisoners thrown from buildings.

Despite such a continuation of ill-treatment of POWs, only a few rigorous academic endeavors have been devoted to the topic. While most existing research on POWs has been descriptive research, statistical studies conducted by James Morrow and Geoffrey Wallace were probably the two most systematic investigation on the subject. Their research, however, left within-case dynamics of violence unexplored by focusing solely on cross-case variations. That is, even though their frameworks specify types of groups which are more likely to perpetrate killings of POWs compared to other groups, they do not shed light on the specific conditions under which the groups are more likely to target POWs in a given war. To narrow this gap, this project investigates the spatiotemporal condition for the ill-treatment of POWs by conducting statistical analysis with a focus on the Korean War, during which a total of nearly120,000 soldiers were taken prisoner by the United Nations Command forces and the Communist forces.

I argue that territorial changes are the main condition under which armed groups target POWs. Armed groups in war usually do not target POWs due to fear of losing their international reputation. Groups, however, may have the incentive to kill the prisoners when in retreat. When retreating, groups need to relocate POWs in the contested area to the rear. Yet relocation is difficult to employ due to the limits of logistics and transportation resources. The provision of sufficient transportation and rations is difficult, even for an army’s own soldiers, when it is one the verge of defeat; the resources should be distributed first to the military personnel and core civilians, such as government officials. Also, bringing a large number of prisoners to the rear makes the withdrawal slow and risky. Thus, killing POWs rather than resettling them becomes a quick and dirty solution, especially when groups are making a hasty retreat. Hence, it is hypothesized that as the pace of retreats of an armed group increases, the group is more likely to victimize POWs.

To evaluate the suggested hypotheses, I apply logistic regression models to an original panel dataset with 6,156 observations, which was created based on archival documents, oral history records, and Truth and Reconciliation Commission reports. The dataset covers military, political, and geographical aspects of the Korean War during each month of the war in each administrative county of South Korea.

This paper aims to make various academic and policy-relevant contributions. The paper contributes to political violence scholarship by developing a theory that explains the micro-dynamics of killings of POWs. It also makes an empirical contribution to the scholarship, which suffers from a lack of reliable data, by creating an original large-N dataset on the treatment of POWs during the Korean War. As for practical contributions, the paper provides an analytic framework that predicts the incidence of killings of POWs to help prevent violence in ongoing militarized conflicts.

Author

• Seung Joon Paik, George Washington University