Spatial Models of International Conflicts: Extending Current Theories of War Behavior

Abstract

Abstract Previous studies on interstate conflict have shown that war is spatially contagious. These studies have, however, suffered from a limited definition of spatial contagion and have ignored other nongeographic explanations. In this paper, the effect of location on war behavior is tested using eight measures of spatial contiguity among 135 states. The tests support the idea that wars tend to cluster spatially, but the relationship is not as simple as had been believed previously. Examination of war data for African states showed that the geographic effect extends beyond first-order (immediate) neighbors. To analyze the global distribution of conflict, two regression models were tested. In the first which used structural predictors (political, social, economic, and military characteristics of the states), only the military variables were found to be significantly related to war, but the residuals showed positive spatial autocorrelation. An alternative regression model, with structural and spatial autoregressive terms, provided a better fit with uncorrelated residuals. The spatial predictor was more important than all other variables except military expenditures in explaining the global distribution of war. Key Words: war diffusioncontiguity indicesspatial autocorrelationstructural explanationspatial explanation