Please be invited to a brownbag on Monday 28th of November from 12:00-13:30, in the Peace Room. Henrikas Bartusevicius will present the paper “Improving the Large-N Analysis of Conflict Causes: A Categorical Disaggregation of Intrastate Armed Conflict.” Henrikas Bartusevicius is a visiting researcher here at PRIO from Aarhus University.
Abstract: Intrastate armed conflicts or civil wars are often analyzed as homogeneous phenomena in large-N research. Conflict events are packed into one category and tested against dozens of predictor variables in regression models. However, recent studies suggest that some variables have a non-uniform effect on the outbreak, dynamics and termination of different categories of conflicts (e.g., ethnic and non-ethnic). This implies that the aggregate models that pack all categories of conflicts under the same heading could significantly underestimate or overestimate the effects of predictor variables on particular categories of conflicts. The present study has tested this implication systematically. First, it disaggregated intrastate conflict into four less-abstract categories: ethnic revolutionary, ethnic territorial, non-ethnic revolutionary and non-ethnic territorial. Second, the study tested these categories in multinomial regression models of conflict onset against a number of commonly employed variables – ethnic diversity, regime type, economic prosperity, conflict history, and energy resources. The models displayed sharp differences in the effects of the predictors on the different categories of conflicts. Analysis showed that ethnic fractionalization only affected ethnic revolutionary conflicts while ethnic polarization only affected ethnic territorial ones. The models also showed that regime type and economic prosperity proxies only had an effect on non-ethnic revolutionary conflicts. Finally, the analysis displayed that effect of conflict history is only significant in the models of ethnic conflicts. On the basis of these results, this study argues that introducing different categories of conflicts into large-N research could substantially improve the validity of the findings.