Bath Papers in International Development and Wellbeing 9
The econometric study of civil war is increasing recognized to suffer from problems of 'over-aggregation'. As such, there is a high risk of estimation biases, ecological fallacies, and endogeneity problems. In this paper, I seek to contribute to the disaggregation of the study of civil war by focusing on the socio-economic dynamics of secessionist conflict as an identifiably distinct subset of 'civil wars', and by using a new subnational dataset compiled for this purpose. I test a series of hypotheses relating to the socio-economic conditions that encourage secessionism and political institutions that might mediate it. In contrast to the mainstream literature on civil war, I find a very strong predictive role for a measure of ethnic diversity in accounting for the incidence of secession. I also find a relatively straightforward set of socio-economic relationships. The relationship between relative socio-economic performance and conflict incidence is non-linear: regions that suffer from high 'horizontal inequalities'-whether relatively poor or relatively rich-in relation to the rest of the country are more prone to secessionism. The presence of hydrocarbon deposits also dramatically increases the likelihood of secessionism. But the institutional story is more complex and contingent upon interaction effect with the degree of ethnic diversity and the level of horizontal inequality.