Recent large-N studies of civil war conclude that inequality does not increase the risk of violent conflict. This article argues that such conclusions may be premature because these studies, which usually test the conflict potential of `vertical inequality' (i.e. income inequality between individuals), tend to neglect the group aspect of inequality. Case studies suggest that what matters for conflict is a concept closely linked to both economic and ethnic polarization: `horizontal inequalities', or inequalities that coincide with identity-based cleavages. Horizontal inequalities may enhance both grievances and group cohesion among the relatively deprived and thus facilitate mobilization for conflict. This article provides a quantitative test of this argument, exploring whether various forms of polarization and horizontal inequalities affect the probability of civil conflict onset across 36 developing countries in the period 1986—2004. National household data from the Demographic and Health Surveys (DHS) are used to construct measures of ethnic, social and economic polarization, as well as vertical and horizontal inequalities along two dimensions: social and economic. The article also introduces a combined measure of ethnic/socio-economic polarization as an alternative to the horizontal inequality measure. Robust results from panel and cross-section analyses show that social polarization and horizontal social inequality are positively related to conflict outbreak. Variables for purely ethnic polarization, inter-individual inequalities and combined ethnic/socio-economic polarization are not significant.