Abstract
This paper highlights the fact that different distributional aspects of ethnicity matter for conflict. We axiomatically derive a parametric class of indices of conflict potential obtained as the sum of each ethnic group’s relative power weighted by the probability of across group interactions. The power component of an extreme element of this class of indices is given by the Penrose–Banzhaf measure of relative power. This index combines in a non-linear way fractionalization, polarization and dominance. The empirical analysis verifies that it outperforms the existing indices of ethnic diversity in explaining ethnic conflict onset.