We investigate whether civil wars have economic causes. The model is based on utility theory, rebels will conduct a civil war if the perceived benefits outweigh the costs of rebellion. Using probit and tobit models the propositions are tested empirically. Four variables, initial income, ethno-linguistic fractionalisation, the amount of natural resources, and initial population size are significant and strong determinants of the duration and the probability of civil wars. One important finding is that the relationship between civil wars and ethnic diversity is non-monotoniq highly fractionalised societies have no greater risk of experiencing a civil war than homogenous ones.