Do Broad Based Governments Discourage Civil War?

This paper links the share of the population supporting the government to the risk of civil war. We study the question both theoretically and empirically. The theoretical model shows that having a smaller group of individuals control the government has an ambiguous effect on conflict risk: although a smaller group is less able to resist rebellion, its high per capita office rent, cost of regaining power once lost, and vulnerability to future victimization push it to lower rent extraction and therefore conflict risk. In the model simulations, when the share of the national population that is excluded from power increases, the risk of civil war may first increase, then decrease over a range, and then start rising again. The empirical analysis supports precisely that polynomial pattern instead of a linear effect. The risk of civil war onset is maximized when a small minority, such as 10% or less of the population, controls the government.