In recent years there has been a debate in the statistical civil war literature over the effects of ethnic heterogeneity on the risk of ethnic conflict. The conventional argument has been that the potential for conflict is higher in countries with many ethnic groups than in countries with few ethnic groups. This has been based on the core assumption that more groups generally raises coordination costs, which in turn is conducive to weak institutions, corruption and eventually higher risk of conflict. Recently, this notion has been challenged by those who contend that the risk of conflict is highest when there are only two large groups in a country. Under such conditions the system is likely to become highly polarized, thus increasing the probability of conflict. According to this argument, more groups would in fact reduce interethnic tension. In parallel, a third line of thought has been advanced that departs from the question of numbers and instead focuses on the relative size of the ethnic groups. This argument has suggested that the risk of conflict should be highest in systems where the largest group has a dominant position, yet without complete control. In such situations the majority will be tempted to abuse its power over the minorities, leading eventually to a revolt by one or more of the excluded groups.All in all, there has been limited statistical support for either of these arguments, prompting the inevitable question of whether ethnic heterogeneity has any particular effect on the risk of conflict at all. As such, the ambiguity of the results has played into the well-known debate over greed versus grievances, and has effectively weakened the case for those who uphold that ethnicity is a relevant determinant of civil war onsets. In this thesis I argue that the reason why the results have been ambiguous is not that ethnic heterogeneity is irrelevant. Rather, it is because the tests conducted have either lacked proper theoretical foundation; been based on over-aggregated data, confounded the different dimensions of ethnic heterogeneity; or generally struggled with the methodological challenges inherent in research involving the concept of ethnicity. The ambition of this thesis is to remedy some of these shortcomings and provide a better understanding of how the landscape of ethnic heterogeneity affects the risk of conflict in multiethnic states. More specifically, it is to examine whether some multiethnic configurations are more prone to conflict than others.
To answer this question, I propose two new measures by which the landscape of ethnic heterogeneity can be mapped. The first is concerned with the number of groups while the second focuses on with the potential effect of dominance by a large majority. In testing these measures on a new dataset of ethnic groups compiled by Wimmer et.al (2009) I find considerable support for the conventional argument that countries with a fragmented ethnic landscape – i.e. many groups – have a higher risk of ethnic conflict than countries with few ethnic groups. These results are consistent across different datasets of civil wars and ethnic groups. I find less support for the notion that dominance increases the risk of ethnic conflict in general, but when disaggregating the dependent variable into conflicts over territory and conflicts over government I find that dominance is in fact a significant determinant for territorial conflicts, but not for conflicts over government. The opposite seems to be the case for fragmentation; it is significant for government conflicts, but not for territorial ones. I conclude by making the case that the two core dimensions introduced to map the landscape of ethnic heterogeneity – fragmentation and balance – could serve as the basis for a classification of multiethnic states, which in turn can facilitate more qualitatively oriented studies of the effect of ethnic heterogeneity on the risk of civil war.