Ethnic distribution, effective power and ethnic conflict

Abstract

This thesis is investigating both in theoretical and empirical terms the relationship between the features of ethnic distribution and the probability of con ict. The thesis is composed of four chapters. The rst chapter is the introduction to the thesis in which I brie y summarize the main problems related to the existing literature on ethnic diversity and con ict and I present an overview of the measures of ethnic diversity commonly used in the literature. In the second chapter I propose a theoretical model that speci es the potential of con ict in a society as a function of the population distribution across ethnic groups. I axiomatically derive a general parametric class of indices of con ict potential that combines the groups' e ective power and the between-groups interaction. The e ective power of a group is a function of a group's relative size but it also depends on the relative sizes of all the other groups in the population. The interaction component, on the other hand, is given by the probability of interaction between the members of one group with those of other groups. I show that for certain parameter values the index reduces to the existing indices of ethnic diversity, while in general the indices combine in a non-linear way three di erent aspects of ethnic diversity, namely the fractionalization, the polarization and the ethnic dominance. The results of the model share some common features with the literature on con icts in contests and the literature on voting power indices. In particular, the power component of the extreme element of the class of indices is intuitively related to the de nition of voting power in a simple majority game. In that particular case, the value of the e ective power is given by the relative Penrose-Banzhaf index of voting power calculated over the shares of populations associated to each ethnic group. In the third chapter I investigate empirically the role of ethnic diversity in the explanation of the ethnic con ict outbreak. The empirical performance of the indices of con ict potential developed in the second chapter is tested against the existing distributional indices of ethnic diversity within the context of the commonly used logistic model that focuses on the onset of ethnic con icts in a time range from 1946 to 2005. Together with the set of the explanatory variables for structural conditions and country characteristics, I take advantage of the recent "Ethnic Power Relations" data set which includes additional information on the political exclusion and competition along ethnic lines and it o ers the best coding for ethnic wars. The results obtained suggest that the indices of con ict potential outperform the existing indices of ethnic diversity in the explanation of ethnic con ict onset. This evidence is robust to the inclusion of a larger set of regressors, time and regional controls as well as to several other estimation techniques. The fourth chapter explores empirically the determinants of con ict duration with a particular attention to the potential role of ethnic diversity together with ethnic politics and competition dynamics. The rst part of the chapter presents an overview of the existing literature on con ict duration, the main data cources and the related econometric issues. The second part of the chapter consists in a non-parametric and a parametric survival analysis of the duration of ethnic con ict where we address in detail the issues of non-proportionality of the hazard function, the unmeasured heterogeneity and the presence of "repeated events". The results suggest that there is a statistically signi cant and robust association between ethnic distribution and con ict duration, together with other commonly used explanatory variables in the literature on con ict duration.