Polarization and Conflict

Debraj Ray's LSE public lecture titled "Polarization and Conflict". One of the best lectures I've ever attended, in the sense that it synthesizes theoretical and empirical works done by economists (including the lecturer himself, of course) to answer a practical question.

The lecture begins with the following question: Does ethnic division matters for civil war? Although there are quite a few reasons you can think of in favour, statistical analysis done by economists and political scientists has failed to identify the effect of ethnic division on the occurrence of civil war (see Fearon and Laitin (2003) "Ethnicity, Insurgency, and Civil War", American Political Science Review, 97, pp.75-90). This does not exclude the possibility that ethnic division indirectly affects civil war through, say, poverty. But the consensus emerging from the empirical literature is that ethnic fragmentation does not have a direct impact on civil war.

Professor Ray now asks this: Is the measure of ethnic division appropriate? The measure those empirical researchers have used is called "ethnolinguistic fractionalization index", which measures the probability that two people randomly picked up from a country belong to different ethnic groups (a good reference is probably Easterly and Levine 1997). Horowitz (1985), a book that takes a historical - rather than statistical - approach to this issue, suggests that broad cleavages, not fractionalization, matter. Civil war is more likely to happen when we see, in a given country, two large distinct groups rather than a large number of different groups. But how can we conceptualize this notion of "broad cleavages"?

Professor Ray, along with Joan Esteban, already made this happen in 1994 (see Esteban and Ray 1994). He created a measure of polarization. This measure captures two ideas: identity and alienation. Identity refers to how many people you are identified with - what can be called "local equality". Alienation refers to how far other people are from you - something called "global inequality". As civil war is not something you can do individually, you need to capture something social. That's what "fractionalization index" fails to do. But Esteban-Ray's polarization measure does.

Now a forthcoming paper in American Economic Review - "Ethnic polarization, potential conflict and civil wars" by Jose G. Montalvo and Marta Reynal-Querol - utilises this concept of polarization and does the same empirical analysis as Fearon and Laitin (2003) (see above). Guatemala, Sierra Leone, Nigeria, and Bosnia, countries well-known for civil war, turn out to be countries with high polarization index but low fractionalization index. Even after controlling for a bunch of other variables - income per capita, population, geography, etc. - it is shown that the index of polarization does have a direct impact on the probability of civil war.

The lecture closes with a remaining issue: Do economic differences across groups matter for conflict? Empirically, this question is hard to answer due to lack of data. But it's beginning to be collected. Theoretically, there are two types of civil war: vertical war (poor vs. rich ethnic groups as in Rwanda and Burundi) and horizontal war (economically similar groups fighting each other as in Nigeria). Research is now going on to figure out how economic differences across groups affect the occurrence of civil war.