Paul Collier is a Professor of Economics at Oxford University. From 1998 to 2003 he was Director of the Research Department at the World Bank. His book, The Bottom Billion: Why Poor Countries are Failing and What Can be Done About It, will be published in Spring 2007.
Present discourse on ethnic conflict is grounded in a mixture of common sense and advocacy, which together form a received wisdom with which everybody is familiar. The task of analytic and quantitative research is to bring us beyond this received wisdom: to discover whether there are counterintuitive behavioral relationships that might be helpful for developing policy.
The phrase “ethnic conflict” hovers between description and explanation. As a description of the large scale organized violence that besets many low-income countries, it is unexceptional. As an explanation, it is radically inadequate. Most societies are ethnically diverse. Where conditions are ripe for internal violence, the organization is indeed likely to be constituted along ethnic lines. During a conflict, ethnicity is often used as a propaganda tool. Indeed, whether or not ethnic divisions are a cause of conflict, they are quite likely to be a consequence of it. After such conflicts, societies are often trapped in the ethnic organizations and categories that were determined by the conflict.
As a literal objective, “ending ethnic conflict” cannot be taken seriously. Political conflict is endemic, and intra-ethnic affinity may form a cost-effective basis for political organization. Political contests between ethnic groups are consequently common in ethnically diverse democracies. What can, and indeed should, be ended is large scale, organized violence. This kind of violence generally takes three variants: communal violence, pogroms, and rebellion. Of these three, rebellion is by far the most deadly, since it results in full-scale civil war. Communal violence, in contrast, is generally the least deadly because it lacks large scale organization.
The prevention of pogroms, in which a government systematically kills the members of some ethnic group, is, in one sense, straightforward: pogroms can be prevented by enforced international norms. The difficulty in preventing pogroms thus comes down to a dispute over the limits on state sovereignty. The prevention of civil war, by contrast, requires a substantially different agenda. Civil war, is by definition, defined by private, non-governmental organized violence. This is not, of course, to lay the blame for civil war at the door of the rebels. Rather, all governments attempt to maintain a monopoly over organized violence within their borders. Where they lose this monopoly due to the emergence of a rebel group, a civil war is virtually inevitable. But reducing the global incidence of civil wars is indeed feasible and would be hugely valuable to international peace and stability. Pogroms and communal violence are different and equally important variants of ethnic conflict; however, my discussion here will not apply to them.
To “end” ethnic conflict there are three possible points of intervention. We can attempt to prevent such conflicts from starting, we can attempt to stop on going conflicts, and we can attempt to strengthen the peace in situations where conflict has recently ended. Herein I discuss each scenario, although I contend that the biggest payoffs would come from improving the record of maintaining post-conflict peace, due to the high level of post-conflict relapses into civil war.
The Deep Prevention of Civil War
Deep prevention is obviously the most desirable form of prevention. The normal, somewhat pious, discourse on the deep prevention of ethnic conflict invariably invokes the need to address its root causes. Obviously, when grievances are genuine they should be addressed, regardless of whether or not they are the root causes of violence. However, the argument that violence will inevitably occur unless such grievances are addressed seems to be both unnecessary as a basis for remedial action and potentially counterproductive, as it becomes subject to hijacking by advocacy groups with their own pet agendas. The root cause method of explanation is inextricably linked to the notion that the cause of rebellion is some objective grievance. But this is incorrect.
Over the past few years, Anke Hoeffler and I have statistically investigated the causes of civil wars. In our latest work, Beyond Greed and Grievance: Feasibility and Civil War, we have been joined by Dominic Rohner and have analyzed a comprehensive set of civil wars from 1965 to 2004. Ethnic diversity does marginally increase the risk of a civil war, but we discovered that other factors are more important in determining whether a country is more or less prone to civil war.
Imagine a typical developing country. In any five-year time period, there is a five percent risk that a civil war will break out in this country. Now, let us vary the characteristics of this situation in five respects, none of which are commonly thought to be root causes of ethnic conflict. Country A has 50 million people. Region B also has 50 million people, but it is divided into five countries. Country A is flat, whereas the five countries of B are as mountainous as Nepal. Country A has only half of the normal proportion of its society made up of young men, whereas the countries in region B have double this same proportion. Country A does not export primary commodities, whereas the B countries rely on these commodities for a quarter of their GDP. Finally, country A is in Francophone Africa, and region B is not. Other than these differences, country A and region B are identical in every respect. I have chosen these five characteristics because, while they are all important for determining a country’s propensity towards internal conflict, it is hard to set any of them very clearly in the context of objective grievances. None of the five are part of the received wisdom of root causes, and therefore these differences between the two places should not have major consequences for the risk of civil war. Yet, based on statistical evidence, the predicted differences in the likelihood of civil war are substantial. Country A is objectively very safe, with a predicted risk of war of only 0.3 percent over a five-year period. In contrast, the chance of civil war breaking out in at least one of the countries in region B is 97 percent.
Of course, pushing a regression to these extremes takes us beyond the bounds of accurate prediction. But the results nevertheless carry a stark implication. While it is difficult to discern differences in objective grievances between these two places, the empirical evidence demonstrates that there are major differences in the likelihood of rebellion. It is much harder to sustain a rebellion in country A than in region B. Let us go through the five characteristics again and see why.
