Paul Collier

As an economist who works on civil wars, I am in a small minority in two senses: very few economists work on civil wars and most of the people who do work on civil wars are not economists. I do not want to imply that economics has more to offer than other disciplines. But to date, the contribution of economics has scarcely even been heard, and it deserves more attention.

I lead a research project at the World Bank which has examined some
160 countries and 78 civil wars between 1960 and 1999. The project seeks to develop a statistical model which will explain the incidence of civil war within a country. The model examines the impact of several explanatory variables and predicts the risk of civil war over a 5-year period.

The Economic Causes of Civil Wars
Our statistical analysis indicates that the level, growth, and income
structure of a country are significant and quite powerful explanations
for the likelihood of civil wars.

First, conflict is overwhelmingly a phenomenon of low income countries.
Obviously, conflict reduces income, but our research controls for this
by examining income before the conflict at the beginning of a 5-year
period and predicts the risk of civil war during the subsequent period.

Second, and more controversially, our research indicates that the
faster the rate of growth in a country, the lower the risk of conflict.
This result runs counter to the common presumption that rapid economic change in a country causes conflict. In fact, rapid economic growth reduces conflict. To provide some order of magnitude, the average developing country faces about an 11% risk of having civil war in any 5 year period. Each time a percentage point is added to the rate of growth, this reduces the risk of civil war by a percentage point, which is a significant impact.

Third, dependence on primary commodities substantially increases the
risk of conflict, unless the primary commodity is extremely plentiful,
as in the case of oil in Saudi Arabia. The difference in the risks is
absolutely enormous. In a country with no primary commodity exports
at all, the risk is about 1% in a 5-year period. In a country with high
dependence on primary commodities, which means about 30% of its national income comes from primary commodities, the risk is around 23%.

The particular primary commodities upon which a country is dependent
does not matter as much as one might think. The big difference is between oil and non-oil, but the impact is not that marked. At low levels of dependence on oil, the effect is not significantly different from dependence on other primary commodities. However, if a country has 40% or 50% in oil income, the likelihood of civil war is quite high.

Besides these economic factors, two aspects of a country’s social
composition are also closely correlated with the likelihood of civil
war. One is “ethnic dominance.” By dominance, I mean that the largest
ethnic group in the country is more or less a majority but not overwhelming.

Our research suggests that when the largest ethnic group is between
45 and 90% of the population, this constitutes conditions of ethnic
dominance. If the country is characterized by ethnic dominance, the
risk of conflict approximately doubles. This might first appear to be
a large effect, but it is small when compared with the economic effect.

We have also investigated both ethnic diversity and religious diversity
and the combination which we call social fractionalization. Controlling
for ethnic dominance, the more the society is fractionalized into different ethnic and religious groups, the safer it is.

Factors Not Correlated with Civil Wars
Some very surprising things are not correlated with the likelihood of
civil war. One is military expenditure. We cannot find any deterrence
effect to military expenditure before a conflict. Of course, it is quite
complicated to examine the effect of military expenditures, because
governments may see a big risk of conflict and increase military spending in anticipation of the conflict. In such a case, the increase in spending might appear to be causing the conflict, when it in fact may be a result of an impending conflict. Our research has controlled for that and we still cannot find any deterrent effect for military expenditures.

I also cannot find any effect from economic inequality on the risk
of conflict. I have looked at both income inequality and land inequality,
and neither are correlated with an increased risk of civil conflict.
In particular countries, there might be a correlation, but globally,
we find no relationship.

The other factor which does not appear to be correlated with conflict
risk is political rights. Democracy, dictatorship, and political repression
seem to have no effect on the risk of conflict. This often disappoints
those who promote democratization as a means for avoiding violent conflict, but the statistical evidence does not support a strong correlation.

The Case of Africa
The case of Africa over the last 30 years illustrates the statistical
model I have outlined. Africa has a lower level of income than other
regions, drastically lower growth than other regions and a chronically
weaker structure of income. Africa alone of the continents has not diversified away from primary commodity dependence. It is more dependent on primary commodities now than it was 30 years ago.

Africa has had a very unfortunate deteriorating trend. Thirty years
ago, it was safer than other regions. Now, it is more dangerous than
other regions in terms of the risk of violent conflict, and that is
fully accounted for by the economic deterioration in Africa.

On the other hand, there are relatively few societies in Africa that
are characterized by ethnic dominance. While there are exceptions, Africa on the whole is just too fractionalized for even the biggest groups to be over 45%. This in turn reduces the risk of conflict.

Thus, on our analysis, Africa’s high incidence of civil war is not
due to its social structure. Its social structure is a factor making
it relatively safe. Its problems are economic, and hence contingent.
This conclusion is deeply against the grain of most thinking.

Policy Implications
Several policy implications follow from this understanding of civil
conflict. First, if we actually want to try to prevent conflict, we
might get more mileage out of reducing the economic viability of violent
movements than addressing objective grievances.

To avoid any misunderstanding, I think that there are very good reasons in all societies for addressing issues of inequality and political rights based on our research. Unfortunately, based on our research I think that it is a false bill of goods to believe that focusing on these issues is going to deliver peace in conflicted societies. Our agenda for conflict reduction has to be substantially focused on reducing the economic viability of violence.

Furthermore, it is clear that some of the variables that matter most
in reducing the likelihood of conflict are economic variables — particularly the level of income, the growth of income, and the structure of income. Regardless of one’s interpretation of the data, that suggests that economic interventions can build a more peaceful world.

What are those economic interventions going to be? One is to increase
growth through a mixture of policies and aid. I have investigated whether aid has any direct effect on conflict risk, and I cannot find any. There are indirect effects through growth, but not direct effects.

Economic development is an effective strategy for reducing conflict.
Based on a simulation, policy improvements and larger aid budgets could bring down the risk of conflict in a typical aid recipient country by about one third over a 5-year period, which would be a significant accomplishment.

Another policy implication is that primary commodity dependence is
quite dangerous. Ten years ago, the World Bank did not know this, but
now we recognize the enormous importance of diversifying the economy. The developing world has diversified massively over the last 20 years, but Africa has not followed the rest of the developing world in this regard. To my mind, the primary economic task now for Africa is to achieve that diversification.

Finally, this analysis has at least one implication for the war against
terrorism. International terrorists are using failed states — states
where there are civil wars — for safe havens. Failed states are the
one type of territory which is absolutely out of reach. Consequently,
I believe that one part of the long term strategy to fight terrorism
must be to reduce the number of safe havens.

Of course, that is easier said than done. It is not achieved by bombing
them. By bombing, you can destroy a government, but failed states have already achieved that. That is not a remark on what American policy should be, but rather a statement that over the next 10 years, to solve the problem of failed states, we must make development interventions to prevent states from falling into failure, and — what is even harder — to rebuild states which are in conflict and coming out of it.

Compared with the sums of money that will be spent on military and
intelligence activities, the amount of money that is spent on a development agenda is absolutely tiny. And yet, our research shows that money spent on a development agenda will substantially reduce conflict risk. I can only hope that this fact will be taken into consideration.

Paul Collier is Director of the Development Economics Research Group at the World Bank and a senior World Bank spokesperson on development economics research. He is currently on leave from Oxford University where he is one of six full time professors of economics and director of the Center for the Study of African Economies. Recently, Collier published Economic Causes of Civil Conflict and Their Implications for Policy, a study of 47 civil wars from 1960 to 1999. This article is excerpted from a keynote address he gave at the Kroc Institute on October 3, 2001, for a meeting of the Institute’s project on Globalization and Local Violence.

Top of Page

Home > Publications > Peace Colloquy > Issue 1 (Spring 2002)