example, Jonas Savimbi launched two civil wars in Angola. Since
civil war destroys the economy, by the time of the second civil war,
Angola was poor: low income preceded that second civil war even
if it did not cause it. So something offstage, namely Savimbi, kept
causing wars, and the first war reduced income. The final one is that
some phenomena are likely both to lower income and to increase the
risk of civil war. Bad governance might destroy the economy and
give people cause to rebel. So just because low income occurs before
war is not enough to conclude that it causes war.
Gradually, economists have become better at guarding against
these problems, introducing steps that leave fewer and fewer am-
biguities. In our new work we have used more of these safeguards:
indeed, having more observations makes it easier to do so because
the safeguards generally need large samples. To give you a taste of
the safeguards, we got rid of the Savimbi problem by restricting the
analysis to the prediction of first-time civil wars. In part, we reduced
the bad governance problem by controlling for it, and by including
Wars
125
as many characteristics as possible in the analysis. We addressed the
problem of anticipated conflict by replacing the actual level of in-
come with the level predicted by a few characteristics that influence
income but do not otherwise affect proneness to civil war: in prin-
ciple this approach should also fix the previous problem. Even with
these safeguards there is room for doubt, but at least we now have
results based on comprehensive data—at its maximum more than
sixteen hundred episodes during which eighty-four civil wars broke
out. This may not be as good as it can get, there is always room for
improvement, but the results are worth serious attention.
Although we are economists we have tried to be agnostic as to
what might explain proneness to civil war, and so we have included
a wide range of possible causes drawn from across the social sciences.
In addition to various characteristics of the economy, these include
aspects of the country’s history, its geography, its social composition,
and its polity. Let me be clear about what we do not include: we are
not interested in the personalities and immediate political circum-
stances leading to the conflict. All wars have multiple causes: one
reason that Saddam Hussein invaded Kuwait was that the Kuwaiti
leadership was sufficiently impolite to doubt whether he had been
born in wedlock. Such things matter for a proper understanding
of any particular war but clutter up and detract from our under-
standing of civil war as a phenomenon. In trying to prevent war I
suppose that it is useful to know that insulting psychopaths is not a
good idea, but my approach has been to try to find structural char-
acteristics that expose a country to risks and could, over time, be
changed. So let’s get started: what actually caused these eighty-four
civil wars?
The economy matters. Low-income countries are significantly
more at risk even when we control for as many of the possible spu-
rious interpretations as possible: poor is dangerous. Nor is it just
the level of income: it is also the rate of growth. Given the level of
income, societies that are growing faster per capita are significantly
126
WARS, GUNS, AND VOTES
less at risk of violent conflict than societies that are stagnant or in
decline. In one sense this is hopeful: it tells us that economic devel-
opment is peace-promoting. I have no patience with the romantics
who believe that we can build peaceful societies by arresting eco-
nomic growth: the vision of the restoration of Eden. I think that the
truth is quite the opposite.
The statistics of the world post-1960 are supported by the deep
historical evidence of the societies of early history. These impover-
ished societies were extremely violent, as Azar Gat has now bril-
liantly shown in War in Human Civilization. Economic development
is a key remedy to violence. The truly difficult issue about the peace-
enhancing effects of economic development is to sort out which of a
number of possible routes might account for it. I suspect that there
is no single magic route that could be isolated and promoted distinct
from overall economic development. My guess is that there are mul-
tiple routes, such as jobs, education, hope, a sense of having some-
thing to lose, and more effective state security services, all of which
contribute something.
The level and growth of income are not the only aspects of the
economy that matter for violence. Dependence on natural resources
also increases risks. This proposition is supported by the grim evi-
dence of resource wars: timber in Liberia, diamonds in Sierra Leone,
a wonderland of minerals in the Congo. It is also now supported by
statistical analysis of where violence occurs within countries. For ex-
ample, in Angola the violence tended to be concentrated in the dia-
mond areas. It is also evident why natural resources might increase
proneness to violence. They provide a ready source of finance for
rebel groups, they provide a honey pot to fight over, and they enable
the government to function without taxing the incomes of citizens,
which gradually detaches it from what citizens want.
Nevertheless, this is probably the most controversial of our re-
sults: some scholars have argued that it is purely an oil effect, and
others that we have run afoul of a particular variant of reverse cau-
Wars
127
sality. Anke and I have learned not to be proud: over the years I
have been wrong sufficiently often to have had the presumption of
infallibility knocked out of me. With our new data we duly tested
whether oil was the real story: as far as we can see it isn’t. We do,
however, find that with sufficient natural resources a country be-
comes safe. Saudi Arabia and the other superrich Gulf States are
peaceful: they can afford good security systems and they can afford
to buy off all potential opponents. Indeed, just this ambiguity—
some resources increasing risks but sufficient abundance reducing
them—is predicted by the sophisticated recent theoretical work of
Francesco Caselli of the London School of Economics.
The reverse causality problem is trickier. It arises because we
measure resource dependence by the ratio of primary commodities
to income. That inevitably creates a problem because countries that
for whatever reason have low income will tend to have a high share
of primary commodity exports, simply because income, the denomi-
nator, is small. Some scholars have recently tried to get around this
problem by replacing our measure with a newly available measure:
the value of natural resource reserves. The World Bank released
estimates, country by country, giving a snapshot for the year 2000.
Unfortunately this runs into another form of the reverse causality
problem. Any estima
te of natural resource reserves depends upon
what resource extraction companies have found through prospect-
ing. Prospecting is costly, and so the value of proven reserves is an
economic concept as much as a geological one. It is only worth doing
in places where the company’s rights of extraction are secure. Be-
tween 1960 and 2000, prospecting thus tended to avoid societies that
were at civil war, and also those places where there was a serious
risk of war. Think what this implies. The places with few proven
natural resource reserves in 2000 will tend to be those with the worst
prior history of civil war. The scholars who followed this approach
duly announced with a confident fanfare that possessing large en-
dowments of natural resources actually makes a society safer. The
128
WARS, GUNS, AND VOTES
problem of reverse causality has recently been overcome by Tim
Besley and Thorsten Persson. They investigate whether increases
in commodity prices affect the risk of civil war in commodity ex-
porters. Consistent with our own results, they find that the risks are
increased. But they find an important qualification: the risks are not
increased if the quality of democracy is sufficiently high. As with
elections and reform, democracy is a force for good as long as it is
more than a façade.
So much for the economy; let’s turn to history. The aspect of a
country’s history that most commonly excites interest when it comes
to explaining a civil war is its colonial experience. Understandably,
many people in developed countries find it convenient to emphasize
the guilt of their own societies, and equally, many people in develop-
ing countries want to avoid any impression that the violence of their
societies is a consequence of characteristics within those societies. So
there is a ready demand for evidence that colonialism is responsible
for the subsequent violence. Unfortunately, Anke and I cannot find
evidence that supports this contention. Neither the length of time
that has elapsed since independence, nor the particular former co-
lonial power, seems to matter. I do not want to push this too far: it
is quite evident that Portuguese decolonization was disastrous. An-
gola, Mozambique, East Timor all went straight into civil war. But
the Portuguese empire was relatively small, and neither the Brit-
ish nor the French empire, which were the two major ones, shows
any distinctive patterns. The empire-free countries of Ethiopia and
Liberia both eventually collapsed into terrible civil wars. I want to
stress that this is not to exonerate colonialism: I am not an apologist.
But blaming colonialism for civil war is a costly illusion because it
detracts from the focus on its real causes, which are often things
that can still be changed. It may make many people feel better, but
it inhibits action.
The other aspect of history that many scholars have got excited
about is the Cold War. Quite evidently, in some instances civil wars
Wars
129
were aided and abetted by each side. As Niall Ferguson has pithily
expressed it, what people expected would be the Third World War
turned out to be third-world wars. But even the effect of the Cold
War is controversial. While it is clear that the superpowers inter-
vened in civil wars, it is less clear that they caused them. Indeed,
they may even have had an offsetting effect: if any petty war had
the potential to scale up into the Third World War, maybe the su-
perpowers tried to prevent conflicts occurring. We tested for this by
investigating whether the post–Cold War period has had a signifi-
cantly different incidence of the outbreak of civil war than we might
otherwise expect. Basically, it doesn’t. There was a brief, significant
surge in new outbreaks of violence in the first few years after the
end of the Cold War, but from 1995 onward the world has been
back to normal. The third world’s wars were not, in general, caused
by fears of the Third World War.
The one aspect of history that really seems to matter is a previ-
ous history of civil war. Once a country has had a civil war, it is much
more likely to have another war. However, this is ripe territory for
the problem of common causality. Suppose that there is some char-
acteristic of the country that makes it prone to violence but that we
have missed: perhaps the people are just inherently violent. Statisti-
cally this will appear as one war causing another, whereas actually
the same underlying factor is causing all of them. We got around
this problem by measuring the number of years since the last civil
war and testing whether that, or the mere fact of having had a previ-
ous civil war, was decisive. It turned out that it was only the length
of time since the previous war that mattered: the risk of further con-
flict gradually declined with the passage of time. This looked more
like a risk of violence caused by the gradually decaying effects of
previous violence than by something underlying and constant.
So much for history: how about the structure of the society?
Perhaps the most important aspect of social structure that we inves-
tigated was the effects of ethnic and religious divisions. This is the
130
WARS, GUNS, AND VOTES
point at which I get mud on my face because our previous results got
overturned by the new data. We had previously found that ethnic
and religious diversity had two opposing effects. We now find that
the relationship is more straightforward: diversity increases the risk
of violence. As far as we can tell, ethnic and religious diversity com-
pound each other.
Another aspect of the social structure that seems to affect the
risk of violence is the proportion of young men in the population:
young men, defined as those aged between fifteen and twenty-nine,
are dangerous. I suppose this is not surprising: violent rebellions are
seldom staffed by elderly ladies. The effect is large: young men ap-
pear to be very dangerous. A doubling in the proportion increases
the risk of conflict from around 5 percent over a five-year period
to around 20 percent. However, there are a couple of caveats here.
It is very hard to distinguish statistically between societies with
many young men and those with many young women: other than
in China the two tend to go together. In most rebellions the fighting
is done almost exclusively by young men, but not always: famously
the Eritrean People’s Liberation Front was one-third female. And
it is also hard to distinguish between societies with a high propor-
tion of young men of fighting age and those that simply have rapid
population growth.
A final aspect of the society that matters is its size. The risk of
conflict increases with population, but the relationship is much less
than proportionate. A country with a population double that of an
otherwise identical country has a risk that a civil war will break out
that is only a little higher than its sma
ller counterpart: specifically it
is one-fifth higher. Think about that for a moment: it implies that
if two identical countries were merged, then, abstracting from all
the nationalism that would of course be provoked, the risk of a civil
war breaking out somewhere in the combined territory would fall.
Let’s say that previously there was a 10 percent risk in each country,
so that the risk that there would be a war in one or the other of the
Wars
131
countries was around 20 percent. Now, the risk of a conflict in the
new super-country is only a fifth higher than that in either of the
former countries individually: that is, the risk is 12 percent. So the
risk of war has fallen from 20 percent to 12 percent.
I think that this is because there are scale economies in security,
and I think that it matters. Most of the countries that emerged with
the dissolution of empires were too small to reap adequate economies
of scale in security. This creates a potential tension between the scale
economies that could be reaped by mergers between countries and
the greater ethnic diversity that would probably be a consequence.
At present all the political pressure is for nations to get smaller. Eri-
trea exited from Ethiopia. East Timor split from Indonesia. Yugo-
slavia split up into six different countries. Southern Sudan will hold
a referendum to decide whether to withdraw from Sudan. Stepping
back from the historical particularities of these struggles for nation-
hood, is the drift in the right direction?
So much for social structure: how about geography? We had al-
ready tried to investigate whether particular types of geography were
well suited for rebellion. The most promising idea seemed to be that
of the safe haven, and two aspects of the landscape seemed likely to
facilitate: forests and mountains. Forests were relatively easy to mea-
sure: the Food and Agricultural Organization had a measure, coun-
try by country. We investigated it and could find no effect. But there
was no equivalent measure for mountains. There were crude prox-
ies such as the highest point in the country, but these seemed to miss
the point as to what rebel groups would actually find useful: they did
not want to perch on the top of Everest; they wanted rugged terrain
where government forces would not be able to find them. We tracked
Paul Collier - Wars, Guns, and Votes Page 16