down the world’s leading geographer on mountainous terrain and
commissioned him to build a quantitative measure of the proportion
of a country’s terrain that could reasonably be judged mountainous.
This measure has since become widely used, and in our new work we
indeed find it to matter: mountains are dangerous.
132
WARS, GUNS, AND VOTES
Finally, let’s get to the politics. Surely that must be where the real
causes of violent conflict are to be found. We investigated a range of
political science variables, but focused on the one most widely used
by political scientists, Polity IV. I have already described the results:
in low-income societies democracy is dangerous, and in high-income
societies dictatorship is dangerous. Other than this we could find no
effects. Many people quite reasonably assume that violent internal
conflict is the consequence of political repression, but we simply do
not find evidence for this in the data. This does not, of course, mean
that repression is all right. Repression is unjust by definition because
it denies political rights. But this can be undesirable without making a
society more dangerous. And it is danger that is my subject.
It is time to try to make sense of this evidence. This involves an
interpretive leap from the statistics. Some interpretations become
implausible in the face of the statistical evidence, but more than one
interpretation is possible, and my own may be wrong. With that ca-
veat I propose the feasibility hypothesis. The feasibility hypothesis is that the key to understanding civil war is to focus on how rebellion
happens rather than on what motivates it.
Why focus on the rebels? Does that reveal a pro-government
bias? The focus on the rebels is simply because it is the act of rebel-
lion that defines the outbreak of civil war. All governments with the
exception of Costa Rica and Iceland have armies, so these cannot
be the defining feature. Sometimes a government army attacks its
own defenseless citizens, but, disgusting as this is, it is a pogrom,
not a civil war. The defining feature of the outbreak of civil war is
that the usual monopoly of force held by the government army is
challenged: a private organization within the society builds its own
army. No government can tolerate the existence of a private army on
its soil, and so even if it is the government that fires the first shot, it
is the creation of the rebel army that defines the war.
Because of the emphasis upon why the civil war is being fought,
it has become natural to focus on what motivates the rebel group to
Wars
133
form an army. My own past work fed this perspective: a paper en-
titled “Beyond Greed and Grievance” questioned the conventional
view that rebels were motivated by a sense of grievance, introducing
the idea that they might also be motivated by greed. But it was es-
sentially a refinement within the motive-based explanation for re-
bellion. I have now moved on from this view. It seems to me that the
key insight into rebellion comes not from asking why it happens but
how it happens. Usually rebellion, at least on a scale needed for civil war, is simply not feasible. The definition of civil war that I have
used, which is conventional, is that at least one thousand people are
killed in combat per year. On this definition the average civil war
lasts around seven years. So we are looking for rebel organizations
that can kill and be killed on a large scale and yet survive for years.
Rebellion on this scale faces two major hurdles. One is money: a
rebellion is going to be expensive. Someone has to pay for the guns,
and someone has to pay for the troops.
Often people think that a rebellion is just another form of po-
litical protest: people fight when they can’t vote. What brought it
home to me that rebellion is not simply a variant on other forms of
political opposition was a comparison I made between the finances
of a medium-size rebel group and a major political party. The rebel
group I chose was the Tamil Tigers. As rebel groups go it is not
out of the ordinary: northeastern Sri Lanka, where it operates, lacks
high-value natural resources; this war is not financed by diamonds.
I chose the Tamil Tigers only because, unusually, its finances have
been reasonably well studied. Its annual revenue is around $350 mil-
lion. This is around 28 percent of the GDP of North East Sri Lanka,
although most of the money is generated outside Sri Lanka from
donations by Tamils abroad.
For a political opposition party I decided to look for a rich one. I
chose the British Conservative Party, one of the longest-surviving and
most successful political parties in history, which, being on the politi-
cal right, is able to tap readily into financial support. I chose the elec-
134
WARS, GUNS, AND VOTES
tion year of 2005, when presumably its revenues were relatively high.
This information was more accessible than that about the Tamil Ti-
gers: its revenue was around $50 million. So one of the best-financed
political opposition parties in the world had an income one-seventh
that of a medium-size rebel movement. Recall that the revenue of the
Tigers was 28 percent of the GDP of the area they sought to control;
expressed in that way, the British Conservative Party was not one-
seventh the size of the Tigers, it was one ten-thousandth. There is no
simple passage from political opposition to private army: there is a
cliff face in the form of a financial barrier. Most would-be rebels just
cannot muster the money regardless of their motivation.
The other hurdle is military. Under most circumstances if a
small group of young men arm themselves and oppose the govern-
ment army, either they confine themselves to the irritant of terror-
ism aimed primarily against civilians or they die. Only if they are
faced by a militarily weak government do they stand much chance
of survival. While a rebel leader in Zaire, Laurent-Désiré Kabila,
was able to hang on for many years, safe because President Mobutu
had undermined all the organs of government, including the army.
So what is the feasibility hypothesis? It is that in explaining
whether a rebellion occurs, motivating factors are of little impor-
tance compared to the circumstances that determine whether it is
feasible. The tough version of the hypothesis, which I am reluctant
to adopt but which I suspect is close to the truth, is that where a
rebellion is feasible it will occur: the rebel niche will be occupied by
some social entrepreneur, although the motivation might be any-
thing across a wide range. Civil war is predominantly studied in
political science departments and so naturally enough they interpret
the motivation as political: sometimes it surely is, although even po-
litical motivations might stray quite some distance from social jus-
tice. Even rebellions that look entirely justified can sometimes be
called into doubt.
Take the rebellion, or rather rebellions,
in Darfur. Quite evi-
Wars
135
dently the government of Sudan is awful, and its conduct during
the conflict has been murderous. But at least part of the impetus for
the Darfur rebellion was the settlement of the rebellion in the South.
The Sudanese People’s Liberation Army, which fought the rebel-
lion in the South, won some remarkable concessions from the gov-
ernment in the North: it was allowed to run its own government, it
received a substantial share of the oil revenue, gilded by the promise
of huge aid inflows from donors, and all this was capped with the
promise of a referendum on full independence six years after the
onset of peace. No sooner was this deal signed than the contingent
from Darfur that had been fighting for the SPLA returned home
and launched its own rebellion. You can certainly see why, with that
precedent, rebellion might be attractive, at least for its leadership.
The top dog would become a president, and the others would be-
come ministers: secession has its rewards. The rebellion is, of course,
justified in terms of the atrocious sufferings of the Darfur people.
But to date the consequences of the rebellion for the people of Dar-
fur have been catastrophic: surely far worse than any plausible alter-
native scenario. Either the rebel leadership radically misjudged the
consequences of its actions, or it was not genuinely motivated by the
welfare of the people of Darfur. When the government was recently
coaxed to the negotiating table, the key rebel organizations refused
to attend. It is hard to see how a refusal to negotiate can be in the
best interests of the people of Darfur.
Sometimes the motivation for rebellion seems to be religious,
with the rebel group more akin to the fringe religious groups such
as those in Waco or Jonestown, but with the violence turned out-
ward. For many recruits the motivation may well be the lure of
violence: only a small minority of any society are psychopathic, but
these people are likely to be in the front of the queue for rebellion.
Sometimes it might even be sexual. Joseph Kony, the leader of the
Lord’s Resistance Army in Uganda, has reputedly accumulated sixty
wives: perhaps a young man’s dream come true?
136
WARS, GUNS, AND VOTES
The statistical results do not prove the feasibility hypothesis, but
they are consistent with it. I used the results to simulate the risks of
conflict in two hypothetical territories, in one of which rebellion was
easier than the other. I varied only five characteristics that seemed
to be most readily interpretable as differences in the feasibility of
rebellion. One territory was very mountainous, the other was flat:
mountains provide safe havens for rebels. One had a high propor-
tion of young men; the other had a low proportion: young men are
the recruits on which rebel organizations depend. Both territories
had a population of fifty million, but one consisted of a single coun-
try whereas the other was split into five identical countries, each of
ten million: the small countries would be unable to reap economies
of scale in security. One was dependent upon natural resource ex-
ports, the other was not: such exports can provide finance for re-
bellion. One was in Francophone Africa and so benefited from the
French security umbrella, the other was not. All the other charac-
teristics were the same and set at the average for all the countries in
the analysis. I then predicted the risks for these two countries. The
easy-rebellion territory faced a risk of 99 percent that conflict would
break out in one or other of its countries during a five-year period:
this territory was basically so dangerous that it was condemned to
perpetual conflict. The difficult-rebellion territory faced a risk of
less than 1 percent: basically it was safe, even over a century, it was
highly unlikely to fall into violence.
Dramatic as these differences are, they are not decisive evi-
dence. Most of the differences in characteristics that I have used to
construct easy-rebellion and difficult-rebellion countries could in-
stead be interpreted in terms of motivation. For example, I have
interpreted the increased risk of rebellion that mountains induce as
being because they are safe havens for rebels. But here is an alterna-
tive, motive-based explanation. The people living in the mountain-
ous areas of a country are usually poorer than those in other parts of
the country. They may storm down from the mountains to redress
Wars
137
this grievance: mountains matter, but because they create pockets of
grievance. While I do not want to discount such alternative inter-
pretations, I think it is striking that the most obviously grievance-
related characteristics such as the polity do not seem to make much
difference to risks, whereas these characteristics that at least have
plausible interpretations in terms of feasibility have such a large ef-
fect.
Wh at d o e s c i v i l wa r ac h i e v e ? Most obviously, war kills and injures people. Most of the dying is not as a result of battle, but due
to sickness. Mass flight takes people into unfamiliar places where
they lack natural immunity, and public health systems collapse. Be-
cause disease is highly persistent, much of the dying occurs after the
war is over.
Also pretty obviously, war is bad for the economy: not only does
it destroy the economy of the country itself, it damages the neigh-
boring economies. Again, these effects are highly persistent so that
many of the economic costs occur after the war has ended. I estimate
that for the typical civil war in a society of the bottom billion, these
economic costs alone are the equivalent to losing around two years
of income, or some $20 billion. However, I have come to realize that
these estimates, though large, grossly understate the true cost.
They make no allowance for the fact that the people affected
by violent internal conflict tend disproportionately to be among the
poorest and most disadvantaged people in the world. A dollar lost
by someone who is poor should be valued more highly than a dol-
lar lost by someone who is better off. The income differential be-
tween the typical citizen in the countries of the bottom billion and
a typical citizen of the other developing countries is already around
one to five. Even within the bottom billion there is a wide range
of incomes, with those countries that have recently been in conflict
grouped right at the bottom. Not only are the war-prone already the
138
WARS, GUNS, AND VOTES
poorest, they are likely to stay the poorest. Since slow growth is itself
a significant risk factor in violent conflict, the most violence-prone
countries are systematically among the slowest growing.
My estimates of cost also make no allowance for the fact that
peace is fundamental to development, so that its absence frustrates
all other potential interventions. The vaccination of children or the
>
reliable provision of anti-retroviral drugs is virtually impossible in
wartime conditions. This creates weakest-link problems in the pro-
vision of global public goods. For example, smallpox was eliminated
globally in a country-by-country campaign that was evidently a race
against time: until it was eliminated everywhere there was a risk
that it would break back out as a global disease. The last country
on earth where it was eliminated was Somalia during the 1970s. It
would now be impossible to eliminate smallpox: since 1993 Somalia
has been a no-go area. The maintenance of peace is thus a logically
prior investment that opens the possibility of all other interventions.
It is even possible to dress this up in the language and formulas of
technical economics. Financial economists now calculate option val-
ues. The true return on a liquid asset such as a bank deposit is greater
than the interest earned because it enables other investment oppor-
tunities to be seized as they arise. Peace also has an option value.
Finally, I have made no allowance for three global spillover ef-
fects: crime, disease, and terrorism. Large-scale political violence
and the resulting breakdown of the state create territories that have
a comparative advantage in international criminality. They provide
safe havens both for criminals themselves and for their material ac-
tivities, such as the storage of illegal commodities, notably drugs.
Some 95 percent of hard drug production is concentrated in civil
war or post-conflict environments. Civil wars also create the con-
ditions for the spread of disease: the breakdown in public health
systems and the mass movement of refugees. Some of this spread
of disease affects neighbors, and potentially it can also affect the
entire world. One of the explanations for the origin of AIDS for
Wars
139
which there is some evidence is that it originated during a civil war.
Finally, civil wars appear to assist terrorism. Al Qaida based its
training camps in Afghanistan because the absence of a recognized
government was convenient. Similarly, the American government
finally decided that leaving Somalia without a recognized govern-
ment was too dangerous, once evidence built up that Al Qaida was
relocating there.
Where this leaves us is that the cost of this form of political vio-
Paul Collier - Wars, Guns, and Votes Page 17
