teered to build a data set on the price of Kalashnikovs on the infor-
mal market around the world, country by country and year by year.
This took real ingenuity, but he had it. After seven months he had
three hundred observations, which was enough for statistical analy-
sis. We were in business. Even a cursory look at the numbers showed
interesting patterns: a secondhand Kalashnikov was only about half
as expensive in Africa as in the rest of the world. So our research
shaped itself around this finding: why were Kalashnikovs so much
cheaper in Africa than elsewhere, and did it matter? Phil faced an
examination on his work: generally, economics departments are not
particularly impressed by the importance of a question; they are con-
cerned about whether the student has applied the latest techniques
to the highest possible standards. Somehow, despite being the first
Oxford student in half a century to go from zero rowing experi-
ence to being selected for the university’s team, and spending seven
months on data gathering, Phil also managed to gain a distinction
grade on his thesis: you can be reasonably confident that his results
are defensible.
So why are secondhand Kalashnikovs cheap in Africa? The an-
swer is that the key source is leakage from government armies. Gov-
ernment soldiers are usually very badly paid, and so they are tempted
to sell their guns or steal from stockpiles. Government armies buy
Kalashnikovs most vigorously when they are fighting a rebellion. So
the guns are officially imported into Africa, stolen, and so become
illegal, but cannot easily be re-exported to the markets in which sec-
ondhand Kalashnikovs fetch a high price. That is because to export
the guns out of Africa they have to be imported into countries that
generally have sufficiently good border controls to make it difficult.
But the guns do not just stay in the country whose government first
imported them to Africa. Africa’s internal borders are highly po-
rous, and so the cheap guns slosh around the continent going to
116
WARS, GUNS, AND VOTES
where there is currently demand: which means wherever there is a
war. The next question was “Does it matter?” By this Phil meant
whether cheap guns increased the risk of rebellion. According to
simple economic theory it ought to do so: cheap guns should make
rebellion easier and so more likely, or, stated slightly differently,
small rebel groups would tend to buy more guns as a result of their
being cheap, and so be more likely to scale up their violence to the
point at which it met the definition of a civil war. I have to admit
that I thought that this was a pretty demanding question to ask of
the data. In the event, Phil found a significant effect: cheap guns
increased the risk of civil war. When Phil went back to Australia he
generously deposited all his data with the Peace Research Institute,
Oslo, where it sits waiting for some other enterprising student to
update and expand it.
One implication of Phil’s work is that dangerous countries
make for dangerous regions. Another is that if Africa’s internal bor-
ders are porous, the way to curtail armaments reaching the current
danger points is to curtail arms flows to the entire region. There
are two ways of going about such curtailment, neither easy. One is
to address the inadvertent leakage of aid into the purchase of ar-
maments: squeezing the finance for armaments should reduce their
inflow. The other is to attempt to impose quantitative restrictions
on the trade flows.
Each of these approaches might seem impertinent: other regions
buy plenty of armaments, why shouldn’t Africa? Before Africans
and sympathizers put on the comfortable clothes of indignation,
step back for a moment and think of Africa’s own interest. Quite
manifestly, Africa no longer faces a military threat external to the
region: all its threats are internal, either threats between neighbors
or fears of rebellion. Threats from neighbors place governments in a
prisoner’s dilemma. Although each country’s increase in its military
spending makes it feel more secure, it does so by making its neigh-
bors feel less secure. As I discussed, the neighbors respond to this
Guns
117
by increasing their own military spending. Arms races in Lilliput
are a menace for the region as a whole. Further, recall that military
spending is higher when countries are small. Africa is divided into
fifty-four countries, despite having a total population far less than
that of India. And so the waste that comes from military spending is
more of a problem in Africa than in regions with fewer countries.
The solution to a prisoner’s dilemma is cooperation. Africa
needs collective action to curtail its arms spending. But the problem
with cooperation is enforcement: each government has an individ-
ual interest in encouraging its neighbors to coordinate reductions
in their military spending while not doing so itself. There are vari-
ous ways of securing cooperation, but the most straightforward is
to persuade a neutral policeman to enforce it. The policeman could
be the donors limiting leakage from aid, or it could be the United
Nations imposing effective embargoes on arms exports. But in one
shape or form Africa should be searching for a policeman. As I ex-
plained, the donors have recently reshaped aid allocations so that
it is even more difficult to prevent leakages. So how about direct
restrictions on arms exports?
I f a i d i s i na dv e rt e n t ly f i na n c i n g arms purchases, and if cheap guns make civil war more likely, one possible remedy might
be to limit the flow of armaments into dangerous places. Fortu-
nately, the countries most at risk of civil war are not sufficiently in-
dustrialized to have an arms industry, so that curtailing the trade
might indeed limit the availability of guns. In recent years sufficient
popular pressure has built up behind this idea that from time to time
it is implemented. For example, during the standoff in Cote d’Ivoire
between the government in the south and a rebel group in the north,
the United Nations placed an embargo on arms shipments to either
side. Are such arms embargoes effective? The work I am going to
describe is by Stefano DellaVigna and Eliana La Ferrara.
118
WARS, GUNS, AND VOTES
They reasoned that if anybody knew what was really going on,
it was the people who were investing their money in companies that
manufactured armaments. Of course, not everyone who buys shares
in a company knows much about how it is performing. But it only
needs a few people to take the time and energy to find out in order
for the share price to be affected: if they discover that the company
is being hurt by an embargo, they will sell their shares and the price
of the stock will fall. Using this reasoning DellaVigna and La Fer-
rara got information on which armaments companies were export-
ing to the country prior
to the embargo. They then checked what
the announcement of the embargo did to the subsequent price of
the stocks.
What they found must at first have seemed confusing: although
the stock prices duly fell for some companies, for others prices ac-
tually went up. Was this just noise: random movements in stock
prices? It turned out that it was not random. The stock prices of
armaments companies based in the OECD—that is, the developed
countries—fell significantly as a result of an embargo. But the stock
prices of armaments companies based outside the OECD rose. They
realized that the most likely explanation for this was that these latter
companies were breaking the embargo and, in the process, profiting
from the absence of competition from the OECD companies.
They concluded from that analysis not that arms embargoes
cannot work, but that there is a simple method of policing them
more effectively. Suspicions should be raised if an embargoed com-
pany’s stock price rises as a result of the embargo. It is a simple idea
with a lot of potential that demonstrates the payoff to statistical re-
search.
O n e k e y c o n c l u s i o n f ro m a l l this is that military spending is likely to be excessive, driven up in an arms race spiral, and so be a
regional public bad. Collectively, the countries of the bottom billion
Guns
119
are spending around $9 billion on the military, of which up to 40 per-
cent is being financed by donors. Similarly, in regions where borders
are porous, a profusion of guns purchased by the government of one
country gradually seem to leak onto the informal markets of neigh-
boring countries. These cheap guns on informal markets increase
the risk of civil war. The final menace is that in post-conflict societ-
ies, which are usually big military spenders, the military is counter-
productive, provoking the very risk that it is meant to deter.
Not only is military spending excessive but aid is paying for it.
If the international community is minded to put matters right it has
two potential strategies: quantitative restrictions on arms purchases
or an incentive through linking aid allocations to the chosen level of
military spending. Arms embargoes can be made to work despite
their past lack of success. Guns are fueling the fire of political vio-
lence, and there is a need for them to be curbed.
C h a p t e r 5
WA R S : T H E P O L I T I C A L
E C O N O M Y O F D E S T R U C T I O N
Why are some places prone to war? Iraq has
deeply confused how people think about twenty-
first-century war. The war in Iraq is not a guide to
the future; it is a rerun of a phase in world history that is essentially
over. Iraq started with an international invasion. So did the two world
wars, Napoleon’s wars, the Crimean War, the Franco-Prussian War,
and the other great set pieces of military history. In the twenty-first
century international invasions are going to be infrequent. The wars
that will fill our television screens this century will be civil wars, not
international wars. Of course, there were civil wars in the nineteenth
century as well as invasions, but even the civil wars of the nineteenth
century are a hopeless guide to what warfare will be like. The major
civil war of the nineteenth century was the American Civil War. This
was in form, if not in law, an international war: one alliance of states
fought another alliance of states, each with its own recognized terri-
tory, government, and army. It’s history.
Future civil wars will take the form of a government pitted
against a private extralegal military grouping. They will variously
be called rebels, terrorists, freedom fighters, or gangsters, but their
essential characteristic will be the same. These wars will also be a
122
WARS, GUNS, AND VOTES
throwback, but to a very different period of history: the time before
nation-states cohered.
To rephrase the question, why are some countries more at risk
of civil war than others? If we could answer this question we might
be able to do something about it: some of the factors that elevate the
risk of civil war might be things that could readily be put right. I
think I have the answers to this question, although I am not fully
sure of them, even after studying the causes of civil wars for some
years. My approach is statistical: I take over all the countries in the
world for as long a period as possible and try to find what accounts
for why civil wars break out in some places at some times but not in
other places at other times: why some places are dangerous.
The core of my approach is to try to predict whether a country
has an outbreak of civil war on the basis of its characteristics prior
to the war. There are many pitfalls in this approach, but the key
problem is the lack of data. Records of the civil wars themselves are
not the problem. Astonishingly, a small team at the University of
Michigan, the university that pioneered the quantitative analysis of
political phenomena, has built a record of all the world’s civil wars
since 1815. There is even now a rival list built up in Scandinavia. But
for most of this period there are too little other data to match against
these outbreaks of civil war to try to explain them: for example, for
most countries reasonable economic data do not exist prior to 1960.
Even if the data were available, prior to 1960, for many years, virtu-
ally all the low-income world was in empires that kept the lid on
internal conflict. The period prior to empires would probably be re-
vealing, but there are as yet too few hard numbers for my approach
to be feasible. Even for the period post-1960 the countries that are
most likely to have a civil war also tend to be those least likely to
have adequate data on other characteristics: they are the dots and
blanks in the global tables produced by the international organiza-
tions. Fortunately, time has been on my side.
When Anke and I first tried the approach, in the late 1990s,
Wars
123
we could muster only twenty-three civil wars to explain. This was
pretty close to being hopeless. By the time of our next effort, pub-
lished in 2004, we were up to fifty-three civil wars and around
550 episodes during which a civil war could have occurred. This
was an improvement but it was still far from ideal. In our most
recent work we have been joined by Dominic Rohner, and time
has helped in three distinct ways. The most obvious way is that
there has been more time in which civil wars have occurred, or
more encouragingly, might have occurred but didn’t. Anke and
I work in five-year episodes, and whereas our previous analysis
took the story only up to the end of 1999, we are now able to take
it up to the end of 2004. Indeed, this is a very strategic additional
five years because it was the period of a major effort by the inter-
national community to settle wars, and so we can test whether it
&n
bsp; also reduced the incidence of outbreaks. But time has helped us by
more than just this. Scholars have been hyperactive in quantifying
phenomena that were not previously measured, and at filling in
the gaps in previous estimates, so that our data for the past are now
much nearer to being complete.
The third way in which time has helped is the embarrassing
one: we have got better at doing the work. We realized that we could
use a fancy statistical program that fills in the blanks of missing data
by randomly assigning a range of different numbers. I had always
been resistant to using make-believe numbers, but the advantage of
this approach was that it filled in each missing number with several
different possibilities, one at a time. Using these numbers in turn,
you could then see how robust the results were to the possibility that
the missing number would have taken these values. We used this
to check on our core results derived only from those numbers that
were genuine.
Another way in which we got smarter was better to control for
reverse causality or common causality, the current obsession in pro-
fessional economics. Take one of our core results, that low-income
124
WARS, GUNS, AND VOTES
countries are more likely to have outbreaks of civil war. Is this more
than just an association between low income and civil war: the two
phenomena tending to occur together but the first not causing the
second? The step from association to causality is tricky, and I will try
to spell out why. A rudimentary way of getting from association to
causality is the sequence in which events happen. If the low income
occurs before the outbreak of war, then this suggests that causality
runs from income to war. But is this enough? There are three ways
in which this step from association to causality could be mistaken.
One is that the civil war could be anticipated. If you know that
you are living in a country at risk of violent conflict, then you are
less likely to invest. And so the country will be tend to be poor due
to the war even prior to the outbreak of the war. But low income
has not caused the war; instead the prospect of the war has caused
low income. Another is that the country might have some charac-
teristic not included in our analysis that keeps causing civil wars: for
Paul Collier - Wars, Guns, and Votes Page 15