to guard against the pitfalls you fear. The world had the decency
to conform to the theorem: controlling for everything else, if the
neighbors spent more, then so did the country itself.
We haven’t quite done with proxies for external threats. In
fact, there is one more that is utterly obvious. If you were free to
chose which neighbor you faced, would you be more frightened of
China or Bhutan? Never mind the politics, or the share of military
spending in income, China is intrinsically more of a threat because
it is so much bigger. More populous countries systematically spend
a smaller share of income on the military. This is an instance of a
proposition that is going to loom large throughout this book: secu-
rity is subject to economies of scale. Big may not be beautiful, but it
is safe: small is dangerous, and expensive.
108
WARS, GUNS, AND VOTES
External threats have historically dominated our thinking on de-
fense because of the horrors of international warfare that character-
ized the last century. There are entire academic departments of inter-
national relations that studied the subject. But actually, international
warfare is largely a thing of the past. The main drivers of military
spending in the countries of the bottom billion are now to be found
within their own societies. The threats are internal, not external.
The most obvious internal security threat for which an army
might be useful is to counter a rebellion. Sure enough, if a govern-
ment is engaged in fighting a civil war, its military spending leaps
by around 1 percentage point of national income. Civil wars are a
lot more common than international wars, and on average they last
more than ten times as long. So this form of warfare is more impor-
tant as a driver of military spending in the bottom billion than is
international warfare. Anke and I estimated that for Africa it was
about twice as important.
But even civil war is not that common. As is the case with inter-
national war, governments spend most of their time fearing it rather
than fighting it. Anke and I developed a model of the risk of rebel-
lion and used the estimates to see whether governments responded
to a heightened risk of rebellion with increased spending. We found
a large effect: governments tried to buy security from the threat of
rebellion by building a big army.
This is a case of out of the frying pan into the fire. A big army
is not just a source of defense, it is an interest group. The nearest we
got to the military-industrial complex was to investigate whether the
military was concerned to look after itself. Quite commonly, profes-
sions tend to lobby in their own interests. Plato had the brilliant idea
that the ideal government would be composed of philosopher kings:
government by professors. Unfortunately, Plato’s splendid idea has
not been implemented with sufficient frequency for it to be amena-
ble to statistical testing, but I hazard that a government of professors
would spend significantly more on universities.
Guns
109
Whereas professors have seldom run governments, with gener-
als it is an entirely different story. Sometimes citizens freely choose
to elect generals because they are war heroes: American citizens
elected General Eisenhower. But more commonly, generals come
to power by a somewhat different route: they elect themselves. Mili-
tary rule happens with sufficient frequency that its consequences
for military spending are amenable to statistical testing: does the
military do for military expenditure what I expect professors would
do for university expenditure? They most certainly do. Following
a coup d’état military spending leaps, and quite generally, military
regimes spend much more on the military, even controlling for the
security risks. Of course, this may well not be ill motivated. Gener-
als, colonels, and what have you may not be driven by squalid ideas
such as “It’s our turn at the trough.” They may be thinking, “At
last we can give the defense of our nation the priority it truly war-
rants.” So the phenomenon of the military spending on itself may
be well motivated. I have come to doubt the efficacy of dissecting
motives: I prefer to look at consequences. Leaders can deliver dire
consequences despite fine motivations.
So far I have considered military spending from the perspec-
tive of what governments might feel they need to spend, or would
like to spend. But as with any other type of expenditure, it has to be
tailored to what can be afforded. Although we measured military
spending as a share of income, this still leaves out a lot of affordabil-
ity considerations. One potentially important consideration is how
rich the society is. Rich people spend a higher share of their income
on luxuries than poor people. This is not a moralizing statement, it
is a definition. Economists define luxuries as those items of expenditure that increase more than proportionately as income rises. The
opposite of luxuries is necessities. Things that are necessities have to
be bought even if income is low: that is why food purchases form a
much larger share of spending for poor people than for rich people.
The current surge in world food prices is an irritant to the budget
110
WARS, GUNS, AND VOTES
of a household with a decent income, but devastating for the world’s
poor: food is half of their expenditure.
So is military spending a necessity or a luxury? Well, we know
from politicians that military security is the most vital priority: they
tell us often enough. In economic language things that are vital pri-
orities are necessities: you buy them even if it means sacrificing the
purchase of things that are less necessary. So military spending must
be a necessity. It is one of those propositions that we know must be
true, but nevertheless it is best to look: known truths, like theo-
rems, sometimes turn out to be wrong. Indeed, it is wrong. Mili-
tary spending came out as a clear luxury: it increases much more
than proportionately with income. In one sense this is good news:
the poorest countries tend to spend a lower proportion of their in-
come on the military. But the unfortunate aspect is that with global
growth military spending is going to loom larger and larger. While
politicians tell us how necessary it is, they behave as though it were
a government luxury good.
There was another aspect of finance that we decided to ex-
plore. Low-income countries get substantial amounts of aid. We
wondered whether any of this inadvertently financed the military.
Any such effect would be inadvertent because aid for development
is distinguished very sharply from aid for the military: it is meant
to pay for education, infrastructure, and suchlike. Aid intended for
military support is recorded separately. Back in the 1960s the United
States gave about equal amounts of money for development aid and
for military aid. Gradually th
e balance has shifted: now military
aid is far less than development aid. This change in priorities was
not foolish: as you will see, security comes with development rather
than with guns.
Whether aid leaks into military spending is an easy question
to pose but a difficult one to answer. The difficulty is in sorting out
causality. We want to test whether aid causes military spending,
but it is quite likely that causality also runs in the other direction.
Guns
111
Those governments that choose to have high military spending may
get less aid because donors disapprove of their priorities. To test for
whether aid leaks into military budgets, we need to focus on varia-
tions in aid that are not caused by donor reactions to military spend-
ing. Economists realized how to do this in 2003, and we followed
the approach that has now become standard. It relies upon the fact
that different donors give money to different countries, largely for
historical reasons. The Italians give to Ethiopia, a former colony,
and the French give to Cote d’Ivoire. Further, national aid budgets
go up and down according to the national economic cycle. So, when
the Italian economy booms relative to the French, Ethiopia tends to
get an aid windfall relative to that of Cote d’Ivoire. Crucially, this
is not caused by anything happening in either country, so if military
spending goes up in Ethiopia and down in Cote d’Ivoire, it is either
a coincidence or aid has changed military spending. Coincidence is
always a possibility, but it becomes a small possibility if the number
of observations is large: this is what is meant by statistical signifi-
cance.
So, after all that, what did we find? We found that aid does
indeed leak into military spending: on average around 11 percent of
aid finds its way into the military budget.
There are many ways in which this might happen. The most
evident is that aid that is ostensibly earmarked for some particular
expenditure in fact releases the money that the government would
otherwise have had to use for that expenditure. The only way to
avoid this consequence of aid would be for the donor to insist on
items of expenditure that the government categorically does not
want. Yet over the past decade there has, quite reasonably, been a
massive shift in the philosophy of aid toward country ownership.
The government itself sets the agenda for how the aid is earmarked.
So it has probably become even easier for governments to use part
of the aid flow for military spending. If 11 percent of aid leaks into
military spending, since total aid to the countries of the bottom bil-
112
WARS, GUNS, AND VOTES
lion is $34 billion, the inadvertent leakage into military spending
totals $3.7 billion. Since total military spending of the bottom bil-
lion is around $9 billion, this implies that around 40 percent of the
military budgets of the governments of the bottom billion is being
financed by aid. Hopefully, this may overstate the problem. Even if
on average the leakage s 11 percent, if donors manage to skew aid to
the countries where leakages are below the average, the total would
be less than 40 percent.
If we combine this result with the evidence on neighborhood
arms races, it has a potentially disturbing implication. Maybe aid is
inadvertently financing arms races in Lilliput. As we are about to
see, it has a yet more disturbing implication in the context of post-
conflict societies.
Do guns deter civil war? It may well be that once a rebellion
is under way high military spending can squash it, but prior deter-
rence is potentially a different matter. Whether guns deter is another
of the questions that is easier to pose than to answer. Since the need
for deterrence is one reason for having plenty of guns, it is possible
to get into a statistical muddle. High spending is likely to be associ-
ated with a high risk even if in fact it reduces risk. You might recog-
nize this two-way causality as being the same underlying problem
as that posed by aid and military spending. The underlying solution
is the same: find something that influences spending but does not
otherwise affect the risk of war. In the parlance of economics, a sub-
ject desperate to look scientific, such influences are termed instru-
ments. In principle, from the differences in military spending that
are only due to such influences, it should be possible to tell whether
the spending causes a change in risk: that is, whether it deters.
We followed this approach and concluded that military spend-
ing did not deter: in fact it did not seem to have any discernible
effect. This may have been because the instruments were not good
enough: in social science it is far more difficult to show convincingly
that something does not matter than to show that it does matter. But
Guns
113
we went one step further, and this was, to my mind, rather more
convincing. As you have seen, the most risky environments are the
post-conflict situations. We asked whether the military spending of
the government was differentially effective in such situations. We
found that there was indeed a significant difference, but it was per-
verse. Far from deterring violence, high military spending by a post-
conflict government provoked it.
That is where our work had got to by the time that Anke was
due to give birth to Henry: we hastily packaged it up and got it pub-
lished. I could try to pretend that our research is an entirely ordered
and coherent sequence of steps along which we march to discover-
ies. It should be, but it isn’t. Somehow I managed to work on the
question of post-conflict risks along two parallel tracks, producing
two different journal articles on distinct but related issues. I realized
that our analysis of military spending in the post-conflict context
had inadvertently omitted other influences that are specific to such
situations, most notably peacekeeping. Evidently the two analyses
should be merged. With Havard Hegre, a young Norwegian po-
litical scientist, I combined them. We again explored the effects of
the government’s post-conflict military spending, but now alongside
peacekeeping. To my relief the result survived: high military spend-
ing significantly increased the risk of further conflict.
As to why it has this effect, we have to leap beyond what statis-
tics can tell us and try intelligent speculation. My own guess is that
the decision of the government to spend on the military inadver-
tently signals to citizens that it is planning to turn nasty, and that
this signal forewarns those rebels who have recently put down their
arms that they were unwise to do so.
I have just taken you through two distinct results: aid leaks into
military spending, and in post-conflict situations military spending
increases the risk of reversion to war. Now for the dilemma: aid
is highly useful in post-conflict economic recovery, which in
turn
brings down the risk of further conflict. Putting all this together, aid
114
WARS, GUNS, AND VOTES
in post-conflict situations is currently a two-edged sword, restoring
the economy but inadvertently inflating dangerous spending.
So far I have focused on government military spending, but
how do the rebels get their armaments? One route is that hostile
governments buy guns for the rebel groups that oppose neighbor-
ing governments. But this is not the only way in which government
guns find their way into informal usage. Guns purchased for official
use leak into the hands of rebels.
The weapon of choice for any self-respecting rebel movement
is the Kalashnikov. One reason for this is that it is such a simple
weapon that it seldom goes wrong, needs little maintenance, and so
can be entrusted to someone who is pretty clueless. This is important
because rebel recruits are generally young, ill educated, and ama-
teur. The other reason that Kalashnikovs are popular is that they are
cheap. That in turn is because the Soviet Union produced vast quan-
tities of them and licensed their production in some of its satellites.
The latest insightful political leader to set up a Kalashnikov factory
is President Chavez of Venezuela. Presumably he plans to present
them to friendly states.
Economists are particularly interested in prices. I remember
one distinguished economist saying to me when I was a young re-
searcher: “It’s all we’ve got.” What he meant was that the heartland
of economics is its theory of behavior: we assume that people and
firms try to maximize something, subject to whatever constraints
they face. The key constraint they face is the prices. Armed with this
insight, economists can predict how behavior will change when the
constraints change. If a price falls, people will buy more of the good.
So if Kalashnikovs are cheap, rebel groups will buy more of them.
Political scientists do not share this fascination with price, and they
are the profession that dominates the study of violent conflict. So the
data on guns were all done in terms of quantities: there were plenty
of estimates of how many guns were being traded, but no figures on
their prices.
Guns
115
For some years I had been trying to find price data: instead I
found Philip Killicoat, an Australian graduate student. He volun-
Paul Collier - Wars, Guns, and Votes Page 14
