need for them. So it is not surprising that the post-conflict election
should be used as the milestone, in the surreal technical jargon of
peacekeeping, for troop withdrawal. Or, in the more familiar sound
bite, elections are the exit strategy. How this strategy played out in
the Democratic Republic of the Congo I will return to shortly.
If you think about it, our results suggest that a post-conflict elec-
tion is inappropriate as a milestone: it is more like a tombstone. Of
course, it depends whether peacekeeping works: if it doesn’t work
then the boys might as well be brought home and any sort of stone
will do. So it is time to turn from elections to peacekeeping.
We asked the United Nations for data on its peacekeeping op-
erations. The good news was that they had pretty complete records.
Unfortunately, the records were not organized for quantitative
analysis: it took our research assistants seven months to put them
into shape. But eventually we had information, country by country
and year by year, on the numbers of troops and their cost. It was time
to see whether peacekeepers helped to keep the peace. Somewhat to
our surprise we got clear results: peacekeeping seems to work. Ex-
penditure on peacekeeping strongly and significantly reduces the
risk that a post-conflict situation will revert to civil war.
By now you will realize that the standard concern is whether
such results are spurious because of reverse causality. For example,
if the troops are systematically sent only to the safer post-conflict
countries, they will appear to be successful in keeping the peace but
the result will not be causal. And so we tried to find something that
would explain the allocation of peacekeeping troops but that was
otherwise unrelated to the risk of conflict reversion. Whatever we
tried, we were unable to get a good explanation for the allocation of
troops, and so we turned to the academic literature. Nicolas Sam-
banis, a young Greek political scientist whom I had once worked
with, had just coauthored a book on post-conflict peacekeeping with
Michael Doyle, who used to be head of research at the United Na-
84
WARS, GUNS, AND VOTES
tions and is a world authority on peacekeeping. They had concluded
that the political decision process that assigned troops to post-con-
flict situations was so complex as to defy being modeled. The vari-
ous members of the Security Council who took the decision were
involved in such byzantine horse-trading that any particular deci-
sion was close to being random. This explained why we were unable
to find good predictors and also suggested that we were not facing a
severe problem of reverse causality.
Nevertheless, we were able to make one helpful check. The de-
cision as to how many troops to send into a post-conflict situation
can conceptually be split into two stages: first, should troops be sent
at all, and, then, if it is decided to send troops, how many should be
sent? We realized that we could learn something about the motiva-
tions for sending troops by looking at that first decision: should they
be sent at all? We found that the decision to send troops at all was
associated with a significantly higher risk of reversion to violence.
The most plausible way of interpreting this is that troops tend to
be sent to places that are more at risk. We cannot tell whether the
same is true of the decision as to how many troops to send. We just
know that given the decision to commit troops, the more that were
sent, the safer was the society. If, in fact, the number of troops sent is
motivated by the same concerns as seem to motivate the prior deci-
sion of whether to send any, then they are being sent in the greatest
numbers to the most dangerous places. How would this affect our
results, which implicitly assume that they are assigned randomly?
Its effect would be that our results would understate the true effectiveness of peacekeepers. The truth would be that places with many
peacekeepers have a lower rate of reversion to conflict despite intrinsically being more at risk. So our assumption that their numbers are
unrelated to intrinsic risk may well be conservative.
I had the results on post-conflict elections and on the efficacy
of peacekeeping by the summer of 2006 and shared them with the
appropriate parts of the international community. I was particularly
Inside the Cauldron
85
concerned that the proposed strategy in the Democratic Republic of
the Congo of troop withdrawals the day after the election, which was
due to be implemented within a couple of months, looked unwise. I
was promptly invited to address the new Peace-Building Commission
of the United Nations, and also shared the results with the French
government, who were supplying the largest component of the peace-
keeping troops. I learned that the military commanders were them-
selves highly skeptical of the plan for troop withdrawal. In the event,
the aftermath of the election became so violently unstable so rapidly
that instead of troops being flown out, they had to be flown in. Within
a few months there was a shoot-out between the private army of Be-
mba, who lost the election, and the government army of Kabila II, the
incumbent winner. Bemba’s forces lost the shoot-out, and he himself
sought protection in an embassy before fleeing to Europe, where he is
now in exile. His exit has not restored order: the Democratic Republic
of the Congo continues to be dangerous.
Even if international peacekeeping is effective it faces problems.
It is expensive and unpopular. Some of the post-conflict govern-
ments get indignant about the intrusion: the Department of Peace-
keeping Operations of the United Nations (DPKO) has become the
new International Monetary Fund, a challenge to the unrestrained
sovereignty of governments keen on asserting their power. It is also
understandably unpopular with the electorates of the countries that
supply the troops: no one wants his son or daughter to be exposed to
the risks of peacekeeping.
Is there an alternative? I could think of two other possibili-
ties. The first is what is known as over-the-horizon guarantees. It is
what the British government is doing in Sierra Leone. For the past
few years there have been only eighty British troops stationed in the
country, but the government has been given a ten-year undertaking
that if there is trouble, the troops will be flown in overnight. Perhaps
this has helped stabilize the society. Sierra Leone is, at least in terms
of reversion to violence, a major success. It has even weathered post-
86
WARS, GUNS, AND VOTES
conflict elections and a change of government. The problem with
the Sierra Leone example is that it is just that: one example. You
cannot perform a statistical analysis on one observation, and so there
is no way of knowing whether in general such guarantees would be
effective. Or is there?
I started to think whether there was anything analogous in
> the past to what the British are now doing in Sierra Leone. Sure
enough, the French had provided security guarantees to their cli-
ent countries in Africa for years. In fact, with the typical logic of
international coordination, they had abandoned it only just before
the British started to do it. The French security guarantees were
informal, but they were most surely for real. They were backed by
a series of French military bases across Francophone West Africa.
They had started with independence and rolled on until the French
government got caught up trying to implement its informal guaran-
tee defending the Hutu regime in Rwanda in 1994. If you remem-
ber, there were French troops stationed in Rwanda as the Tutsi rebel
forces invaded from Uganda and as the Hutu regime set about its
genocide. The French came disturbingly close to finding themselves
propping up a regime implementing genocide and only just pulled
back in time. After that President Chirac ordered a rethink and an-
nounced a new policy toward Africa: military intervention began
to look anachronistic. The first test of this new policy was the coup
d’état in Cote d’Ivoire in 1999. The French old guard wanted to
intervene to put it down, but President Chirac vetoed intervention.
So we can date the credible prospect of French intervention from
independence until the mid-1990s. After their military catastrophe
of the battle of Dien Bien Phu in Vietnam the French were in no po-
sition to extend their military guarantee across the whole of the Fran-
cophone world; it was basically credible only in West and Central
Africa, and it had lasted for around thirty years. This was, however,
a large enough group of countries, for a sufficiently long period, to
be amenable to statistical analysis.
Inside the Cauldron
87
The key question was whether this guarantee had actually re-
duced the incidence of civil war. This question needs a model of the
risk of civil war. Such a model can be used to address a range of im-
portant questions, but here I will just give you this particular answer.
Did the French informal security guarantee reduce the incidence of
civil war? We found that it was highly effective. Francophone Af-
rica had characteristics that would otherwise have made it prone to
warfare: the actual incidence was much lower than would have been
expected. Statistically, the guarantee significantly and substantially
reduced the risk of conflict by nearly three-quarters.
But was the military guarantee the reason for this remarkable
reduction in conflict? Could it have been something else associated
with the French presence? For example, in response to French op-
position to the invasion of Iraq, some Americans accused the French
of an excessive aversion to force: what was that ill-judged phrase,
“cheese-eating surrender monkeys”? Perhaps French culture had
inculcated pacific values? While this might seem implausible to
anyone aware of French military history, we decided that no stone
should be left unturned. If the reduction in the risk of conflict was
due to culture rather than the security guarantee, it would reach the
parts of La Francophonie that the guarantee did not cover. It didn’t:
the enhanced security was unique to West and Central Africa, the
region covered by the French military bases. To my mind this is rea-
sonably convincing. Over-the-horizon guarantees look as though
they work. As I was finishing this book Chad exploded into civil
war: rebels reached the gates of the presidential palace. As the crisis
unfolded the French position rapidly shifted. Initially the French
announced that they had no intention of intervening militarily.
Within a week they had thought better of it and issued a security
guarantee: the rebels would be repelled by French force unless they
withdrew. The French had a large military base right there in Chad:
the rebels withdrew.
It is time to move on from the politics and the military. What
88
WARS, GUNS, AND VOTES
else drives post-conflict risks? Surely the economy enters some-
where? In fact it enters twice over. The lower the income, the higher
the risk of conflict reversion; and the slower the economic recovery,
the higher the risk. Both of these have implications. If low-income
countries face higher risks of conflict reversion, other things equal,
the international community should be allocating peacekeeping
troops disproportionately to those post-conflict situations with the
lowest income. This would indeed provide a useful rule-of-thumb
to cut through all the horse-trading that Doyle and Sambanis found
to be dominating the decision on the Security Council. A further im-
plication is that, again other things equal, strategies that enhance the
economic recovery are going to be peace-enhancing: raising growth
and cumulatively augmenting the level of income.
So how to rebuild a shattered post-conflict economy? The
problem with economic interventions is that they are not exactly the
cavalry. It is possible to destroy an economy quite rapidly, as Presi-
dent Mugabe has convincingly demonstrated, but putting Humpty
together again takes time. If average income can grow at 7 percent
a year, which is entirely possible in post-conflict situations, then the
level of income doubles in a decade, and so by the end of the decade
risks are substantially lower. But this is the time frame for economic
recovery, not two or three years.
The story so far is that the post-conflict decade is dangerous
and that there seems to be no clear political quick fix. In particular,
elections and democracy, at least in the form found in the typical
post-conflict situation, do not bring risks down. Economic recovery
works but it takes a long time. The one thing that seems to work
quickly is international peacekeeping, but it is politically difficult to
sustain peacekeeping for the length of time needed for the economy
to recover. Is prolonged peacekeeping necessary, even if only in the
form of over-the-horizon guarantees? There is one remaining pos-
sibility. Perhaps the key risks occur early in the decade and are fol-
lowed by a safe period. In that case peacekeeping could be brief.
Inside the Cauldron
89
That would make it politically much easier. Since an option that is
politically easy is far more likely to be adopted, it was worth inves-
tigating. The risk of going back to conflict does seem gradually to
decline with time, but don’t hold your breath. Time heals, but its ef-
fects seem to be decade by decade rather than year by year. The first
four years after the end of a conflict are perhaps somewhat more
dangerous than the next six, but the effect is not statistically signifi-
cant. Within the post-conflict decade there is no safe period.
So where does this leave us? Economic recovery is to my mind
the only genuine exit strategy for peacekeeping. I think we need to
dismiss the illusion that elections are the milest
one and face the long
haul of building the economy. It may well not be necessary to keep
many peacekeeping troops on the ground throughout the decade:
an initial military presence may well be able to evolve successfully
into an over-the-horizon guarantee. But any such guarantee must
be credible: the French guarantee was made credible by its military
bases, and the British guarantee was credible because during the
conflict they indeed flew in overnight to check the Revolutionary
United Front (RUF) forces that were set on occupying the capital,
Freetown. The British forces held off the RUF on the outskirts of
the capital at little place called Waterloo. But they only arrived just
in the nick of time: as Lord Wellington said of the former, some-
what grander battle, it was “a damned close thing.”
S o , i f e c o n o m i c r e c ov e ry i s the exit strategy, how can it be facilitated? What policies work, and can donors help? Anke and I
had already done a little work on the payoff to post-conflict aid: we
found that it was significantly more effective than aid at other times.
This is not surprising: post-conflict recovery was the initial rationale
for the international aid agencies. But I decided it would be worth
looking more closely at what could be done to revive the economy.
For this work I teamed up with Victor Davies, a doctoral student
90
WARS, GUNS, AND VOTES
from the classic post-conflict country of Sierra Leone, and Chris
Adam, a close colleague at Oxford. I also worked with Marguerite
Duponchel, a doctoral student at the Sorbonne, where I teach as a
visitor. Although I will try to make what you are about to encounter
come across as a seamless web of research, it was not like that at the
time.
Some important uses of aid for post-conflict are blindingly ob-
vious: it pays for the reconstruction of infrastructure. But here is
one that is much less obvious: countering inflation. High inflation
is a pretty disastrous macroeconomic strategy: essentially a policy of
desperation. Normally governments keep inflation moderate. This
is despite the fact that in the short run they could fleece the econ-
omy by printing money. Inflation is a tax that most people do not
recognize as a tax. Governments restrain themselves because of the
alarming dynamics that lead to hyperinflation. The only govern-
Paul Collier - Wars, Guns, and Votes Page 11
