important, just as in the risk of civil war. Coups are more likely the
148
WARS, GUNS, AND VOTES
poorer the country and the lower its growth rate. So if the president
adopts policies that promote the development of the economy, he
becomes safer. Perhaps the president musters a flicker of interest
at this point, or perhaps his eyes glaze over: not another homily on
good economic policy. A further economic effect is via aid. After
allowing for the possibility of reverse causality, an additional 4 per-
centage points of GDP of aid increases the risk of a coup by around
a third. This may be because aid works like a honey pot, making
control of the government more attractive. So, inadvertently, donors
may be exposing governments to an enhanced risk of a coup.
What else did I find that might cheer a sleepless president? Ah,
yes, coups are getting less common with the passage of time: they are
gradually going out of fashion. The president concludes that all he
has to do is tighten the repression and hang on: time is on his side.
Unfortunately for presidents, this is in part offset by a countereffect.
Each year that a leader stays in power increases the risk of a coup:
far from gradually becoming indispensable, political leaders who
stay in power for decades overstay their welcome.
For an incumbent president the passage of time and the length
of incumbency offset each other. In any particular year, say 2008, a
president who has been in power a long time faces a higher risk of a
coup than one who is a newcomer. President Mugabe, who has been
in power for twenty-eight years, is more likely to find himself past
his sell-by date than President Mwanawasa of Zambia, who came
to power much more recently. Similarly, we can compare the coup
risks faced by two equally long-serving presidents at different times.
The year 2008 is President Mugabe’s twenty-eighth year in power.
For President Eyadéma of Togo the same long-service milestone of
twenty-eight years was reached back in 1995. At that time such long
service implied that a president was living dangerously, although in
the event Eyadéma reigned on and on until gathered up from the
presidential palace by the Grim Reaper. Now, thanks to the passage
of time, Mugabe is safer in 2008 than was Eyadéma in 1995.
Coups
149
What, apart from repression and economic development, can
a president do to guard against a coup? One strategy that is widely
touted in the literature is to divide the military into many different
units so that each one can function as a check on the others. Dur-
ing the Kenyan attempted coup of 1982 the government was saved
because the air force fell out with the army. In Zaire, President
Mobutu split his military into so many units that were not allowed
to communicate with one another that coups were made extremely
difficult. He did, however, pay a price for that strategy since the
same process made his security forces completely ineffective: despite
its massive size, Zaire was unable to defend itself from an invasion
by its tiny neighbor Rwanda.
Unfortunately, there are very few data on the internal struc-
ture of military forces, and so, while the divide-and-rule hypothesis
sounds eminently sensible, it is very difficult to test. We hit upon one
possibility: since landlocked countries did not have navies, all other
things equal, their military was likely to be less divided, and so a
coup attempt was more likely to succeed. We investigated whether
this was borne out empirically. Although we indeed found that coup
attempts were more likely to succeed in landlocked countries, the
effect was not statistically significant, so it may well be pure chance.
However, since the sample size for this test was small by the stan-
dards of statistical testing, little can be concluded from the lack of
significance. My guess is that divide-and-rule works. The president
is getting impatient reading all this: he has already divided his mili-
tary into seven distinct groups, each headed by a cousin.
So let me try to be more helpful. I have found something simple
and within the power of any president: adopt a term limit. At the
start of the 1990s term limits became fashionable. If an incumbent
president agreed to a limit of, say, two four-year terms, but declined
to make the rule retrospective so that the clock only then started
ticking, he had the prospect of a further eight years in power, and
that seemed a long time. The adoption of term limits significantly
150
WARS, GUNS, AND VOTES
and substantially reduces the risk of a coup: in fact it more than
halves the risk. Armed with term limits, the incumbent presidents
of the 1990s were a whole lot safer than they had been previously.
But as the fateful year approached when the term limit was actually
supposed to preclude them from continued power, presidents began
to have second thoughts. Should they really step down? Perhaps this
would be irresponsible? Surely they were indispensable? So, with
heavy hearts, they bowed to the pressure from their sycophants who
were themselves scared witless at the prospect of losing their access
to patronage. They began the process of changing the constitution
so as to remove the term limit.
The degree of difficulty that presidents faced in abolishing term
limits was, in fact, a good measure of how vigorously the society had
built constitutional defenses. The presidents of Chad, Zimbabwe,
and Uganda succeeded in abolishing them; the president of Russia
found an ingenious way around them by shifting to become prime
minister. The presidents of Zambia, Nigeria, and Venezuela tried to
abolish them but failed.
The evidence that term limits are effective in reducing the risk
of a coup is the most encouraging result so far: it suggests that to
some extent coups do function as last-resort checks on power. But
whether term limits will continue to be so effective in reducing the
risk of a coup depends upon their credibility. With so many presi-
dents waiting until near the end of their final term and then re-
moving the limit, those restive for power must now ask themselves
whether a term limit is merely a trick. For those presidents who
fooled their army into believing that they had committed them-
selves to an end-point only to remove it, the adoption of a term limit
actually had the perverse effect of lengthening their expected period
of office rather than shortening it.
To focus the mind, ask yourself what President Mugabe would
make of all this. He would note that the collapse of the Zimbabwean
economy has seriously exposed him to the risk of a coup. He would
Coups
151
note that this is compounded by his having been in office for twenty-
eight years, a mere blink of the eye compared with presidents Castro
and Ghadafi, but nevertheless rather a long time. As to term limits,
forget them. The only hopeful bit is that repression works
. But the
army is restive. What else could a worried president do?
The president controls the size of the military budget. If he is
worried about the possibility of a military coup, he could change the
budget. But life is complicated: in which direction should the presi-
dent change it? He is, it appears, on the horns of a dilemma. If the
army is the menace, then perhaps the safest thing is to slim it down.
If each officer is a potential Napoleon, then the fewer officers, the
safer the president. But offsetting this, if the army is demanding
more money, then perhaps the safest thing is to pay up. The presi-
dent dithers: up or down, which is best? At moments like this he
has learned to turn to the Internet, and sure enough, he swiftly finds
some research that provides the answer. At first the answer seems to
be as hedged around as a Delphic oracle, but he sorts it out.
In most countries for most of the time the risk of a coup is negli-
gible. If the head of the army comes along and starts muttering that
the troops are restive, then a sensible president tells the army what
he thinks of it. The chance of a coup being successful is so low that
no sane army officer would try it: the penalties for failure outweigh
the payoff to success because failure is so likely. So the extortion
threat is not credible. The president points to the example of Costa
Rica, which has managed perfectly well without an army, and cuts
the budget. We find something like this in the data. In the normal
range of coup risk, the level of military spending does not affect the
risk, and governments respond to small increases in risk by cutting
the military budget: if the military is a nuisance, you might as well
have less of it.
But there is a different range of coup risk. If the risk is high,
then the extortion threat becomes all too credible. The president
knows that a coup would have a sufficiently high chance of success
152
WARS, GUNS, AND VOTES
that the payoff might well be worth the risk. Unless he pays up he
is living dangerously. If, however, he pays up and does what the
army wants, then the payoff to a coup is reduced: he is safer. We also
find this in the data. In the range in which coup risk is high, a high
level of military spending significantly reduces the risk of a coup,
and, consistent with this, in response to a high risk the government
increases the military budget. I think of this as grand extortion: the
army is menacing the government for money in much the same way
as a gang of thugs run a protection racket, except that this is on a
grander scale.
So, to interpret the Delphic oracle of economic research, all the
president needs to work out is whether he is facing a high risk of
a coup or a low risk. If the risk is low, he can do what he would
sometimes rather like to do and slash the military budget, showing
all those useless officers with their gold braid what he really thinks
of them. If, however, the risk is high, then he had better raise the
military budget. He will have to steel himself to face down the dis-
approval of the donors as he raids the health budget and announces
a pay increase for the army. President Mugabe is in no doubt. The
meltdown of the economy, his long period of incumbency, the ab-
sence of credible term limits: each of these raises the risk of a coup.
He is in deep trouble. Repression plus money for the army seems to
be his best way out. As a preliminary measure he doubles the police
force.
I t i s t i m e t o r e t u r n to that awkward-looking question that I parked: do coups typically lead to improvement or deterioration?
We know that they do not come cheap, but if they are the only way
of removing a bad regime, then perhaps they are to be welcomed.
There are two aspects of the legacy of coups that we might rea-
sonably judge them by: their political consequences and their policy
consequences. We decided to use the standard measures for each
Coups
153
of them. For the political regime we used the score of the Polity
IV index, and for economic policy we used the rating of the World
Bank called the Country Policy and Institutional Assessment. Each
of these has limitations, but they are a reasonable guide to the legacy
of a coup. While the immediate impact of a coup might well be ad-
verse, to judge its legacy you need to look longer term. We decided
to look at the five years following a coup, year by year. We confined
ourselves to successful coups: only these produce regime change. Be-
fore I describe what we found, think for a moment what the effects
of a benign surgical-strike coup should look like. Quite possibly,
even though the coup is benign, in the first couple of years outcomes
might further deteriorate. But thereafter they should rapidly im-
prove. We might reasonably hope that five years after the regime
change, both the polity and economic policies should be significantly
improved.
Such hopes would not be justified. In the first couple of years
following a coup, the political regime does indeed significantly dete-
riorate. But even after five years it is still worse than before the coup.
The story is similar for economic policy. For the first three years
after a coup there is a sharp and significant deterioration, and even
by the fifth year policy is worse than prior to the coup. Indeed, if you
think back you may recall that one legacy of a successful coup was a
sharp increase in military spending. Not only do presidents increase
the military budget to try to ward off a coup, but if one succeeds, the
ringleaders reward the army by slamming up spending. No wonder
the World Bank assessment of economic policy deteriorates. There
is one further legacy of a coup: it significantly increases the risk of a
civil war. So the political legacy of coups is not particularly impres-
sive.
There remains one possibility that I must confess I have not
investigated. It may be that although actual coups are detrimen-
tal, the fear of a coup keeps politicians on their toes, forcing them
to deliver reasonable policies. I have not investigated it because it
154
WARS, GUNS, AND VOTES
is an extremely difficult proposition to test. Quantitative analysis
depends upon differences: differences in the risk of a coup would
have to show up as differences in government performance. I have
not got a sufficiently strong stomach to try: there are too many
obvious ways in which causality could flow in the other direction,
with differences in government performance affecting the chances
of a coup. I doubt that it is possible to find a test that would be
convincing, and there is considerable scope for results that are mis-
leading. However, accountability to the army need not improve
government performance for ordinary citizens: it may well im-
prove government performance for the army at the expense of or-
dinary citizens.
Evidently, if things are sufficiently desperate, a
coup is to be
welcomed. A coup is sometimes the only bloodless way of deposing
a disastrous and illegitimate regime, and in such circumstances mili-
tary officers do have a responsibility to take action. The alternatives
may come down to popular protest and rebellion. Recall that popu-
lar protest against autocracy becomes increasingly pronounced only
as incomes rise. At the very low levels of income that characterize
the bottom billion, protests are relatively rare and readily squashed.
Rebellions are too costly and unreliable to be a worthwhile route to
political change. So coups have a role to play in maintaining decent
governance, and the fact that they are getting less common is not
necessarily good news. Yet the historical record is not encouraging.
Surgical strikes do sometimes happen, but more commonly coup
leaders are not surgeons wielding a scalpel, but rank amateurs hack-
ing away at the body politic. To date, coups have been unguided
missiles that have usually hit the wrong target. Rather than be elimi-
nated, perhaps they need a guidance system.
C h a p t e r 7
M E L T D O W N I N C O T E
D ’ I V O I R E
Cote d’Ivoire brings it all together in one disas-
trous meltdown: a fraudulent election, a coup, another
coup, and a war. Yet it used to be known as the African
miracle. Its capital, Abidjan, was regarded as Africa’s Paris.
To make sense of the meltdown we need to start with the prior
success: what was the Ivorian miracle? Success had not been based
on democracy, but on the vision of an autocrat, Félix Houphouët-
Boigny. As you will see, his strategy was risky, but it nearly came
off. Along the way, the president transferred the capital to his home
village, Yamoussoukro, built an astonishing basilica modeled on St.
Peter’s, and induced the pope to come and open it. Since the ba-
silica was financed partly by the diversion of aid, it was viewed with
something between horror and derision by the donors. But societies
throughout history have used monumental buildings to construct
a shared identity. The anthropologist Colin Renfrew suggests that
Stonehenge was such an edifice, and as I will argue, the creation of
a sense of shared identity is very much what leaders should be doing
in these societies. Whether a cathedral in the president’s home vil-
Paul Collier - Wars, Guns, and Votes Page 19
