political leadership. A few political leaders of low-income societies
have succeeded in countering the problems posed by ethnic diversity
by superimposing a constructed national identity. Two outstanding
instances were Sukarno, who was president of Indonesia from 1945
until 1967, and Julius Nyerere, who was president of Tanzania from
1964 until 1985. More recently Nelson Mandela set South Africa on
the same path. Both Sukarno and Nyerere got their economic poli-
cies seriously wrong, falling victim to the fashionable nostrums of
their times, but on the key issue of building the nation they were
political giants. Sukarno had the more difficult task, a vast territory
of more than six thousand inhabited islands.
This has indeed always been how national identity comes about:
it is a political construction. But here I want to stick with the rare
Ethnic Politics
67
instances of the construction of a sense of nation in the new post-
colonial countries. What can leaders do?
Both Sukarno and Nyerere focused on language: indeed, lan-
guage is so fundamental to ethnic identification that it is the main
way in which social scientists have measured it. Sukarno created a
national language, Bahasa Indonesia, so simple that I have heard
Australian schoolchildren chatting away confidently in it. Nyerere
made Kiswahili universal across Tanzania. From now on I am go-
ing to focus on Nyerere’s strategy, for reasons that will soon become
clear.
Language was not the only strategy for surmounting tribal iden-
tity that he adopted. He took charge of the primary school curricu-
lum, inserting a heavy dose of pan-Tanzanian history into it. Chil-
dren were taught in school to see themselves as Tanzanians. While
language and education policies tried to reshape cultural identity,
Nyerere also transformed the processes whereby political decisions
were taken. He eschewed multiparty electoral competition, sensing
that it would be divisive. Instead, at the local level the colonial sys-
tem of enhancing the power of the tribal chief was completely up-
rooted. The national political party created village committees. At
the national level resources were allocated between localities, and
hence between ethnic groups, on principles of equity. Nyerere also
constructed physical symbols of national unity, most notably build-
ing a new national capital, Dodoma, in the center of the country,
an act much derided by the donors. Partly due to lack of funding,
Dodoma has not succeeded, but it clearly demonstrated his larger
purpose of moving beyond the inherited localized identities. Above
all, Nyerere developed and hammered home the rhetoric of national
unity: people were Tanzanians, and that was something to be proud
of. Ethnic identities were not forcibly suppressed; they were simply
downplayed. Even when Tanzania introduced multiparty politics it
was circumscribed: no party was allowed to campaign on an ethnic
platform. By chance, the current leader of the Tanzanian opposition
68
WARS, GUNS, AND VOTES
is an old friend of mine: a fine economist, he is about as far removed
from the politics of the gutter as it is possible to get.
Did Nyerere’s strategy work? That is one of those questions that
it is intrinsically difficult to approach scientifically. One guide is the
Afrobarometer survey, which asked the same attitudinal questions
in many African countries. One of the questions got pretty close to
the heart of identity: it asked, “Which specific group do you belong
to first and foremost?” and the potential responses were left open-
ended. Elsewhere in ethnically diverse African societies, nearly half
the responses were couched in terms of ethnicity: first and fore-
most people defined themselves in ethnic terms. In Tanzania only
3 percent responded with an ethnic or linguistic identifier. Having
to identify themselves more specifically than simply “Tanzanian,”
three-quarters gave their occupation. I think I would do the same:
proud as I am of my origins, I identify myself more strongly as an
economist than as a Yorkshireman.
But these responses to survey questions may reveal no more than
what is deemed acceptable in polite discourse: people may reply to the
interviewer by saying whatever makes them look good. Economists
are generally rather suspicious of reaching conclusions about behav-
ior just on what people say about themselves; we prefer to infer true
opinions from what people do. So the real issue is whether differences
in the sense of identity drive differences in behavior. This question is
more difficult. Difficult, but not, as it happens, impossible: Edward
Miguel of Berkeley recently did it. This is how.
Nyerere’s attempt at nation building in Tanzania stands in stark
contrast to political leadership in neighboring Kenya. Kenya’s first
president, Jomo Kenyatta, was in many respects also a great man:
his economic policies were far better than Nyerere’s. When Tan-
zanian socialists accused Kenyatta of running a “man-eats-man”
society, Kenyans aptly responded that Nyerere had built a “man-
eats-nothing” society. But Kenyatta could not bring himself to rise
above ethnic loyalty. He favored his own tribe, the Kikuyu, mas-
Ethnic Politics
69
sively skewing public resources to the Kikuyu heartland. Like many
African leaders, Kenyatta had not made adequate preparations for
his own succession. Two of Kenyatta’s henchmen, both Kikuyu,
wanted the job, and each blocked the other. In a sea of confusion
they decided to appoint someone so hopeless that they could rule by
proxy: they chose a poodle from a minority tribe. Step forward onto
the world stage President Daniel arap Moi. In one key respect Moi
was considerably less hopeless than the kingmakers had anticipated:
he swiftly marginalized both them and the Kikuyu selectariat. Ev-
erything was reversed except for one constant: massive favoritism
toward the president’s own tribe, the Kalenjin.
As it happens, the Kalenjin tribe is itself a nice demonstration
of how identity can be constructed. You might imagine that African
tribes go right back to the primordial times of the birth of man.
In fact, the Kalenjin go back all the way to 1942. With the Second
World War being fought out in North Africa, the British wanted
recruits for the Kings African Rifles and, sensibly enough, targeted
their recruitment toward a large low-income area. The cheapest
means of recruitment was to use the radio, but the area covered a
wide range of dialects. Choosing one of the dialects in the middle
of the range, each radio broadcast opened with the attention-grab-
bing phrase “I tell you, I tell you,” not, of course in English, but in
the dialect: “Kalenjin, kalenjin.” In the appalling aftermath of the
2007 Kenyan elections, the Kalenjin led the violence. The tribe is the
product of a radio program. Such is the stuff of ethnic identity.
While b
oth Kenyatta and Moi favored their own tribes, neither
devoted any serious priority to building a sense of national iden-
tity. There was no attempt to create a national language, and in the
school system the history of each locality was given precedence over
national history. Politically, the colonial system of chiefly power was
largely left in place: the local big man became all important. As to
interethnic equity, forget it. And despite its greater wealth, Kenya
made no effort to build national symbols such as Dodoma.
70
WARS, GUNS, AND VOTES
The Kenyan elections of December 2007 provided an oppor-
tunity for a new set of politicians to fan the flames of a fire that
had been lit by their predecessors. By far the main culprit was the
opposition leader, Raila Odinga. Recall that the incumbent has the
advantage in respect of bribery and miscounting, so the opposition
is indeed more likely to resort to the cheaper strategy of playing
on ethnic identity. Odinga ran a campaign that was tantamount to
promising ethnic cleansing. His strategy was electorally successful
because the Kikuyu, whom he targeted, constituted less than a quar-
ter of the population. Odinga probably won the most votes. That
he lost the election is probably due to ballot fraud. But if so, he was
cheated out of a victory that was won by a strategy that in a proper
democracy would have been illegal.
The difference in post-independence political strategies be-
tween Tanzania and Kenya was sufficiently stark to lay the founda-
tions for a natural experiment: an attempt to build a sense of national
identity, versus an attempt to reinforce tribal identity. However,
a natural experiment needs much more than divergent strategies:
the two places need to be otherwise comparable. The two countries
were indeed pretty similar and certainly ethnically diverse: Kenya
had forty-eight tribes, Tanzania even more. Miguel enhanced these
country-level similarities by focusing on two districts, one Kenyan,
the other Tanzanian. He selected them because they were even more
similar than the countries themselves: Busia in Kenya, and Meatu in
Tanzania. The international border, established in colonial times,
had basically driven an arbitrary straight line through what until
then had been one area. But divergent strategies and comparability
are still not enough for a natural experiment. There needs to be some
quantitatively measurable difference in outcomes: identity is a slip-
pery sort of entity to observe. Miguel decided to measure the supply
of some key public goods, such as the amount of money raised lo-
cally for schools, the provision of school facilities, and whether wells
were in working order.
Ethnic Politics
71
But if Busia was to be one observation and Meatu the other,
there was not going to be any statistical power whatsoever: either
Busia is going to be better than Meatu or it is going to be worse,
and a priori, there is a 50 percent chance of finding either outcome.
Miguel’s key inspiration was to use the fact that both Busia and Me-
atu were composed of many localities. Some of these localities had
high degrees of ethnic diversity whereas others were homogenous.
He realized that he could use these differences in the degree of di-
versity between different localities within Busia and Meatu to see
how much damage diversity was doing in each society.
In Busia, the Kenyan district, he found exactly the pattern that
researchers have usually found when they investigate the conse-
quences of ethnic diversity. The more diverse localities within Busia
had worse public-goods provision than the more homogenous lo-
calities. What is more, the effect was really big. The average, fairly
diverse locality had 25 percent less school funding per pupil than
the homogenous localities. This was a problem fully recognized by
head teachers in the ethnically diverse schools: they blamed ethnic
rivalries for the unwillingness of parents to support the school.
How about Meatu, the Tanzanian district? The key test in the
research design was whether ethnic diversity was similarly damag-
ing there. There was just as much variation between localities in
Meatu as in Busia: some localities were highly diverse and others
were homogenous. It turned out not to matter at all: diversity had
no discernible effect on public-goods provision. The statistics were
supported by the interviews: Miguel received comments such as
“We’re all Tanzanians” and “This is Tanzania, we do not have that
sort of problem here.”
I hope I have given you a flavor of Miguel’s study: it was, in
fact, a beautifully crafted piece of social science. It is important be-
cause it provides pretty convincing evidence that Nyerere’s strategy
of building national identity had actually worked. Over a period of
forty years, between independence and the survey on which these
72
WARS, GUNS, AND VOTES
results are based, the damage normally caused by ethnic diversity
had been dramatically reduced and perhaps even eliminated. Nyer-
ere had turned a new country into a new nation.
Nyerere and Sukarno showed what could be done by leader-
ship. Unfortunately, their approach was rare in the societies of the
bottom billion. Far more common was that of Kenyatta and Moi in
Kenya, where the consequences of a strategy of emphasizing ethnic
identity over Kenyan identity are now all too apparent. As I write
this I am trying to follow events in the aftermath of Kenya’s election.
Around one thousand Kenyans have died in ethnic violence. It is
hard to discuss research in such a context. But recall that in Nigeria
Pedro Vicente and I had conducted surveys during the presidential
elections of April 2007. Since they had proved feasible, I decided
to try the same approach during the Kenyan elections, which I an-
ticipated would be rough. I put together a team. As you have seen,
ethnic diversity in teams can be a source of strength: ours had a Ke-
nyan, an American, a Belgian, a Mexican, and a German. This work
was so recent that I can report only a few preliminary results.
The survey was conducted prior to the explosion of violence
that followed the election. But even at this stage, five in every six Ke-
nyans feared becoming victims of political violence, and one in ten
had already been threatened about the consequences of voting the
wrong way. Just as in Nigeria, electoral violence looks to have been
a strategy of the weak: it was the government supporters who were
the most fearful, and events proved them right. But the threats did
not well upward from community-based antagonisms. The incite-
ment to violence was seen as coming down from the organizations
of the political parties. Violence against the Kikuyu was a deliberate
electoral strategy of Raila Odinga.
Consistent with the allegations of fraud that followed the gov-
ernment declaration of victory, we found that at the time of our sur-
/> vey, which was a few days prior to the elections, the opposition was
poised to win. Nor would this have come as a surprise to the Kenyan
Ethnic Politics
73
electorate: when asked how free and fair they expected the elections
to be, 70 percent expected problems, and these fears were dispro-
portionately high among opposition supporters. Ethnicity was all:
only half of voters regarded their primary identity as being Kenyan.
More revealingly, voting intentions were massively skewed by eth-
nicity. Not only did the Kikuyu vote for Kibaki and the Luo vote
for Odinga, but even the tribes other than those of the candidates
largely voted as ethnic bloc votes.
But here are the results that I think toll the death knell for eth-
nic politics. They concern the discipline that electoral competition is
supposed to provide on government economic policies. In the years
leading up to the election the Kenyan economy had been doing
rather well: its fastest growth for more than two decades. Nor had
the benefits of growth been confined to the Kikuyu. Even the Luo
recognized that they had become better off. Kibaki even managed
to get amazingly strong approval ratings from Luo respondents. It
didn’t help him. He was the wrong tribe and they were not going to
vote for him: 98 percent of the Luo voted for Odinga. With this sort
of voting behavior, there is little incentive for a president to provide
national public goods: he might as well favor his own. The strong
ethnic identities that Kenyan political leaders had fostered had ef-
fectively deprived electoral competition of its potential for hold-
ing a government to account. As for the other supposed benefit of
elections, legitimacy, here is another comment from Koki Muli, the
head of Kenya’s Institute for Education in Democracy: “Do these
people not care about legitimacy?”*
* “Kabaki Win Spurs Kenya Turmoil,” Financial Times, December 31, 2007, p. 6.
C h a p t e r 3
I N S I D E T H E C A U L D R O N :
P O S T - C O N F L I C T S E T T L E M E N T S
With the millennium came peace. The
international community finally started to pay
serious attention to the running sores of long-
lasting civil wars. Peace conferences were called, pressure was
Paul Collier - Wars, Guns, and Votes Page 9