Kenyans dont always vote tribally

Mutahi Ngunyi’s “Tyranny of Numbers” article – now tyrannizing the unwary in cyberspace – says that the Jubilee Alliance of Uhuru Kenyatta will win the 2013 presidential election in the first round with a substantial majority over the Cord Coalition of Raila Odinga. In contrast to this, numerous opinion polls have consistently placed the CORD Coalition ahead of Jubilee with the most recent poll saying that the election is now a virtual tie. The polls say, in effect, that no party can win this election in the first round. Who is right? The polls, without question. Though Mutahi’s argument has gained traction in certain circles, it is wrong-headed and the numbers on which it rests have been tortured to yield results they cannot sustain. This article explains the basic errors and spurious conclusions on which the “Tyranny of Numbers” stands. Put simply, the analysis is so flawed Mutahi should refund the money of the client that paid for it.

The “Tyranny of Numbers” makes a plausible but trivial claim, that is, that Kenya’s voting is historically influenced by ethnicity. This claim is plausible because identity is central to our national life- as it is in Belgium, Lebanon and any other multicultural country you may care to name.  It is trivial because no one has ever won an election in Kenya through exclusive ethnic votes without mobilizing at least three ethnic groups and so undercutting the argument that people vote ethnicity.

On the basis of that claim, Ngunyi then uses electoral demographics to build what, on examination, is a house of cards. The Jubilee Alliance, he says, begins with such a large numerical advantage the effort needed to secure an electoral victory is infinitesimally small. Barring a major catastrophe- including an assassination- Jubilee should win the presidential election in the first round. This inevitable victory comes from the Jubilee Coalition’s “bankable” ethnic vote of 6.2 Million (or 43.2% of the total vote). This number is basically a totting up of the registered GEMA and the Kalenjin voters. On that same ethnic logic, Mutahi reckons that CORD Coalition starts off with 19.2% of the vote or 2.74 million votes. For CORD to win, he says, they need to double their support. Or as he puts it, for CORD to catch up with Jubilee, it must multiply each of its Kamba and Luo vote by 2.3. Virtually impossible is the implication. Well? This is numerology not numeracy. Here is why.

The “Tyranny of Numbers” is built on five easily exposed errors and fallacies.

First, to say that Kenyans vote along ethnic lines does not lead to the conclusion that they vote for only their tribesman. Kenyans may mobilize along ethnic lines, a fact probably driven by weak political parties, but ethnic votes can and have historically split or moved tactically to politicians from other ethnic groups. In fact the only vote that has historically not moved is the Kikuyu vote. Mutahi analysis assumes that what is true of the Kikuyu is true of all ethnic groups in Kenya.

Secondly, even though the ethnic numbers seem plausible on the face of it scientific surveys, such as opinion polls, have consistently shown that Kenyans won’t always vote the way Mutahi predicts. And that in fact, the CORD Coalition has a much larger base – nearly double- the Kamba/Luo ethnic core Mutahi’s calls their “bankable vote.” Contrary to everything Mutahi says, the opinion polls suggest, as they did in 2007, a rather close election that will most probably be decided in a run-off. In fact the polls suggest that the so-called Jubilee “”” head-start does not exist. By nomination day, CORD had already made up ground on Jubilee by increasing its core support way beyond its ethnic base of Luo and Kamba whereas the Jubilee leaders had not. In effect, CORD has no catching up to do.

Thirdly, Mutahi misses the effect of incumbency on elections. In 1992 and 1997 and 2007, the leading candidates for president, Daniel arap Moi and Mwai Kibaki, ran as incumbents. So is Raila in 2013.

Fourthly, there are almost 3.5 million newly registered voters who nobody knows as yet how they will vote.  It is not possible to project historical voting patterns on this group.

Finally, Ngunyi’s voter-turnout figures for the GEMA block are not plausible. The turnout blind spot leads him to grossly overstate the overall GEMA numbers.


These five errors and fallacies mean that Mutahi’s conclusions are not just implausible, they are plainly mistaken and probably propagandist. We consider each of them in turn and provide evidence that refutes every one of Mutahi Ngunyi’s claims.

Proposition 1:
Ethnic groups do not always vote for their own (Even when there is a viable ethnic candidate)

Leaving the fact of incumbency, which is discussed below, past presidential elections show that any candidate who relies on a narrow ethnic base, two communities, for instance, is unlikely to win no matter how substantial their ethnic base. More critically, except for the Kikuyu, an analysis of the elections that have taken place in Kenya since 1992 shows that all other communities have done one or both of the following: split their votes between different candidates from other ethnic groups or voted as a group for a candidate from another ethnic group. This is tactical voting not ethnic voting. It is based not on loyalty but on calculation.

Let us look at the election results since 1992 to see whether this conclusion holds up. Below is Table 1, with election results, expressed in percentage terms for the leading candidates in that presidential contest.

Table 1: The 1992 Presidential election result by province, in %age

Province Moi % Matiba % Kibaki % Odinga % TOTAL  
Nairobi 16.6 44.1 18.6 20.2 373,548
Central 2.1 60.1 36.1 1.0 1,026,952
Eastern 36.8 10.2 50.5 1.7 782,800
North Eastern 78.1 10.1 4.5 7.1 73,394
Coast 64.1 11.4 7.6 16.1 310,476
Rift Valley 67.8 18.7 7.6 5.7 1,463,898
Western 40.9 36.3 3.6 17.9 524,200
Nyanza 14.4 3.3 6.4 74.7 800,678
Actual vote 1,956,866 1,404,266 1,050,617 944,197 5,398,037

Source: Constructed from defunct Electoral Commission of Kenya, ECK, results

To win a presidential election under the rules then in place, a candidate had to win a majority of the votes cast plus a regional threshold of 25% of the vote in five of Kenya’s eight provinces. If no candidate met this requirement, a second round election between the two front-runners would be held.

The 1992 results show that 96.1% of the vote in Central Province, an overwhelming majority of it Kikuyu, went to the two Kikuyu candidates, Mwai Kibaki and Kenneth Matiba. But the two candidates did not do as well outside of Nairobi and Eastern Provinces, both of them GEMA strong-holds. Matiba made 25% in Nairobi, Central and Western, but he also made 25% in Western Province because Martin Shikuku partnered with him. Kibaki made 25% in Central and Eastern. But Eastern, mainly Ukambani, gave Moi 36.8%.

When we drill down, some interesting things come up. Even though the Luhya’s may have mobilized as a community, they nevertheless split their vote, giving some to President Moi and some to Matiba. This probably reflects internal communal disagreements about who would be best placed to protect their interests between Moi and Matiba, exactly the kind of argument we would expect to hear in tactical voting. It may be argued in rebuttal that the Luhyas would have been no different from the Kikuyus if they had a presidential candidate from their own community.

Well, in 1997 they did and their voting that year clearly shows tactical rather than ethnic voting at work – see election results in Table 2 below. These results also show split ethnic voting in Eastern and Nyanza.

Table 2: The 1997 Presidential Elections by province, % age

Province  Moi % Kibaki % Raila % Wamalwa% Ngilu % Turnout
Central 5.6% 89.4% 0.7% 0.3% 3.1% 74.1%
69.5% 21.0% 2.2% 6.2% 0.7% 75.9%
Western 44.9% 1.4% 1.9% 48.2% 0.5% 68.1%
Nyanza 23.6% 15.1% 56.8% 1.6% 1.7% 67.2%
Eastern 35.6% 28.5% 0.7% 0.7% 33.6% 72.6%
Nairobi 20.6% 43.9% 16.3% 6.8% 10.9% 50.2%
Coast 63.4% 12.8% 6.1% 2.8% 9.4% 50.6%
73.2% 21.2% 0.3% 4.6% 0.5% 55.9%
Total  40.6% 31.0% 10.8% 8.2% 7.9% 68.2%

Source: Constructed from defunct Electoral Commission of Kenya, ECK, results

The results of the 1997 election do not support Mutahi Ngunyi’s conclusion that ethnic votes follow ethnic candidates. Though the Luhya community had a presidential candidate, Kijana Wamalwa, they split their vote between Wamalwa and President Moi, clearly a case of tactical voting. That fact also underlines the fallacy of the “Tyranny of Numbers” assumption that in 2013 the Luhya vote is Musalia Mudavadi’s.  President Moi also won in Eastern Province, where it had been assumed, on ethnic voting projections, that the Meru vote would go to Mwai Kibaki, the GEMA candidate, and the Kamba vote to the Kamba presidential candidate, Charity Kaluki Ngilu.

Indeed, based on exactly the assumptions that Mutahi makes in his analysis, the opposition failed to agree on one candidate to contest against President Moi on the logic that they could force a run-off by using their regional ethnic votes to stop Moi from satisfying the 25% in five provinces rule. But this logic collapsed with Moi comfortably winning more than 25% in Western and Eastern Provinces.

Once again, the only community that voted almost exclusively for their own candidate was the Kikuyu. Kibaki got 89.4% of the vote in Central. It is clear that the Luos also voted by a very large majority for their own, Raila Odinga. But the fact that both Kibaki and Moi- who are neither Luo nor Kisii, the two communities resident in Nyanza – got as many votes as they did suggests more diversified voting choices among the Luo and Kisii than among the Kikuyu.

However, and as the results of the 2002 election show, the Luo vote can move to a non-Luo candidate even in an election in which there is a credible Luo candidate. In 2002 they would ignore James Orengo, one of the leading lights of the pro-democracy, to move with Raila Odinga to support Mwai Kibaki. This is another example of tactical voting and not merely emotional ethnic voting. The 2002 election shows that the opposition leaders had recognized and their supporters had accepted that splitting the vote would hand an electoral victory to President Moi’s candidate, Uhuru Kenyatta.

To round off the analysis of 1997 results, let us look at the outcome in the Coast Province. In this multi-ethnic province, nearly 2/3 (63.4%) of the vote went to President Daniel arap Moi, a Tugen. This means that the Coast Province actually voted as if they were one ethnic group even though it is one of Kenya’s most ethnically diverse regions.

Table 3: The 2002 Presidential Elections results by province (as %age)

Province  Kibaki %  Kenyatta %  Nyachae %  Orengo % Turnout
Central 68.9% 30.2% 0.4% 0.1% 66.1%
Coast 62.7% 33.4% 3.2% 0.4% 42.1%
Eastern 72.5% 26.1% 0.8% 0.3% 60.9%
Nairobi 76.4% 20.8% 2.4% 0.2% 42.0%
North Eastern 28.1% 67.1% 4.6% 0.2% 57.8%
Nyanza 61.3% 7.6% 29.7% 1.1% 55.6%
Rift Valley 43.2% 53.3% 3.1% 0.3% 60.8%
Western 76.3% 21.5% 1.4% 0.5% 57.1%

Source: Constructed from defunct Electoral Commission of Kenya, ECK, results

In the 2002 election, there were in effect two main Kikuyu candidates both on the government side- Uhuru Kenyatta- and on the opposition side- Mwai Kibaki. This means that the Kikuyu question was irrelevant: the election would have to be won outside Central Province. Indeed, all previous elections have actually been won outside Central Province. The results suggest that there is no two-tribe alliance between the Kikuyu and one other large tribe that could win an election in the first round, whether under the old constitution- a majority of the popular vote plus 25% of the votes in five provinces or even under the new Constitution’s requirement of an absolute majority plus 25% in 24 counties.

In all past elections, Kikuyu candidates – not supported by at least two or three non-GEMA leaders- have not generally done well outside of the core GEMA electoral areas. Of course this could also be said to be true of other candidates, except that no other ethnic based candidate begins with the same numerical advantage as a Kikuyu-GEMA candidate does. In other words, there is no basis for Mutahi’s assumption that because Uhuru’s core block has so many votes it necessarily follows that there will he will be able to easily mobilize the additional support he needs from other parts of Kenya to win the election in the first round.

Some background will help fix this point. In 1992 the two Kikuyu candidates, Matiba and Kibaki only managed to get 25% or more in two and three provinces respecitively: Kibaki in Central and Eastern and Matiba in Central, Nairobi and Western (because of the Shikuku factor). Under the old Constitution, a combined platform would have had more votes than Moi but that does not mean that it would have won the election in the first round. It may be argued that the combined Matiba and Kibaki vote in the Rift Valley would have made the 25% required in that province. But, given that Kibaki failed to make the 25% vote in the Rift Valley in 1997 when he was the only GEMA candidate, it is possible that some of the votes that Matiba got in the Rift Valley in 1992 were not transferable to Kibaki and vice versa. In other words, there is nothing to suggest that if Kibaki and Matiba had joined forces they would have won the election in the first round without one more ethnic block.

In fact in 1997 when Kibaki was the only serious GEMA candidate he got 31% of the total vote but managed to make 25% in only 3 provinces:  Central, Eastern and Nairobi, core GEMA areas. It is only in 2002, with the support of Raila, Mudavadi and Kalonzo that he got more than 25% in all eight provinces.

No GEMA Candidate has ever made more than 25% outside of GEMA strong-holds without substantial support of at least two regional heavy-weights. In fact, in 1992 Kibaki got less than 10% in all non-GEMA provinces. In contrast, Oginga Odinga who started out with a core vote of less than half of Kibaki’s core vote (if we believe Mutahi) scored more than 15% in three non-Luo provinces: Nairobi (20.2%); Western (17.9%) and Coast (16.1%). Though Matiba scored 18.7% in the Rift Valley, a reflection of the Kikuyu diaspora vote, he scored less than 11% in all other non-GEMA areas except Western.

Uhuru is the only Kikuyu candidate who has ever done well outside of GEMA core areas- in 2002 – but then he was not seen as a GEMA candidate in that election. This fact was clearly manifest in the fact that he performed very poorly in all GEMA areas: he trailed Kibaki by a ratio of 3:1 in Nairobi and Eastern provinces and by a ratio of 2:1 in Central Province.

But, if GEMA candidates do not perform well outside of GEMA cores areas, how does one explain Uhuru’s seemingly strong performance in precisely those areas in 2002? Unfortunately, that good performance outside of the GEMA core does not look as impressive analysed closely.

Consider: in percentage terms, Uhuru got 10.4% fewer votes than Moi had in 1997 (see Table 4 below). What the data shows is that even with President Moi’s support, Uhuru lost very substantial grounds in all non-GEMA core areas that Moi had won in 1997. This means that his relatively modest 10.4% overall loss is thanks to the high population numbers in Central Province. His only gain outside of core GEMA areas was in North-Eastern, a region that had never voted against incumbency since the advent of multi-party democracy. The inability of Moi to transfer his 1997 vote- especially in the Coast, Eastern, Rift Valley and Western- confirms the general poor performance of GEMA candidates outside of their core areas. The question then is: 11 years on, how will Uhuru perform outside of core GEMA areas seeing what ground he lost even with Moi’s support in 2002? Is the Kalenjin gain from his Ruto partnership the only gain he is likely to make?

Table 4: How Uhuru Lost the Moi Vote

Province  Uhuru’s % age  vote in 2002 by Province Moi’s %age vote in 1997 by Province Change between 1997-2002
Central 30.2% 5.6% +24.6%
Coast 33.4% 69.5% -36.1%
Eastern 26.1% 44.9% -18%
Nairobi 20.8% 23.6% -2.8%
North Eastern 67.1% 35.6% +31.5%
Nyanza 7.6% 20.6% -13%
Rift Valley 53.3% 63.4% -10.1%
Western 21.5% 73.2% -51.7%
Total 30.2% 40.6% -10.4%

Source: Constructed from defunct Electoral Commission of Kenya, ECK, results

The Candidate Counts: Support for Moi in 1997 vs. Support for Uhuru in 2002

These numbers are not encouraging for GEMA candidates. It could be argued that the 2002 election changed the dynamics. That what was true before then is no longer so. It may be said that both the 2002 and the 2007 show the potential of GEMA candidates to do well outside their own core regions. This would be a plausible basis for Mutahi’s argument that with Uhuru and Ruto starting off with such a large base, getting the additional one million or so votes needed to beat CORD in the first round is a no-brainer.

However, before making that conclusion it is important to realize what happened in 2002. To begin with, contrary to received wisdom, there was no pro-opposition wave in 2002, if a wave is understood as an efflorescence of new support. The most instructive thing about the 2002 presidential result is how closely it tracks the 1997 presidential results. In 1997 Kibaki had 31% of the vote. Raila, Wamalwa and Ngilu had a combined total of 28%. In other words, the combined Kibaki, Raila, Wamalwa and Ngilu vote in 1997 was 58%. Compare that to Kibaki’s results in the 2002 election. The total vote for Kibaki in 2002- garnered with the support of Raila, Ngilu and Wamalwa – was 61.3%. In other words, the total Kibaki vote for 2002 in percentage is only 3.3% more than the combined total percentage of his votes in 1997 and those of Raila, Wamalwa and Ngilu. It is the combining of the 1997 opposition result, not the explosion of a new pro-opposition wave that explains what happened in 2002.

What do these numbers tell us? They tell us, even if somewhat tentatively, that in a three or four horse race GEMA candidates have tended to have very concentrated regional support from their core areas and not much support outside those areas. There is therefore no basis for assuming that once a GEMA candidate locks in the GEMA/Kalenjin vote of 6.2 (as calculated by Mutahi) then the rest of the 5.1 million votes (removing the 2.7 that he has allotted to CORD) will necessarily be a walk-over for Uhuru/Ruto.

Let us now consider the 2007 election. That election tells us two additional things which are relevant to how we analyze the way things will pan out in this election.

First, the fact is that there is now an in-built tendency in our electoral system- with the new electoral thresholds – to move Kenyan politics towards a two party system. In such a system, swing votes such as the Kambas would have been in 2007 and Luhyas will be in 2013 are critical.

Second, that election also tells us that starting from a high support base – as Mutahi claims that Uhuru and Ruto have- does not translate automatically into a huge electoral advantage over those who start from a low support base- as he says that Raila and Kalonzo have.

The 2002 election showed that it was the fragmentation of effort that had cost the opposition the previous two elections, not their poor performance or Kanu’s good performance. All sides recognized that again in 2007 and that race became, in effect, a two horse race with the Kamba as a potential swing constituency which neither Kibaki nor Raila could take advantage of to decisively win the election. The 2013 election is also a two-horse race with the Luhya this time as the swing constituency.

A little probing of the 2007 numbers is revealing. If we assume that throughout 2002 to 2007, Kibaki had not lost the 31% electoral base that he brought into NARC in 2002, then if the opinion poll figures for October 2006, which gave him a 41% support base are correct, incumbency had given Kibaki only a 10% gain. Over the next 14 months – from October 2006 to December 2007 – Kibaki would increase his support by a marginal 2%. Is this yet another case of the inelasticity of the GEMA candidates votes outside of their base?

That question brings us back to the second thing we learnt from the 2007 election: high starting thresholds guarantee nothing in electoral terms: the polls gave Kibaki a support base of 41% in October 2006. This was a seemingly insurmountable 28% lead over the 13% support base that the polls gave to Raila. Yet in 14 months Raila Odinga had raised that support to 45% according to the December 18, 2007 polls (see the trend in polls from October 2006 to December 18th 2007).

The conclusion to draw is, once again, the opposite of the one Mutahi has drawn from his fancy math: the numbers one starts with do not make it easier to mobilize the additional votes required to win an election. If they did, the 2007 election should have been a landslide for Kibaki given where he had started in October 2006.

Table 5: The 2007 Presidential Election by Province

Province Kibaki % Odinga % Kalonzo %  Registered Voters


Nairobi 47.7% 44.0% 8.1% 1,275,021 51.5%
Coast 33.1% 59.4% 6.5% 1,045,629 57.0%
North Eastern 50.3% 47.2% 2.3% 315,664 61.3%
Eastern 50.4% 5.0% 43.8% 2,516,998 65.9%
Central 97.0% 1.9% 0.7% 2,186,315 82.1%
Rift Valley 33.5% 64.6% 1.4% 3,358,285 72.8%
Western 32.2% 65.9% 0.7% 1,564,854 62.0%
Nyanza 16.9% 82.4% 0.3% 2,041,686 76.2%
Totals 46.4% 44.1% 8.9% 14,304,452 69.0%

Source: Constructed from defunct Electoral Commission of Kenya, ECK, results

Past elections then give us two conclusions that should unsettle Mutahi’s smug conclusions.

First, all ethnic votes are movable to non-ethnic leaders. The only vote that so far has refused to move, even if partially, to non-ethnic candidates in all four multiparty elections is the Kikuyu vote. So the assumption of an effectively bound ethnic vote for each of the leading candidates is false, a fact clearly underlined by the fact that in the polls CORD is actually leading Musalia Mudavadi in Western Province.

Secondly, the fact that Uhuru and Ruto appear to start off with a large so called “bankable vote” does not make it any easier for them to mobilize the additional “small vote” they need to win the election in the first round. If this were a logical necessity, then Kibaki who started off with 41% support in October 2006 polls should have had an easy time mobilizing the 9% or so additional support he needed to guarantee himself 50% of the national vote in 2007. Instead, we had a situation where Raila, starting off with 13% of the vote in 2006, increased his support by 32% in 14 months to catch up with and surpass Kibaki in the polls and force Kenya’s closest election.

More subtly, in spite of their high starting numbers, GEMA candidates running outside a three or four way alliance have not, historically, done well outside their core areas. Does Mutahi sense this and is he trying to wrong-foot debate about what could happen in the election and the widely predicted run-off by planting red herrings all over cyber-space?

The core of the “Tyranny of Numbers” is thus easily shown to be a spurious massaging of the numbers. But there are other equally strong objections to Mutahi’s numbers. Let us now turn to the opinion polls.

Proposition 2:
Opinion poll numbers are bankable; Mutahi’s numbers are not.

Since Opinion polling was first introduced in Kenya in 1992, polls have generally called elections right. The early polls in the early 1990s, as with all new things, were bedeviled by methodological challenges, political manipulation and a general culture of distrust based on misunderstanding how sampling and surveys work. But since 2002 opinion polls have generally been within the eventual results.  As a review of opinion polls in 2010 concluded the 2002 “pre-election and exit polls” “reasonably and accurately” predicted the outcome of the general election. The 2002 polls predicted victory for Kibaki with a 62% vote, gauged Uhuru’s support at 31% and Simeon Nyachae’s at 5%.

If the polls are generally on the money, how can they help us see how the 2013 election will turnout? To answer that question and to see why the “Tyranny of Numbers” assumptions are spurious, it would help to review the polls conducted from October 2006 to the election in 2007. Those polls are important for two reasons: First, they show us, as already noted in the preceding section, that the support that one begins with (the GEMA/Kalenjin numbers, for instance) have no relationship with where one ends up. Secondly, they also show that polls are able to capture the evolution in the political landscape quite accurately.

So, what did the polls say in 2007?
Table 6: Summary of Key Presidential Candidates’ Polls from 2006-2007

Pollster/ Date of Poll Kibaki Musyoka Odinga Mudavadi Ruto
Steadman, Oct. 2006 41% 20% 13% 3%
Steadman, Dec, 2006 42% 20% 14% 3%
Steadman, March 2007 51% 14% 17% 2% 2%
International Republican Institute, April 2007 44.3% 15.3% 18.7% 2.7% 2.6%
Research and Marketing Services, June 2007 45% 14% 28% 4% 3%
Steadman, July 2007 45% 11% 25% 3% 2%
Infotrak Research and Consulting and Harris Interactive Global, August 2007 42% 11% 25% 8% 6%
Steadman, August, 2007 47% 13% 36% 1%
Steadman, Sept. 2007 38% 8% 47%
Steadman, Oct 13th, 2007 37% 8% 53%
Steadman, Oct, 23rd 2007 39% 8% 50%
Steadman, Nov. 9, 2007 41% 11% 45%
Consumer Insight, Nov. 21, 2007 41.4% 14.7% 40.7%
Gallup, Nov. 17, 2007 42% 11% 45%
Steadman, November 23, 2007 43.3% 11.4% 43.6%
Steadman, Dec. 7, 2007 42% 10% 46%
Steadman, Dec. 18, 2007 43% 10% 45%

Source: Wikipedia; Also Odera Kiage and Kwame Owino

In polls conducted in October and December 2006, a full year before the election, Raila’s support averaged 13.5% whilst Kibaki’s was at 41.5%. In the early polls conducted in 2007, in March and April, Raila’s average support rose to about 17.3% whilst Kibaki’s rose to 47.3%. Instructively, Kalonzo’s support, which had averaged 20% in October and December, 2006 fell to 14% during this period.

This would seem to suggest that both Kibaki and Raila ate into Kalonzo’s support base. In polls done in June, July, August of 2007 Raila had pushed his electoral support to 26%. Kibaki’s had declined to 44% and Kalonzo’s had fallen further to 12%. This is significant because it suggests that Raila had nearly doubled his support by eating into both Kibaki’s and Kalonzo’s support bases. More significant, he had done so before joining up with Musalia, (average 6% for the period) and Ruto (averaging about 4% for the period) in the Pentagon. Since the Luo vote is only 12% of the total vote in Kenya the doubling of Raila’s support means that he had an additional 14% support beyond the Luo ethnic core which Mutahi Ngunyi says is the primary determinant of voting in Kenya.

In the nine polls reviewed here, mainly by Steadman International (now IPSOs Synovate), and one by Gallup and Consumer Insight) conducted from September to December, 2007 during which time the Pentagon was in full flow, Raila led Kibaki in all but one poll, the Consumer Insight poll of November 21st 2007. Raila’s support averaging 46.1% and Kibaki’s 41%. However, the averages for the period are a little misleading because the last three polls in November and December were, in effect, a statistical tie. The last poll before the elections – conducted on December 18, 2007 by Steadman- put Kibaki’s support at 43% and Raila’s at 45%, a tie given a margin of error of plus/minus 3.

For those who doubt the science of polling, it would help to look at the three polls conducted by three different organizations on November 17 (Gallup International); November 21st (Consumer Insight) and November 23 (Steadman).  Consumer Insight gave Kibaki a 0.7% lead over Raila, Gallup gave Raila a lead of 3% and Steadman gave Raila a lead of 0.3%. In short, a tie.

When the election was finally held, it was unclear who had won it. In fact, when Justice Kriegler led a commission of inquiry into the management and conduct of the 2007 election, he concluded that the Commission could not say who had won the election.

Given that background, let us review what we have learnt from opinion polls so far and whether they offer a more plausible set of numbers to answer the “Tyranny of Numbers” (Table 7 below, summarizes the recent polls).

Table 7: Opinion Poll Results since December 2011

Pollster/ Date of Poll Raila Uhuru Ruto Kalonzo Musalia
Ipsos Synovate, Dec. 2011 32% 22% 16% 10% 2%
Ipsos Synovate, April 2012 34% 22% 13% 8% 5%
Ipsos Synovate, July 2012 33% 23% 9% 12% 7%
Ispos Synovate, Sept 2012 36% 30% 6% 5% 7%
Ipsos Synovate, Nov 2012 33% 26% 9% 8% 4%
Ispos Synovate, Jan 2013 46% 40% 5%
Strategic Research, Feb 2013 44.4% 43.9% 6.4%
Consumer Insight, Feb 2013 45% 43% 5%
Infotrak Research and Consulting, Feb 2013 45.9% 44.4% 6%
Poll of polls, (Infotrak, CI and SR) 45.1% 43.7% 5.8%

Before Raila joined up with Kalonzo, his support was steady above 32%. In fact, before December 2011 it had touched a peak of 48% (in October 2010). In the four polls done by Ipsos Synovate in 2010 Raila’s support averaged 40.5%. This means that throughout 2010, Raila had 28% support over and above the 12% Luo core that Mutahi Ngunyi makes so much of. Though that support base shrunk to an average of 34% in the five polls conducted in 2011, that number was still 22% above the Luo support base that Mutahi calls Raila’s “bankable vote.” That means that throughout 2011, Raila had nearly twice his so-called bankable vote outside his home turf.

According to recent polls, that support has increased since the nominations were conducted in December 2012. Since then Raila’s support has been consistently above 40% and in the poll of polls combining the three opinion polls done between February 14-17, the Raila-Kalonzo ticket has 45.1% of the vote and the Uhuru-Ruto ticket has 43.7%, a statistical tie.

However, when we do the ethnic maths that Mutahi asks us to do, the combined support for Raila and Kalonzo should be no more than 19.2%. If this is the correct ethnic figure, then if the polls are right, Raila and Kalonzo have at least 25.9% support outside of their ethnic core. In other words, the Raila/Kalonzo group has more support outside of their ethnic core than their combined possible ethnic vote. In contrast, the Uhuru-Ruto ticket has 43.7percent, only 0.5% above their combined ethnic vote of 43.2%.

Now let us do the maths that apparently shows that the election was won on December 18th 2012 when the voter registration exercise closed. According to the narrative of the “Tyranny of Numbers”, the combined GEMA/Kalenjin vote, assuming 100% voter turnout, is 6.2 million votes or 43.2% of the total vote of 14,337,339. And CORD has only 19.2% or 2.74 million votes.

The fallacy here is easily revealed by the polls. On the CORD side, Mutahi assumes, correctly, that Raila and Kalonzo have the total support of their communities- the Luo and the Kamba- but concludes, wrongly, that that support – 19.2% of the vote – is the full extent of their support base. On the Jubilee side, he makes the opposite error. He assumes, once again correctly that Uhuru and Ruto have the total support of their community but concludes, once again wrongly, that they have more votes outside their core support that they can mobilize.

If the opinion polls are right, then there really is no real gap between CORD and Jubilee in terms of the support base. In fact, the polls seem to suggest that the total Jubilee support is entirely drawn from their ethnic base. This ought to be a matter of concern to Jubilee. Consider this a little closely. Until November last year, polls had given Uhuru a support base of about 23%. This was before he joined up with Ruto. That number is very close to the total Kikuyu vote, reckoned to be between 21%-22%. It is instructive that for a long while Uhuru’s support base was below his ethnic block vote, indicating that for a long time he did not even have the support of his own community. It is only after Uhuru joined up with Ruto that his poll numbers approached 40% and now 43.7. That number, as already pointed out is extremely close to the outer limit of the total GEMA/Kalenjin vote, 43.2% of the vote on Mutahi’s calculations.

If Kenyans are voting primarily along ethnic lines, as Mutahi says they are, then poll numbers are suggesting that Uhuru/Ruto combination basically have no support outside their ethnic base of Uhuru’s GEMA region and Ruto’s Central Rift Kalenjin core. The conclusion to be drawn from the demographics and the polls is therefore the opposite of the one Mutahi draws: The Uhuru/Ruto core support is ethnically concentrated but the Raila/Kalonzo core support is ethnically diverse. Using the banking metaphor so beloved of Mutahi, the vote that Uhuru/Ruto can take to the bank closely matches the numbers of their ethnic base whereas the vote that Raila/Kalonzo can bank is larger than their ethnic core. There is just no other way of explaining the poll numbers against the demographics.

Proposition 3:

Incumbency Matters
So far, the argument in this article has tried to show that the effect of ethnicity is over-estimated in Mutahi’s analysis. But as important is Mutahi’s blind spot with regard to the effect of incumbency on electoral outcomes.  One of the missing explanations for the Moi victory in 1992 and 1997 as well as Kibaki’s performance in 2007 is the fact of incumbency.

President Moi did not run just as a Tugen, he ran as the President of Kenya. In 2007 Kibaki ran as the President of Kenya, not just as a GEMA leader. This means that one of the explanations for the geographical spread of his support base in that election is the fact that he was the sitting president. Similarly, in 2013, Raila is not running as a Luo leader, he is running as the Prime Minister of Kenya. The analysis from the 1992, 1997 and 2007 presidential elections show that ‘incumbency’ is a factor.  If Mutahi’s ethnic logic is correct, then Moi should have got only the Kalenjin and Kamatusa votes in 1992 and 1997. But he was the incumbent. He was not presented as a Kalenjin candidate but as the President of Kenya – hence the number of votes (widely spread in geographical coverage) he got in both elections. Similarly, one possible explanation for Kibaki’ ability to get votes outside the GEMA constituency unlike other GEMA candidates before is the fact that in 2007 he ran not as a Kikuyu candidate but as the President. Mutahi’s argument stretched to its logical limit means that Kibaki should have only got votes from the GEMA communities.

The truth is that Raila is now campaigning as the incumbent prime minister. The fact that the 2007 election was inconclusive means that, in effect, we have a re-run of that election, both at the personal, psychological level and at the ethnic level. Raila has, in truth, remobilized his support base and the fact that the poll numbers he has now look remarkably similar to his poll numbers in 2007 is not an accident.

Proposition 4:
Nobody can presume to know how the 3.5 million new voters will cast their ballots

There is much play with voter registration numbers and, by implication, prophesying on the impact of high voter turn-out. It is not easy to say what the voter turnout will be but the historical trends are clear and there is nothing to make us think that this election will be any different. One large unknown, which has not been factored into the “Tyranny of Numbers” analysis, is how the newly registered voters, thought to be about 3.5 million, will actually vote. As we saw in the United States in the election last year, changes in electoral demographics can have a very unsettling effect on presumed voting patterns. One signal that there may well be surprises is the growing influence of Waititu and Sonko. These two are not ethnic leaders, contrary to what pundits might like to think. They are class candidates representing the underclass and enjoying support across ethnic lines. How mobilized is the under-class in this election? No one can tell until the results are in.

Proposition 5
Turn-out Trends since 1997 show that Mutahi’s turnout figures are bogus

A casual look at turnout figures for the last three elections-2007, 2002 and 1997- indicates that Mutahi’s turnout numbers and the extrapolations he makes from them are totally bogus. Excluding the referenda, the turnout figures clearly show very significant regional variations and that the 100% turnout assumption for Jubilee strongholds is just not tenable. As the table below indicates, national turnout figures for the last three elections were 69% for the 2007 election; 55.3% for the 2002 election and 68.2% for the 1997 election. In terms of Mutahi’s numbers the variation in turnout between Central Province and Nairobi, whose 40% Kikuyu vote he gives in total to Jubilee, was extremely high: 29.7 in 2007; 24.1 in 2002 and 23.9 in 1997, averaging a gap of 25.9. There is no basis for an assumption that this years’ will election will any different.

Table 8: Voter registration and turn-out since 1997

Province 2007 Presidential Results 2007 Presidential Results 2002 Presidential Results 2002 Presidential Results 1997 Presidential Election 1997 Presidential Election Av. % turnout 2007-1997
Registered  Turnout Registered Turnout Registered Turnout
Nairobi 1,275,021 51.5% 884,135 42.0% 726,779 50.2% 47.9%
Coast 1,045,629 57.0% 879,741 42.1% 800,689 50.6% 51.7%
North Eastern 315,664 61.3% 216,336 57.8% 172,337 55.9% 58.3%
Eastern 2,516,998 65.9% 1,734,209 60.9% 1,433,737 72.6% 66.5%
Central 2,186,315 82.1% 1,563,084 66.1% 1,346,189 74.1% 74.1%
Rift Valley 3,358,285 72.8% 2,415,555 60.8% 2,160,453 75.9% 69.8%
Western 1,564,854 62.0% 1,202,104 57.1% 1,028,732 68.1% 62.4%
Nyanza 2,041,686 76.2% 1,555,986 55.6% 1,361,251 67.2% 66.3%
National 14,304,452 69.0% 10,451,150 55.3% 9,030,167 68.2% 64.2%

Source: Constructed from the defunct Electoral Commission of Kenya, ECK, data, various.

Suppose that turnout remains true to historic trends? If we use the average turnout rate of the last three elections as the basic turnout rate then we can work out, as the table below does, more plausible scenarios of how the election might pan out. We construct three scenarios. First, we average the two most recent polls. Second we construct one scenario where we assume that the candidates will get the votes that the polls give them. In other words we assume that the polls are correct. Thirdly we impugn the polls and assume that they are systematically off by 6% either in favour of or against either candidate. If we allocate votes to both Raila and Uhuru based on average regional turnout figures for the last three elections, the results of this election will be as set out in the table below.

The first thing one notes is that this allocation gives us a more plausible voter turnout figure of 9,183,458. On this turnout, the winning majority is 4,591,730 (50%+1). We assume that if each candidate gets the regional distribution of votes that the opinion poll indicates that they will get, they are also likely to make the 25% in 24 counties threshold. According to these three scenarios that emerge for each candidate, there is no case in which Uhuru-Ruto Jubilee Alliance can win the election in the first round. Not even if we assume a worst scenario for Raila (that the opinion polls have systematically overstated his support by 6% across the board) and a best-case scenario for Uhuru (that the opinion polls have systematically understated his support by 6% across the board).

On that assumption Uhuru would get 4,272,502 votes and Raila would get 3,807,073, a difference of 465,429. This would mean that though Uhuru could make the 24 counties threshold he couldn’t make the 50%+1 absolute majority vote needed to win in the first round.

On the other hand a first round win for Raila can only happen if we assume that the opinion polls have systematically understated Raila’s support by 6% across the board. If they have, then he would get 4,916,222.  If these assumptions are right, there is no scenario in which Uhuru can win the election in the first round and in only one highly optimistic scenario can Raila win it in the first round. It therefore does not seem plausible that either candidate can win this election in the first round, as Mutahi claims Jubilee will.

Table 9: Simulated Election results for 2013 based on historic turnout and poll data

Provinces Projected Turnout (Based on past turnout) Uhuru allocation (based on latest poll of polls) Uhuru allocation (based on latest poll of polls) Uhuru allocation (based on latest poll of polls) Raila allocation (based on latest poll of polls) Raila allocation (based on latest poll of polls) Raila allocation (based on latest poll of polls)
  If correct If off by 6%(-) If off by 6% (+) If correct If off by 6%(-) If off by 6% (+)
Nairobi 852,095 323,796 272,670 374,922 460,131 409,005 511,257
Central 1,620,953 1,410,229 1,312,971 1,507,486 129,676 32,419 226,933
Rift valley 2,361,697 1,369,784 1,228,082 1,511,486 755,743 614,041 897,445
Nyanza 1,301,867 104,149 26,037 156,224 1,145,642 1,067,531 1,223,755
Eastern 1,391,767 320,106 236,600 403,612 807,224 723,719 890,730
Western 895,431 40,294 0 89,431 519,349 465,624 573,076
Coast 601,831 126,384 90,275 162,494 403,227 367,117 439,336
North Eastern 202,567 54,693 42,539 66,847 141,797 127,617 153,950
Totals 9,183,458 3,949,435 3,209,174 4,272,502 4,368,789 3,807,073 4,916,222

Source: Various, turnout data, latest opinion polls.

The interesting thing is what would happen if Mudavadi threw his weight behind either of the candidates. If we discount how his regional vote is distributed and assume that he has, as the polls say, 5% of the national vote, he controls about 459,200 votes. There is only one scenario in which Uhuru could win a first round election even if he got all of Musalia’s votes, namely scenario 3 in which the polls have understated his support by 6% across the board. However, the only scenario under which Raila would not win the election in the first round with Musalia’s support is one in which we assume that the polls have systematically understated his support by 6% across the board.

In short, there is no plausible scenario in which one can see Mutahi’s “Tyranny of Numbers” yielding the results he claims.

It only remains to answer the argument that has now begun to filter through, namely that people do not tell the truth when they answer pollsters’ questions: the so-called Bradley effect. There are two responses to that.

First, the alternative scenarios constructed in Table 9 simulations above on the assumption that polls have been wrong by 6% (plus or minus) should eliminate most errors of that type.

Secondly, there is a more important question. Is there a basis for assuming that this time around, people have more systematically lied to pollsters than they did before when the polls accurately called the election?

It is clear that if we take a hard look at the real numbers rather than the wishful ones that the “Tyranny of Numbers” wants us to consider there really is nothing to Mutahi Ngunyi’s much-publicized piece of numerology. There is moreover something quite sinister and irresponsible in the narrative he has created out of these numbers. He calls the coming election a transition election- which is true- and notes portentously and alarmingly that such elections “have attracted assassinations” and notes, “how or whether these occurrences will manifest themselves is still unknown.”

This is arrant nonsense: the 1963 election attracted no assassination; the 1978 election that brought Moi to power after Kenyatta’s death attracted no assassination; 1992 election attracted no assassination and the 2002 election attracted no assassination. What transition election is Ngunyi referring to? Assassinations in Kenya have arisen in the context of internal party struggles between reformers and anti-reformers or between radicals and moderate and have only been tangentially connected with elections or transitions. Mutahi’s reckless posturing is a particularly odious case of scare mongering and it is not helpful in a tense and close election.

*Wachira Maina is a Constitutional Lawyer

By Mzee Posted in kenya

3 comments on “Kenyans dont always vote tribally

  1. I think this long winded post is trying to deny the obvious. We have been voting tribally since the advent of multi-partyism. It is important to accept the truth so that we can be in a position to find a solution. Mutahi made very broad assumptions. For example, he wrongly assumes that CORD is a Luo-Kamba affair. In the real sense, CORD is a 75%Luhya-Luo-Kamba-Coastal-Kisii-Turkana coalition as its primary building blocs. It also enjoys the support of majority Somali, Pokot and Samburu. It is likely to split the Maasai vote equally with Jubilee The GEMA Kalenjin will not vote CORD. Let us be honest here. Jubilee currently doing 2% in Western, 7-9% in Nyanza and 10-17% in Coast. It is correct to assume that these blocs will also vote CORD unanimously. The Primary support base for CORD is about 6.1 million strong as opposed to Jubilee’s 6.2m strong base. Evenly matched but turnout will take its toll but CORD can still count on the support of Mudavadi’s 25% Luhyas plus advantage in the Maasai/Somali swing vote. CORD has its own tyranny of numbers and it needs to bring out this broad based tyranny to counter Mutahi’s narrow two-tribe tyranny which is not accurate. Denying the obvious does not help.


    • LKay,

      I had an interesting exchange with one of those CORD heavy lifters and my theory to them was that, yes Kenyans do vote on tribal lines but it is just one of the many factors and in fact the only times we have achieved anything as a country it has been when we discarded tribalism.

      In 2002 in the Narc revolution Luos, Luhyas (even with their key leaders with Moi), Kikuyus etc all voted for Kibaki? Why? Because the country was fed up with Moi and his dictatorship. Mudavadi who was going to be Uhuru’s deputy could not even win a seat in his backyard. Some will say the Luos voted for Narc because of Raila but the reality is is Raila did not even have any leadership role in Narc.

      If we look at the dyanamics of 2002 one thing we see quite clearly is that at any one time the nation is faced by some issues that transend their tribes. At such time it is up to the leaders to read the mood of the nation. Those who don’t will fall on the wayside and that is what happened to Mudavadi in 2002.

      In this election there are some issues that may also transcend tribes. The issue of land for example is a big deal. This is the first election in which land is central to the election. Even the attempts to suppress the land debate have fallen on deaf years.

      People are sick of a few individuals owning the whole country. That is one issue that complicates Uhuru’s alliance with Ruto. It is the one reason that will guarantee CORD a whole ton of votes from the Kalenjin communities that would have otherwise gone to Jubilee. There is a reason CORD rallies in key kalenjin counties still attract very big and energized voters. It is that 20-30% vote from that constituency that people like Ngunyi just ignored while assuming tribe overtakes everything. It is also part of the reason a Mudavadi Ruto team with Uhuru’s blessing would have made Jubilee almost unbeatable. I still shake my head remembering how close these people were to running away with this thing and were blinded by political stupidity of legendary proportions.

      The other interesting point you have raised is that the Luhya vote has its own dyanamics and CORD found themselves very lucky with that. They still have their vote and they didn’t have to give Mudavadi anything. That gave room to bring Kalonzo in which is a big deal for CORD.

      But remember that never before has Raila had the support of the Bukusu part of Luhyaland. This time he has that vote. That is about 40% of the vote there. Raila still has great support in Mumias and Busia counties. Of course Ngunyi never figured that out. In effect CORD will probably get more votes from Western than ODM had in 2007. One big deal there is voter turn out. It was terrible in 2007 and Kibaki still got about 40% of the vote there. This time with all the competition for endless number of posts the turnout might be very good.

      And speaking of turnout and the strange idea that Jubilee voters will turn out in large numbers while CORD folks will be sleeping. We forget that for the first time all the 14.3 million voters just collected their voting cards in the last 3 months. Before only a small fraction of voters actually registered near the election date. Majority of the voters had voter cards obatined decades ago and some may be did not even know where their cards were. This is the first time we have this situation and it is going to have an impact in voter turnout.

      Then we have the Gussii factor with another 500,000 plus votes. That is in the CORD column.

      And back to tribalism and Kenyans always voting according to tribes we forget that the very first election held in our country was in 1963 and tribe was never a factor then.

      What forget that it is our leaders who always drag tribalism monster back to the house after Kenyans kick it out. Kenyatta did it in 1963 and Kibaki in 2002. This time we may just kill this tribal thing forever.


What Say You Now?

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s