Africa Matters is a blog that follows the news and offers analysis of African affairs. Our aim is to delve deeper into the issues of African politics and development. We don’t presume to be experts, and we don’t presume to have all the answers—we are just trying to ask the right questions.

Saturday, January 31

Revisiting Kenya's tarnished election

A year ago, Kenya was in turmoil—a bastion of stability in a region better known for violence and anarchy had suddenly cracked. On December 27, Election Day, presidential challenger Raila Odinga seemed to have a healthy lead over incumbent Mwai Kibaki by a few hundred thousand votes; three days later, after a suspicious delay, when the chair of the Electoral Commission emerged from the vote count, Kibaki was named the winner. Riots erupted in cities throughout the country, and, soon, anger and frustration turned into ethnic clashes, with hundreds dead and hundreds of thousands displaced. It took months to broker a power-sharing deal to end the crisis.

Now, a year later, Mike McIntire and Jeffrey Gettleman have a very interesting article in The New York Times that takes a look back at an under-reported thread in this story: an exit poll, commissioned by the U.S. government, which would have shown Odinga as the clear victor, had it not been suppressed until it was too late to make a difference.

The poll, conducted by the International Republican Institute, found Odinga the winner by six percentage points, a stark contrast to the official tally, where Kibaki emerged with a 1%, 230,000 vote victory. (IRI is an American organization funded, along with its sister organization the National Democratic Institute, by the National Endowment for Democracy; IRI has a Republican slant—John McCain is the chair of its board—while NDI leans Democratic.) The subtext of the article is that IRI, under pressure from U.S. Ambassador to Kenya Michael Ranneberger and members of its board, quashed the report. McIntire and Gettleman float the theory that Kibaki, a Bush ally and supporter of U.S. anti-terrorism activities in the region, was preferred to Odinga, who "was viewed skeptically by some in Washington because of his flamboyant manner and his background: he was educated in East Germany and named his son after Fidel Castro."

The Times lacks any hard evidence of foul play; the investigation relies on the testimony of a few key players in the affair, including a former senior IRI official, Kenneth Flottman, and political scientist Clark Gibson, who designed the poll—as well as the generally suspect nature of the whole ordeal. Sources in the story cite repeated meddling by Ranneberger, which indicates biased government interference from the start, as well as the seemingly specious nature of IRI's justification for withholding the results—that the poll was inaccurate.

IRI, for its part, has released a fiery response to the Times's article, defending the claim that there were serious questions about the poll's validity, opening with the tactful, "Once again a New York Times 'investigative' story, written like a sloppy high school term paper, has slipped by any editors worthy of the newspaper’s former reputation." The statement goes on to explain:
Immediate poll results were unavailable to IRI because post-election violence in Kenya prevented data reaching Nairobi for almost three weeks after the balloting. Once the results did reach Nairobi on January 17, 2008, IRI and James Long, a Kenya-based polling consultant with the political science department at the University of California at San Diego, attempted to resolve discrepancies in the data. As early as January 19, former IRI employee Ken Flottman – who in the Times story charges that IRI withheld the poll at the behest of U.S. government officials – stated in an email that “This is enough to tell me that we need to check against the actual questionnaires to ses [sic] what they do and don’t say.” On January 25 Flottman wrote “James [Long] agrees we should do an audit. We will set up the meeting and then I’ll just have him start auditing. Will let you know what we are finding.”
The most bizarre line in IRI's response, though, is this gem: "IRI had no political motivation to suppress the poll; if anything, having worked with defeated candidate Raila Odinga for more than 20 years, IRI had motivation to want Odinga to win." Wait, what? Countering an accusation of bias towards the winner by claiming bias towards the loser seems, well, half-baked. Further, IRI is proudly displaying this letter of support from Odinga himself, which, if you actually read it, is less than a ringing endorsement.

There's no way to know what effect an exit poll like this would have had on Kenya's crisis. On the one hand, exit polls are notoriously unreliable; on the other, this one would have carried the weight of an experienced election monitor with the backing of the U.S. government, and in those chaotic days last January, evidence like that might have been just enough to scuttle Kibaki's machinations. 

I find it hard to believe, though, as Odinga claims in his letter and as McIntire and Gettleman suggest in their article, that had this exit poll gone public, it may have saved people's lives. More likely, I think, it might have been enough to lift Odinga to an outright presidential victory, but it probably would have precipitated even more violence in the short term, since much of the mayhem was a result of Odinga supporters, who, frustrated by their disenfranchisement, lashed out at Kikuyus, Kibaki's tribe. This poll may have served only to aggravate their anger.

The larger question, of course, is what business the U.S. has interfering (yet again) in a foreign election? This poll may have had little impact in terms of subduing violence at the time, but it could have altered the course of Kenyan politics—the country may now be at peace, but the electoral crisis and subsequent political bargaining and constitutional maneuvering required to end it have done permanent damage to Kenya's democracy. In an interview with the Times's Celia Dugger, Seretse Khama Ian Khama, president of Botswana, had some thoughtful words on this subject, in the context of the similar scene playing out in Zimbabwe at the moment:
You know, these power-sharing agreements, it's not the way to go on the continent. You can't have a situation where a ruling party, when it senses it may lose an election, can then manipulate the outcome, um, so that they can stay on in power—and then agree to having a power-sharing agreement, in order to extend their stay in power, as what happened in Kenya and here in Zimbabwe. What we must ensure, is that we get our elections right, and make sure and ensure that they are held in a credible manner. The basic principle is that people should have the final say.
A dangerous precedent is being set. A family in California invests in their home for the long run; if an earthquake strikes, they don't tear the house down and rebuild it anew once the shaking ends, just to stop the dishes from rattling. The desire for short-term stability in Africa is winning out over the principles of democracy—we're reaching a point where the less legitimate an incumbent can render a vote, the more likely he is to hold on to power.


Clark said...

The NYT article and the IRI response did not present all of the useful facts about the exit poll. As one of its designers (along with James Long from UCSD and Strategic Research in Nairobi), I would like to present some relevant details.

First, this exit poll was not intended to call the results of the election. Exit polls are, after all, polls, and not the counting of real ballots. The intent of our exit poll was to conduct research that explores the determinants of Kenyan votes. We asked questions about voters’ backgrounds to see if they were associated with how they voted. We were especially interested in ethnic backgrounds. We had no intention of becoming involved in any controversy. Exit polls are a standard feature of elections around the world.

The best check on an election is a rigorous process under the supervision of the relevant authorities. Failing this, there can be a parallel vote count, which did not occur in Kenya. A third choice might be an exit poll -- certainly a well-run exit poll is far better than all of the forensic exercises that have been tried on the Kenyan vote. These rely entirely on the assumptions of the analyst; exit polls do not. There are clear and basic rules about conducting polls, which we followed. The dozens and dozens of exit polls conducted around the world – and the millions of dollars spent on them -- attest to their accepted usefulness.

I believe our exit poll was, outside of perhaps South Africa, the most rigorous ever staged on the continent. The actual results of the exit poll were Odinga 46.07%, Kibaki 40.17%, Kalonzo 10.22%. The margin of error was 1.32. Since IRI’s original reviewer did not review the actual 5,495 surveys, his comments are barely relevant; we had technical responses to all of them, and none imperil the results. We also have no idea how IRI came up with the "2% off" figure. They did not have any discussions with us during any of their "reviews." I doubt the accuracy of their reviews because of this, as they would need guidance about many features of the poll. They have not made public any of the methods they used in their reviews. We are happy to provide our procedures and methods to anyone. We are scientists, not activists.

We had a contract with IRI. IRI had a contract with USAID. Our job was to design and manage the poll, and to provide the results to IRI and AID. We did this, and very well in my opinion: the margin of error is extremely low, and our coverage of Kenya included 179 out of 210 constituencies.
We had no control over the publication of the results for six months. Now we are allowed to publish and discuss our findings, which was the intent of our participation in the first place.
I hope this information is useful.
Professor Clark Gibson
University of California at San Diego

Jackie IRI said...

Wilson Strategies, at the request of IRI, did in fact review more than 6,000 questionnaires through the re-entry of the data. After the data was re-entered it showed a 6 point spread between Odinga and Kibaki (46.8– 40.7%) as opposed to the original 8 point spread (48.5– 40.9%) reported on January 31, 2008.

IRI’s only interest in the Kenyan election and the exit poll was to ensure an open and transparent democratic process.

Anonymous said...


Who was reporting anything on January 31? To whom? Where were those results published?

In any event, both tallies show clear Raila victory, and whole point of NYT story is that if IRI had acted and reported the poll as is the standard practice for exit polls violence may have been avoided.

Anonymous said...

How does IRI ensure an open and democratic process when no matter what their tabulations, the wrong candidate (Kibaki) was declared the winner? They still are not clear on the source of 6 or 8 point result, but no matter as Gibson and Long have reported the same 6 point spread, so appears everyone is in agreement that Raila won it, and decidedly so. They were the only people in possession of nation-wide data to challenge Kibaki's victory (as opposed to the numerous accounts of rigging in specific locations, helpful but hard to quantify in total); so I continue to fail to see how a democracy promotion organization fulfills its mandate by subverting democracy. Does anyone else find this funny?

Anonymous said...

Odd--but certainly not funny

Anonymous said...

Gibson and Long have published their results. You may need an academic login to view for free: