AMBASSADOR’S TELEX REVEALS US ASSESSMENT OF KENYA’S CONTROVERSIAL 2007 POLL: ‘DID KIBAKI STEAL THE ELECTION?’
The hotly disputed 2007 presidential elections in Kenya between Mwai Kibaki and Raila Odinga led to civil disorder and post election-violence (PEV) in which perhaps 1,300 people died and hundreds of thousands were displaced. To this day the answer to who really won the election and how it was or wasn’t ‘stolen’ rages on.
Whilst I was researching for a major TV documentary series on the murder of Dr Robert Ouko I came across a telex from the then US Ambassador in Nairobi Michael Ranneberger to the State Department in Washington DC setting out his assessment of what may or may not have happened.
The telex was released into the public domain on 9th July, 2012, but for some reason to date no journalist in Kenya has picked up on it, or if they did they didn’t publish it.
This is the full transcript of that telex.
FM AMEMBASSY NAIROBI
TO SECSTATE WASHDC
Subject: Kenya: DID KIBAKI REALLY STEAL THE ELECTION?
Summary: Kenya’s hotly contested December 27 presidential election has been controversial on a number of levels, and many observers still question who actually won. When we looked at and all available data to try and answer that question, we found evidence of rigging on both sides and confirmation that some of the rigging took place inside ECK headquarters itself. By analyzing various datasets (available on request), we developed scenarios that could point to either Kibaki or a Raila victory. We do not think it will ever be possible to tell definitively who actually won the election. This is due in part to the compromise of election officials and election-related ballots and forms, but also because our estimated number of “ghosts votes” (i.e., stuffed ballots) from both sides easily exceeded President Kibaki’s margin of error. End summary.
Ref B provides the context for the disputed results of Kenya’s December 27 presidential election, in which incumbent President Mwai Kibaki was announced the winner by the Electoral Commission of Kenya (ECK) late in the afternoon of December 31, and then immediately sworn in as president. Raila Odinga of the Orange Democratic Movement (ODM) continues to insist that he in fact won the election. Specifically, he claims that he was cheated out of the presidency when a politicized and/or compromised ECK leadership altered constituency-level vote tally sheets in the tense days between closure of polls late on December 27 and the controversial announcement of results in Kibaki’s favour on December 30. Right or wrong, the perception of an election brazenly stolen by the incumbent administration was the initial spark for tribally-based attacks and mass demonstrations that have since left over 600 Kenyans dead and a country embroiled in a major, unresolved political crisis.
But Did it Provide the Margin of Victory?
There is little doubt that there were major irregularities in the way constituency tallies were received, verified and reported by the ECK in Nairobi between December 28 and December 30, and there appears little doubt that the cheating that took place at this level was done so exclusively by Party of National Unity (PNU) partisans in Kibaki’s favour. There was cheating at the constituency level by the ODM and PNU. An interesting question arises: Numerically, did this unprecedented form of central-level, 11th hour cheating in fact make the difference in who won and who lost? In other words, in taking into account all the available data, was the cheating that occurred at KICC significant enough numerically to provide the margin of victory for Kibaki? The answer has important implications for how the current crisis might be resolved.
A Major Caveat About Local-Level Cheating
An important caveat hangs over the analysis below, and over the election results more generally. This is despite praise from all quarters that election day voting was generally free and fair, in fact there is strong circumstantial evidence indicating that more traditional forms of cheating, such as ballot stuffing at polling stations and/or constituency-level tallying centers, probably played an important role in determining the final results of the 2007 presidential election. This was possible because despite the extensive monitoring, there were no party agents or international observers at many polling stations and vote tallying centers (and domestic observation that was more widespread was of mixed reliability). In Kenya, it is very unlikely that a voter would cast a presidential ballot and not a parliamentary one. Yet there were significant discrepancies in six of Kenya’s eight provinces between votes cast for parliament vs. those cast for president.
These discrepancies total 459,100 votes, or 4.6% of all votes cast, dwarfing Kibaki’s margin of victory (230,478 votes, or 2.3 percent of all votes cast). It is impossible to conclude definitely how many of these “ghost votes” went to each candidate, but the margin of uncertainty these extra votes create easily exceeds Kibaki’s margin of victory.
But in moving back to trying to determine how many votes were “stolen” at the central level after polls closed and the results were being returned to Nairobi, it is critical to find credible discrepancies between the presidential vote tallies unofficially disclosed at the constituency-level tallying centers the night of December 27, and the official results announced by the ECK in Nairobi in the days thereafter. Significant discrepancies would indicate that the results were doctored enroute to, or after arriving in Nairobi, as per the claims of ODM and others. To try to quantify these discrepancies, Post gathered polling data from a variety of sources and ran the numbers in several different ways, each discussed below.
The Official Results: Kibaki Wins
The official ECK documentary results show Kibaki winning 4,583,358 votes to Raila’s 4,352,880, for a margin of victory of 230,478. (Note: This dataset comes from hardcopy tally sheets for all constituencies obtained from ECK sources on January 4. These results vary from the ECK results verbally announced December 28-30. But the difference is insignificant – a victory for Kibaki that is 1,363 narrower. End note.) This dataset forms the baseline against which others below are compared for signs of discrepancies.
The Standard Newspaper: Kibaki Wins, But by Less
A spreadsheet obtained January 4 from the Standard Newspaper (which was considered pro-Odinga) reports unofficial results gathered by Standard journalists at the constituency tallying centers on the night of December 27. The Standard dataset has results for all but 12 of 210 constituencies, and in 117 constituencies, the results match those of the ECK. Comparison against official ECK results shows significant discrepancies (500 or more added votes) in 35 constituencies in Kibaki’s favour totalling 191,894. But these gains are counterbalanced to a large extent by the 149,479 votes inexplicably gained by Raila, indicating there was vote rigging by ECK Returning Officers on both sides. The net result: Kibaki still wins with 4,391,464 to Odinga’s 4,203,301. Margin of victory: 188,163.
Other Allegations: Raila Wins Narrowly
In another scenario, Post subtracted from the official ECK results discrepancies reported from the constituency level by a number of observers, monitors, and other sources, including two ECK contacts who provided documentary evidence to us of vote padding in favour of Kibaki in six constituencies by Nairobi-based ECK officials. Other sources include the European Union election monitoring group, which documented anomalies in six constituencies, domestic observers, ODM party agents and partisan blogs. The latter two sources would have to be considered less reliable, but we threw them into the mix to see what would happen. All together, the dataset included alleged discrepancies in 28 constituencies. Numbers from the Standard database were not included in this dataset. With these important caveats in mind, we ran the numbers, and found that Raila comes out on top with 4,375,539 votes to Kibaki’s 4,349,001. Margin of victory: A slender 26,538.
Merging Allegations: A Mixed Bag
Finally, we ran the numbers one more time, combining the numbers from the Standard’s database with discrepancies reported by other observers. The Standard’s discrepancies sometimes matched reports from other sources, particularly with regard to disputed votes for President Kibaki (nine matches for Kibaki versus three for Raila). In other cases, however, ECK, the Standard, and other observers all reported different results for the same constituency (this occurred a total of ten times).
The results of this final analysis were mixed. When giving Kibaki the benefit of the doubt over differing discrepancies, he beat Raila by 26,364 votes. When Raila gained the benefit of the doubt, he however, he came out ahead by 57,425 votes.
Finally, if we combine the four estimates of stolen votes at the central level (paras 7-11) with our estimate of stuffed ballots at the local level (paras 4-5) and then subtract these grand totals from the official ECK results, the scenarios change slightly. Either Raila or Kibaki wins, depending on how the votes won in each constituency by each candidate, the margins of victory are between 30,331 and 114,130 votes. All of these scenarios assume extensive cheating on both sides. In all cases, the margin of victory for either side is slim and ultimately unknowable.
Comment and Conclusions
This analysis is not the last word. Additional data could emerge that would alter the bottom lines, which are heavily caveated to begin with. Thus, our conclusions are by definition very tentative. It is now clear that the ECK became a victim of partisan political interests during the election, and the ECK’s failure as an institution constitutes a dark spot on Kenya’s democratic evolution. (Note: After the results were announced and ECK officials were sent home, a “break-in” at the heavily guarded KICC building occurred. We do not know what was taken or altered, but since there was heavy police security around the ECK, we can only conclude that it was an inside job. End note.)
What the analysis does tell us is that it’s not at all clear who actually would have won the presidency had the election been truly transparent, free, and fair. The only thing we know for sure on the basis of our incomplete and possibly flawed data is that we don’t know for sure, and that whoever won likely did so by a very slim margin. This flies in the face of the position adopted by ODM and others as a mantra: That the election was brazenly stolen by Kibaki’s ECK insiders at the last moment and that Raila should have won by a wide margin. It also contradicts any perception or conviction within the Kibaki camp that the latter clearly won the race. The fact of the matter is, as ECK Chairman Kivuitu said publicly on January 1, “I do not know if Kibaki won the election.”
One implication of our analysis is that a recount or an independent audit of the December 27 results as a practical matter would probably not be able to definitively determine the true winner, particularly since it appears the election’s paper trail has since been compromised. And even if an audit could be conducted, the results might be so close as to fuel further controversy and unrest. Another implication is that holding a run-off may likewise result in a very close race. After what happened during this election, Kenya’s electoral institutions cannot credibly conduct a free, fair, and transparent election until fundamental reform has been carried out.
Martin Minns was the series consultant for the groundbreaking six-part Citizen TV documentary –Murder at Got Alila: Who Killed Dr Robert Ouko and Why?