The December 2007 Kenya presidential election: where did fraud occur?
Dr. David Ndii
and
Dr. Karuti Kanyinga
Electoral fraud is hard to prove. This is one of the reasons as to why political operatives frequently engage in it the world over, though not always successfully. The predicted closeness of the presidential race before the election may have contributed to an irresistible temptation on all sides to participate in fraud—or at least predict its occurrence. Since the counting began, members of political parties, accredited observers, both domestic and international, as well as civil society organisations and wananchi have lodged allegations of rigging. Commissioners of the ECK including its chairman have also cast aspersions on the results.
But what truth is there to these claims? There are two opposing claims. There are those who contend that it is still possible to get to the truth, that is, to know who Kenyans elected president of the republic on the 27th of December. The other view is that the records have been compromised to the extent that the true outcome will never be known. In this article we provide some preliminary analysis of the official presidential election results released by ECK which provides statistical evidence consistent with the claims that significant fraud may have taken place. Our analysis entails showing the discrepancies between parliamentary and presidential vote tallies. We use the 2002 results as a benchmark to arrive at an estimate of the magnitude of fraud that could have taken place. Numbers by themselves do not “prove” every instance fraud, but they are highly suggestive and allow us to provide information on whether realistic attempts at fraud might have swung the result of the presidential election. Our analysis suggests that the magnitude of fraud would have been sufficient to affect the outcome.
Presidential and Parliamentary Turnout
The basic premise of our analysis and contention is the following: the overwhelming majority of Kenyan voters cast ballots for all the three offices that they are offered the opportunity to elect. It is rare for a voter to cast a ballot for his preferred presidential candidate and ignore or decline to cast a ballot for his/her preferred MP and Councillor. Variances between the presidential and parliamentary election will arise primarily as a result of differences in the number of spoilt ballots in the two elections, but there will be also a small number of abstentions – conscious or otherwise - from one or the other.
Numbers for both the 1997 and the 2002 elections support our argument. The turnouts for the parliamentary and presidential races were almost identical. We use the 2002 elections as a baseline because that election did not carry claims of presidential rigging. Although the 1997 presidential election carried this claim, we note that there was no marked difference between the total valid votes cast for presidential and the total valid votes cast for parliamentary candidates (except in about 10 constituencies where MPs were elected unopposed).
In 2002, valid votes cast for parliamentary candidates exceeded valid votes cast for presidential candidates in about 48 constituencies by a total of 114,000 votes. This is equivalent to equivalent to 1.9% of the presidential votes in those constituencies. However, two constituencies Bomachoge and Kasarani had unusually large variances, 40,000 votes between them, close to one third of the total. If these two are excluded, the variance is 74,000 equivalent to 1.2 percent of the valid votes. Ninety six constituencies had variance in the other direction, that is, where presidential votes exceeded parliamentary votes. This amounted to 64,000 votes, equivalent to 1.1 %. As is evident, the variance in the two directions almost cancels out leaving about 10,000 votes difference countrywide. This variance is consistent with differences in the number of spoilt ballots and a few voters who may have voted for one office and not the other. . Regardless of the reason, however, the variance could not swing the presidential election in 2002. Neither could the difference in the 1997 presidential election affect the overall outcome. Going by this analysis a difference of around 1.2% between the presidential and parliamentary valid vote is what we have taken to be normal.
The variance between valid parliamentary and presidential votes in 2007 is startling. A review of the result – excluding five constituencies for which one or the other results are unavailable s—produces a number of unrealistically high turnout variance.. There are as many as 35 constituencies where the variance is above 5 percent which translates to over 237,000 votes. These constituencies include instances where the variance is above 10,000 votes (one constituency has a variance of over 30,000 votes). This would mean implausibly, that about 10,000 voters in some of these constituencies deliberately chose not to vote for an MP. They voted for their presidential candidate and walked out of the polling station. There are about 70 constituencies where the variance is above 2%.
The parliamentary election has 25 constituencies where the parliamentary vote exceeded the presidential vote by more than 2%, a rather liberal cut-off given the norm of 1.2%. Looking at raw votes, this disparity produces about 116,000 ballots that are anomalous. This means presidential candidates lost about 116,000 votes. Factors responsible for this loss or wasting of presidential vote are not clear. The main observation to make is that this number of votes is critical for shaping the final outcome of presidential election and in particular an election that was too close to call.
.
Does it Make a Difference?
What do these numbers mean? If we add votes where parliamentary turnout was unrealistically high to ballots where the presidential turnout is suspect, we the anomalous votes. By adding up the variances, and adjusting for the normal variance derived from the 2002 election data, we have arrived at 365,000 suspect votes.
A parliamentary vote may exceed a presidential vote because parliamentary supporters “stuffed” ballots, presidential supporters “wasted” ballots (or reduced those of the rivals), or some combination of the two. A presidential vote may exceed a parliamentary vote because parliamentary voters were wasted, or presidential votes stuffed, or a combination.
Our analysis is restricted to those constituencies where on balance; the differences between stuffing and wasting were great enough to produce abnormal variance in the turnout rates. There may in fact be a number of constituencies where either stuffing or wasting occurred in both races, such that turnout rates are close but parties committed fraud nonetheless.
We have not attempted to apportion the suspicious votes to either presidential or parliamentary rigging. It would be surprising if shenanigans at the parliamentary did not occur. The winning margin in the presidential election declared by ECK is 231,728. The suspicious votes exceed the winning margin by close to 130,000 votes. Fraud of this magnitude is more than sufficient to have altered the outcome of the presidential election.
What next?
Debate rages as to whether fraud did in fact take place; whether it took place at the polling station level, at the constituency count, or in the ECK; and whether Kenyans will ever truly know the results of their 2007 election. The hotly contested terms of this debate suggest that Kenyans deserve an internationally monitored audit of the vote count.
We generate our results from only looking at one type of anomaly. If fraud was indeed committed in other ways, the magnitude of the problem may even be larger than what we suggest here. Re-tallying of the votes – no matter the status of the records – and close scrutiny of results in each of the 28,000 plus polling stations will contribute to knowing the truth about this election.
There are those who fear that the truth about the election will divide the country even more—that we are better of not knowing. We believe that the truth is the first stage in the process of reconciling a nation in a political crisis. Denial—clinging desperately to the belief that we are right and they are wrong—will only deepen the crisis.
Dr. David Ndii is an economist and researcher based in Nairobi and Dr. Karuti Kanyinga is a political scientist and researcher at the Institute for Development Studies (IDS), University of Nairobi.
Next entry: Moderate majority must defeat ethnic hawks
Previous entry: News Roundup 21 January (2)