Election Fraud in Kenya 2007

Would Rivest-Smith Antifraud Voting Protocols have been worth it?

(Return to main page)

Press reports like these give me about 95% confidence that the 27 December 2007 Kenya presidential election "won" by the incumbent Mwai Kibaki, was rigged. The real victor should have been opposition leader Raila Odinga.

To summarize the low points, Odinga led comfortably in 50 opinion polls, including 90% of all major polls during September-December 2007; and also led in all the early election returns. But conveniently, the late-arriving returns 2+ days later swung it to Kibaki despite more than half of his cabinet having been voted out and Odinga's ODM party winning a majority in parliament. Odinga demanded a recount but just 3 days after the election Kibaki was sworn in. An official of the Kenyan elections office came forward claiming it was rigged, naming rigged districts, and claiming computers were used. An EU observer noted how one total announced in front of him changed by 25,000 votes... The EU's head observer, Alexander Graf Lambsdorff, said "the presidential elections were flawed" since "the tallying process lacks credibility." The EU called for an investigation.

The result was violence, riots, and chaos (Over 1000 dead so far as of 12 February – A BBC reporter saw 43 bodies with gunshot wounds in a mortuary in the opposition stronghold of Kisumu, with a witness saying police shot protesters; over 100 homes were burned in Mathare; 250,000 people are estimated to have been forced from their homes with 100,000 now at risk of starving).

The government reacted by suspending television coverage.

And even if by some miracle it actually was not rigged but just looks rigged, then all that, please note, would have happened anyway. Election appearances matter, sometimes at the life or death level of seriousness.

Rigged elections – or perceptions/accusations thereof – have led in the past in other countries to civil wars and/or long periods of repression.

What if Kenya had had Rivest-Smith Antifraud Voting Protocols?

With Rivest-Smith, we get

  1. Essentially total confidence the election was not rigged (if it wasn't rigged).
  2. But if it was rigged, then we get either
    1. Essentially total confidence it was rigged (and a decent estimate of how much rigging there was and in what direction; plus ways to try to track down "whodunit" and "wheredunit");
    2. Or it will be obvious to observers at the polling places that they are not following the protocols.

In case (a), Kenya would be saved from all the suspicion, riots, possible civil war, etc. In case (b), international alliances of countries would be in a position to denounce Kenya and/or to try to impose penalties. (As opposed to the USA's unwise move, which was immediately to congratulate Kibaki on his "victory" and call for the result to be "respected." "The United States congratulates the winners and is calling for calm, and for Kenyans to abide by the results declared by the election commission. We support the commission's decision [to elect Kibaki]," said [US State Dept.] spokesman Robert McInturff. But the USA later questioned anomalies and backpedaled.)

The main price you have to pay for Rivest-Smith, is it's slightly more complicated for voters to vote, although for Luddite voters there could be a "use old-style voting" option which still could be made secure thanks to the remarkable "contagious protection 'easy upgrade' property" of Rivest-Smith (in which the secure new-style ballots "contagiously protect" the old-style insecure ballots).

Are Rivest-Smith's benefits worth the price?

Let's do a numerical calculation. I'm (re)writing this on 15 January 2008 (2 weeks after the election) and so far at least 600 Kenyans have been killed in election-related violence and rising. Probably a lot more will die. If they have a civil war and/or enhanced repression/anti-democracy measures, you can bet on it. Plus there are now (as of 6 January) press reports that 250,000 have been made homeless and 100,000 are going to be at risk of starving. I'm going to consider that to be equivalent to at least 1500 more deaths. So

Benefit of having Rivest-Smith ≥ 2100 lives saved.
(But only 1000 lives saved if the victims are reckoned to have lost 47% of a life on average.)

But suppose, in order to use Rivest-Smith, voters have to think more and work harder when voting, and/or waste time being educated on how to use the new protocol. Suppose that extra effort is N minutes per voter. There were 9 million voters. That is 9N million minutes of voter-time lost, which figuring the 2007 life expectancy of Kenyans is 55 years, is 0.311×N lives worth of time, lost.

Cost of having Rivest-Smith ≈ 0.311×N lives worth of time lost.

So the benefit of adopting Rivest-Smith in Kenya outweighs the cost provided

N<6751
(or in the view that the victims only lost 46% of a life, if N<3106)

There is no way voters will need anywhere near as much as 3106 minutes worth of extra education and/or thinking to figure out Rivest-Smith. The truth is somewhere between 1 minute and a few hours per voter, probably a lot closer to the former. (Plus, even if it did take that long, then not all that time should be reckoned as "wasted" since learning how Rivest-Smith voting works is a useful educational experience with side-benefits.) Also, unfortunately, it is realistic to think the death toll is going to keep rising... [in fact as of later on 31 December 2007 and AFP now reports at least 185 deaths... rising from 149 reported earlier... later, as of 1 Jan 2008, press reports say "at least 200" deaths and some may be starting to claim 250... as of 2 Jan AP claims 300 dead... latest AP News claim on 15 Jan is 600 dead... as of 12 February it is claimed "over 1000" have died... Both BBC and AP news reported on 11 February: "UN humanitarian chief John Holmes said Monday that up to 600,000 people have been displaced in the violence. 'There are something like 300,000 people displaced in camps,' Holmes told reporters in Helsinki after [visiting] Kenya. 'Beyond those 300,000 there are probably just as many who are not in camps who... are sheltering with friends and neighbours somewhere else'."]

So Rivest-Smith is clearly worth it.

Attacks on (and defenses of) this conclusion

One could argue that only waking hours of your life count. We don't necessarily agree with that, but if so, the magic N estimate needs to be reduced by about 33%.

One also could complain that we have not reckoned the money (=time) required to equip Kenya with Rivest-Smith voting-machines and technology. (Which, by the way, do not yet exist.) That objection is serious because in the 2007 USA, equipping populations with voting machines costs about $8 per person. With Kenya's 2007 population of 37 million, that would add up to $261 million, which is equivalent to 261 lives if a life is worth $1 million. And while $1 million might be in the right ballpark for a 2007 US life, one could argue Kenyan lives are worth a lot less than $1 million. Cold hard reality is that Kenya's 2007 GDP in PPP$ per year is equivalent to about $1300 per citizen, i.e. only about $72,000 per life. At that price, this money is equivalent to 3625 Kenyan lives. Also it would take additional money to train election workers. Since "only" 600 people have died as of 2 January (AP), Rivest-Smith might not be worth that cost.

A reply to these critics is that Kenya would have to spend this kind of money anyway over time for whatever other voting machines and election worker training it is going to have. Therefore this "cost" (or a large portion of it) isn't real. Plus, the rioting etc has already probably cost the average Kenyan around $8 worth of some kind of cost (time, hassle, precautions... the entirety of usually-traffic-choked Nairobi was shut down except for soldiers, and Kenya's $billion/year tourism industry is likely to lose over $261 million just this next year alone since who's going to go there? [Reuters: "Bewildered tourists were stranded by delayed flights at Mombasa airport on the Indian Ocean coast..." News.com.au: "Avoid Kenya, travellers told." And from canada.com: "[Kenyan] business leaders said this weekend's tribal clashes were costing more than $30 million a day in lost taxes – not to mention looting damage – and threatened investment in Kenya."]. Plus, it unfortunately is not likely the death-toll is going to stop at only 600, and deaths from protesters and rioters are likely to be only the tip of the iceberg of the price Kenya pays if it has a fraudulent democracy.

Bottom line: although these critic-attacks may have a point, I (Smith) now believe that the costs to Kenya of all this have now risen to such a absurdly huge degree, that that, combined with our defenses to the critics's attacks, is sufficient to cost-justify Rivest-Smith no matter how you reckon it.


Update (Kenya, March 2013)

Although Odinga lost the 2007 presidential election, he took office as Prime Minister in April 2008 (following the violent post-electoral crisis) serving as supervisor of a national unity coalition government. That violence has been estimated to have killed at least 1300 and displaced 600K. The widely-blamed Electoral Commission of Kenya was disbanded and replaced by an "Interim Independent Electoral Commission" (IIEC) on 7 May 2009, then changed to the "Independent Electoral & Boundaries Commission" (IEBC) in 2011 by act of parliament. Then in 2013, the two top contenders for the presidency were Odinga and Uhuru Kenyatta. Kenyatta was the son of former ruler Jomo Kenyatta and one of the richest men in Africa, but both Kenyatta and his running mate William Ruto are due to be tried by the International Criminal Court for "crimes against humanity" for allegedly bankrolling gangs to exact "tribal revenge" (those gangs killed at least 40) during the 2007-8 crisis. The court announced that the start of Kenyatta's trial would be delayed from April until July 2013.

Election Result: On 9 March, Kenyatta was officially declared victor with 50.03% of the vote (provisional) then 50.07% (final announcement). That 0.03% or 0.07% was crucial since without it he'd be forced into a second runoff election versus Odinga. Unfortunately I have major doubts that any country, including Kenya, is capable of counting votes with this degree of accuracy even if we had no worries whatever about fraud. This 0.07% is about 7000 votes, and the difference between the provisional and final tallies thus was about 4000, indicating a "noise level" of at least 4000 in the winning margin.

Contrast: In the Florida 2000 USA presidential election, a media consortium including the New York Times, Wall Street Journal, Tribune Company, Washington Post, Associated Press, St. Petersburg Times, Palm Beach Post and CNN hired the National Opinion Research Center at the University of Chicago in January 2001 to examine the 6 million ballots (using three coders per ballot to avoid possible coder mistakes). The resulting database of all Florida ballots gave different totals depending on what precise definition of "a valid ballot" was employed, e.g. "detached chads," dimples, chad hanging by 1 corner, 2 corners, what sort of scribbles were regarded as valid votes or not, how to handle "overvotes" of different kinds, including the kind where somebody votes "Gore" and writes in the name "Gore," whether two or all three of the coders had to agree about a ballot, etc. Using the most restrictive standard – fully punched ballot – NORC found that 5252 more votes would/should have been added to Florida's 5.96 million, and that the new total would have elected Gore by 115 votes, not Bush by 537. (The state of Florida was counted using non-uniform standards and never officially recounted, hence Bush officially won by 537. Gore tried by court challenges to force a partial recount, i.e. in some but not all counties. Gore lost the case in the US supreme court; if Gore had won it, it appears he still would have lost the election by 225 votes because he foolishly did not ask to recount the whole state of Florida.) Other standards would have produced other results, although with most combinations of choices, Gore won. NORC made available an interactive online tool where you could input your favorite combination of definitions of "votes you'd accept," and it would tell you the election result under that definition-combination.) My lesson from this very heavy and expensive study is that the "noise level" in Florida 2000 was about ±5000 votes as far as counting the total number of ballots was concerned, and about ±200 as far as the Bush-Gore margin was concerned (using uniform statewide standards) and much larger, say 1000, with the nonuniform biased-toward-Bush standards they actually used. And that is just the tip of the iceberg – if we further tried to investigate which voters were and were not really permitted to vote (since, e.g, felons, fraudulent addresses, falsely accused of being felons, etc) then far more "noise" would have been introduced...

Returning to Kenya, The Associated press, Nairobi, Thursday 7 March 2013, reported:

The political coalition led by Kenya's PM Raila Odinga, said on Thursday that the vote-tallying process now under way to determine the winner of the country's presidential election "lacks integrity" and should be stopped, and alleged that some vote results had been doctored.

Odinga claimed there were several districts where the number of votes cast exceeded the number of registered voters. Also the spoiled ballot count was claimed wrong by a factor of 8 due to "software bug," and Kenya's "secure vote-results communications system" malfunctioned. This report by Peter Greste discusses the situation.


Return to main page