“There is no reason why Uhuru Kenyatta’s stock should suddenly have risen the way it has. Nothing has changed. Kenyatta is still an ICC indictee (not a suspect, but an indictee, someone against whom what judges consider credible evidence has been laid, which is much more serious).”
Five years ago, as the 2007 election approached, something strange began to happen in the opinion polls conducted by several local polling firms.
A year earlier, on December 23, 2006, a public opinion poll conducted by the then Steadman Group had put Kibaki in the lead in popularity in all provinces except Nyanza.
Overall, he had a 42 per cent approval rating, with Musyoka in second position at 20 per cent and Raila Odinga in third place with 14 per cent.
Over the succeeding year, this picture changed gradually but very decisively, always in the same direction – Kibaki and Musyoka gradually losing support, Odinga gradually gaining.
In September 2007, when Steadman announced the results for its latest monthly poll, Odinga had finally taken the lead, with 47 per cent to Kibaki’s 38 per cent and Musyoka’s eight per cent.
Government supporters now came out to criticise the polls – something they had not done when Kibaki was leading. “Even if Raila is leading, we are not worried,” said roads minister Simeon Nyachae.
“Steadman is a businessman and he is just making money.” Weekly polls by three pollsters were then commissioned by the Daily Nation.
The first results, published simultaneously the following week, on October 7, put Odinga at 46, 46, and 49.6 per cent, with Kibaki at 35, 41 and 39.6 per cent respectively.
(As an interesting aside, Kanu, the monolithic monster people had fought so hard to tame for so many decades, now trailed in the polls with a tiny one per cent rating. It was a situation that would have been unimaginable less than 10 years earlier.)\
The next Steadman poll, published on October 13, just after Kibaki’s launch of his presidential campaign, gave Odinga a 53 per cent approval rating to Kibaki’s 37 and Musyoka’s eight.
PNU was quick to dismiss the poll as “sheer propaganda”, accusing Steadman of working for ODM and the US government. Polls by Infotrak Harris and Strategic Research confirmed Odinga’s widening lead.
Only a poll by the little-known Consumer Insight seemed out of sync with the others, giving Odinga 44 per cent to Kibaki’s 41. It was the beginning.
Another round of polls published on November 4 showed Odinga leading 50 to Kibaki’s 35 (Strategic), and 51.3 to Kibaki’s 32.2 (Infotrak Harris).
Once again, Consumer Insight was out of sync with the others, at 41 per cent for Odinga and 40.6 per cent for Kibaki. The degree of its variation from the other polls began to raise questions.
The next Infotrak Harris poll a week later gave Odinga 50.2 per cent to Kibaki’s 31.8 per cent. But the Steadman poll conducted at the same time showed a narrowing of the gap – 45 per cent for Odinga and 41 per cent for Kibaki, with Musyoka up to 11 per cent.
Six days later, a Consumer Insight poll gave Odinga 40.7 per cent to Kibaki’s 41.4. It was the first time in eight weeks that any poll had rated Kibaki higher than Odinga.
By the following week, all the polls showed a marked narrowing of the gap between Kibaki and Odinga, at (Odinga figures first) 43.2/38.4 (Strategic), 39.5/40.4 (Consumer Insight), and 43.9/ 38.4 (Infrotrak Harris).
On December 7, Infotrak Harris published the results of a poll that showed 52 per cent of voters thought the election results would be rigged.
They were right. There was no reason why Kibaki’s stock should have risen so suddenly. Nothing had happened to make Kenyans spontaneously grow more in favour of Kibaki by a significant percentage.
Ratings do not change just like that, for no good reason. In the same way this year, 2013, there is no reason why Uhuru Kenyatta’s stock should suddenly have risen the way it has.
Nothing has changed. Kenyatta is still an ICC indictee (not a suspect, but an indictee, someone against whom what judges consider credible evidence has been laid, which is much more serious).
Of course, it is well-known that Kenyatta is pouring billions into his campaign – the kind of billions that no other candidate can come close to matching.
Less well-known to most people is that a prominent Kenyan (I have withheld his name) has privately admitted being the conduit for large sums of money that were paid to pollsters in 2007 to raise Kibaki’s profile.
This year, it would appear that there is a much more complex plot afoot, and we are seeing just the tip of this particular iceberg. Let us begin with purpoted political scientist Mutahi Ngunyi’s so-called analysis of purported voting patterns a few weeks ago, which skewed figures massively in favour of Kenyatta and for a few days caused a sensation – as well as alarm and despondency in some quarters.
That was what it was supposed to do, and it was a good piece of propaganda – you have to give Ngunyi that. It fitted seamlessly with William Ruto’s repeated assertion after the launch of Jubilee’s manifesto that “We have the numbers.”
In fact, this was virtually all Ruto had to say to every newscaster who interviewed him at KICC on that day. There was nothing like, “We have the policies” or “We have the programmes”.
No. Just – jubilantly – “We have the numbers.” It sounded alarm bells. Especially because Jubilee does not have the numbers. Ngunyi’s purported analysis of the numbers was itself analysed at length by this newspaper in last week’s Siasa section, which asked, ‘Was Ngunyi’s purported analysis just a hoax?’
I believe it was. Ngunyi’s half-baked figures and selective data do not stand up to scrutiny. Most notably, he completely ignored all the regions of the country where the coming election will be won and lost.
For me, it continues to be astonishing that TV stations and others persist in putting Ngunyi in the spotlight as some kind of ‘expert’ – when all his predictions as far as I can remember have been spectacularly wrong.
His services as a columnist, making very wild and garbled assertions, have also twice been discontinued at another newspaper (whether by him or by the newspaper I don’t know).
The articles he wrote were characterised by a variety of analogies through which Ngunyi attempted to explain what he purported to present as the vice in the country – the articles themselves finally metamorphosing into incomprehensible nonsense.
Ngunyi in fact has a very chequered history. Just before the 2007 elections, when Ngunyi was working for President Mwai Kibaki, he was named as the author of a report entitled ‘Anglo Leasing: Some Thoughts’, a brutal document advising the president to criminalise whistleblower John Githongo’s release of the dossier on the huge Anglo-Leasing corruption scandal.
The aim was clearly to silence Githongo and to wipe details of the scam from Kenyan collective memory, as well as to ensure there was no Anglo-Leasing-related loss of votes for Kibaki.
Ngunyi denied being the author of the document, as reported in the Standard newspaper of January 28th 2006, but suspicions persisted.
The document’s author had demanded shs.2.2 million for consultancy services to counter Githongo’s damaging information. Ngunyi had reasons for revenge on Githongo.
In 2001, Ngunyi was sued for allegedly embezzling more than shs.10 million shillings, awarded some time in the 1990s by the Ford Foundation to an NGO called Series on Alternative Research in East Africa Trust (Sareat), of which Ngunyi was a director.
Ngunyi was sued alongside Zimbabwean minister Professor Jonathan Moyo, who was working as a Ford Foundation programme officer at the time the money allegedly disappeared.
Moyo is also a purported political scientist – one who switched from being a sharp critic of President Robert Mugabe’s government to being his chief propagandist and, according to various Zimbabwean publications, was a major player in Zimbabwe’s crackdown on journalists, the judiciary and the opposition Movement for Democratic Change of (now prime minister) Morgan Tsvangirai.
There are many guns for hire in Zimbabwe. Kenya is no different. In 1998, Ngunyi’s NGO had engaged Githongo to edit a regional political economy magazine called East African Alternatives.
According to reports, the magazine folded after four issues, and its closure was as a result of audit queries made by PriceWaterhouseCooper.
This led to the suit against Ngunyi by the Ford Foundation. And the Foundation’s star witness for the prosecution was to be none other than John Githongo.
The case was apparently later withdrawn, after Sareat agreed to repay the funds in question – but not before Ngunyi had written to the US Federal Bureau of Investigation (FBI) asking them to probe alleged forgery of documents and money laundering at the Ford Foundation’s East Africa and New York offices.
In another strange twist of events, Ngunyi also reportedly turned turtle and offered to act as a prosecution witness against Moyo. In the document ‘Anglo Leasing: Some Thoughts’, the author recommended that Githongo be repatriated from Britain, where he had sought asylum after releasing the Anglo-Leasing dossier, and that he be charged with high treason.
“Githongo has arrogated himself a high moral ground,” the document told Kibaki. “He is telling the world that ‘you are bad and he is good’. You need to deflate him. You need to cast him as an over-enthusiastic officer with good intentions but completely naive.
But more fundamentally, you must criminalise what he is doing.” The author warned that, without this kind of action, the government in the court of public opinion was “guilty as charged”.
The author continued, “The President is implicated directly or otherwise as per the Githongo dossier. This is why we must interpret public anger as a indirect assault on the President, not the implicated ministers... we must not attempt to resolve this scam, not before the 2007 elections.”
The document also advised the government to release the Goldenberg report in a “mirror image tactic” that would “infuriate and confuse”, and recommended that the government must not fire or suspend any implicated politicians.
“They form a ring around the president,” the document continued. “To remove them is to expose Kibaki. They must therefore remain in place and take the blame while inside.
If you remove them, the public will inevitably go for the president, demanding his resignation as well.” In his ‘tyranny of numbers’ scenario a few weeks back, Ngunyi served up several “estimates” of Kenyan voter numbers, but provided no coherent logic or calculations to justify these numbers, making wild assertions arrived at by no known polling methodology.
And that brings us to the current opinion polls and their methodologies and curious anomalies. All the pollsters in the last poll published indicated that they did face-to-face interviews using a digital pad.
While a digital pad is convenient for the researcher, a process audit would find it hard to unearth instances where individual questionnaire entries had been altered.
A hard-copy questionnaire lends itself well to such an interrogation process. Another concern regarding face-to-face interviews is the place of interview.
The pollsters claimed to have done interviews at households between February 13 and 15 – Wednesday to Friday, all working days. Who would have been interviewed in these houses? Househelps?
Househelps have their valid opinions too, but such an information-gathering approach totally obviates the participation of those of working age,18 years and above, who happen to be the main body of voters and thus should be the target respondents for this kind of poll.
Household face-to-face interviews would be appropriate for household consumer market research (Do you use Harpic? Omo?) but definitely not for this kind of political opinion polling.
A house-based poll also works against the main benefits of the random process. If the pollster knows your house, it means your stated political opinion can be traced back to you.
Since your political opinion can get you killed, evicted or isolated, imagine how a respondent in Bondo would feel about publicly professing support for Kenyatta, or one in Gatundu doing the same for Odinga.
A true random process leaves a respondent anonymous, while giving them an equal chance to participate. The process reportedly being used is not random.
More worrying is that, according to evidence leaking out of polling houses, and contrary to reports made to the public, there were actually no house-based face-to-face interviews.
Pollsters have instead reportedly been conducting phone interviews based on a frame of registered mobile-phone users. While this is convenient and significantly cheaper, it is scientifically lacking in the necessary discipline.
The sampling frame for political polling must be the national voters’ register. A database of mobile-phone users does not reveal such crucial demographics as the age of the respondent (to qualify them to vote), whether they are registered voters or in which voter region the phone-user resides.
Sample size distribution is done by regions – yet you cannot verify the regional location of a respondent in a mobile-phone interview. They could be anywhere.
In addition, mobile-phone accessibility in rural areas, where more than 60 per cent of registered voters reside, is unreliable at best, absent at worst.
And crucial is the fact that, not unlike house-based interviews, phone interviews do not support the random process. Even more worrying is the fact that any such non-random process is fraught with potential abuse.
Any pollster with a mind to do so can buy 1,000 mobile lines and distribute them to that number of persons, who then participate in the phone interviews.
There is no proof with this kind of non-random polling that this is not happening, and largely influencing responses. There are many other questions regarding sampling.
Opinion polling is about measuring differences, and thus analysis and presentation focus on measures of dispersion –variance, range and percentile.
For this reason, concern is about factors that occasion difference in opinion. Such factors include age, gender, social class, education level and region.
The first four cannot be determined before the poll and are usually allowed to emerge naturally through the random process, but the overall required sampling frame is the national voter register, and the sample size distribution is according to the weight for each electoral area.
In the poll results released on February 18, pollsters Ipsos Synovate, Strategic and Infotrak claim to have employed the appropriate sampling factors, but the published results raise more questions than answers.
There was no firm attempt to distribute the sample on the basis of counties, the largest electoral boundary. Strategic gave a list of 32 out of the 47 counties where their research was allegedly conducted but not how the sample was distributed to those counties.
No intra-county sample size distribution is discussed by any of the pollsters. Nairobi, Rift Valley, Western and Central are significantly over-sampled by Strategic Africa, while Rift Valley and North Eastern are significantly under-sampled.
A closer look into the counties they reached reveals a significant bias towards Nyanza, Central, Western, Lower Eastern and North / Central Rift. The pollsters’ rationale is that they focused on the top 30 or so counties in terms of registered voters.
But the 32 counties sampled by Strategic Africa account for only 33.3 per cent of the registered voters. The other two pollsters’ samples also average less than 35 per cent.
The implication is that the rest of the counties that are allegedly insignificant account for nearly 65 per cent of the registered voters! The sampling distribution is clearly not representative.
This is of particular importance considering that this significant ‘minority’ (actually, not the minority but the majority, and also the same persons who were ignored by Ngunyi in his lamentably inadequate – read deliberately abstruse and misleading – ‘analysis’) will form the deciding vote.
The counties sampled are traditional strongholds of the two main presidential candidates and tend to cancel each other out in terms of registered voters, thus explaining the ‘tie’ purportedly captured by the pollsters.
Surely, any pollster worth his or her salt who professed to have polled a ‘tie’ at 45 per cent – to which a candidate needs to add only six per cent to win – would consider a constituency of 65 per cent somewhat significant!
The numbers presented in the analysis tables thus appear to reveal a deficient weighting system. It gives a false picture. Ruto does not have the numbers he thinks he has.
It is difficult to avoid the conclusion that, overall, these polls are conducted to a methodological and presentation design aimed at confirming an ‘election too close to call’, which would justify demand for a run-off, or a particular desired result – to be achieved by hook or by crook. Shades of 2007.
Altogether, the methodology and resulting figures appear to make nonsense of the polls being bandied about, regardless of who might seem to be in the lead. The potential difference, however, appears to be largely in Odinga’s favour.
The writer is a freelance journalist.