From: Mark Kirario

Steadman credibility under censure

THE EAST AFRICAN STANDARD ONLINE EDITION
  Sunday December 2, 2007
By Dennis

The Steadman Group stood by the poll results it released a week ago and denied that the company is facing credibility problems even as questions continued to engulf its latest findings.

But even as Steadman stood by its polls, a fresh cloud of doubt was building around the firm after text messages started doing rounds this week. The messages indicate that the company, under pressure from the Government, will this week release findings that put President Kibaki ahead with 48 per cent against ODM presidential candidate Mr Raila Odinga’s 43 per cent.

There were claims that the scheme is being masterminded by operatives in the Kibaki-re-election campaign with doctoring of polls done with collusion of executives of two media houses and Steadman.

There have been questions from the public, readers and politicians after the figures from Steadman’s latest polls failed to add up.

There were also concerns over the sample frame Steadman used this time round, the pollsters interviewed more people from Central Province, the bedrock of Kibaki’s support, than other parts of the country.

One citizen, who identified himself as Bill and wrote directly to The Steadman Group, asked: How is it that the national poll difference between Raila and Kibaki is a mere 0.3 per cent whilst your consolidated provincial predications have it at a gap more than seven per cent? Are these not the sample votes that will be cast?"

He went on: "I, like hundreds and thousands of Kenyans have come to the conclusion that Steadman has finally bowed to political pressure and manipulated figures that don’t and cannot correspond. Has Steadman forgotten that there is life after December 27? What of its credibility?"

The writer asked whether Steadman’s last poll was a statement "aimed at getting headline news thinking that Kenyans would not read and analyse the finer details."

"It is clear - Raila has more than 15 per cent majority vote in six provinces. Please explain how this then translates into a 0.3 per cent lead at the national level."

At the centre of this controversy is the statistics Steadman released on November 23. In that poll, ODM Presidential candidate was said to be having a thin lead of 43.06 per cent over President Kibaki of PNU, who had 43.3 per cent. Mr Kalonzo Musyoka of ODM-Kenya had 11.4 per cent.

But an addition of the provincial averages showed that the gap between Kibaki and Raila was at least seven per cent. A casual addition of President Kibaki’s support per province as released on November 23, then divided by eight (the number of provinces) showed that Kibaki had scored 39.5 per cent, Raila had 47.6 while Kalonzo had 8.5 per cent, not the figures Steadman released.

In the latest polls, there were also questions about figures that did not add up in the provincial vote.

An addition of the figures from Nyanza Province showed that some 10 points were missing, making the figures not add up.

The real question & Reply from Steadman

Prof Tom Wolf, responding to one of the protesters who demanded answers from the company said there was "no credibility problem that I’m aware of."

"Of course, we’ve got some further fine-tuning to do with the ECK numbers as they continue to whittle down the 188,000 multiple-registered," Wolf wrote in response to one of the inquiries.

Wolf even indicated that "the real questions" should go to the three firms whose pools the Nation Media Group has been carrying with no explanation given for wide variations in their findings.

Steadman Group’s Managing Director Mr George Waititu, responding to questions from The Sunday Standard, said it was wrong to try to arrive at final rating by adding provincial averages.

"We would like to caution that adding up the provincial ratings for each candidate and then dividing by eight (the total number of provinces) is not valid," Waititu said.

"This is because the weight of the proportions within each province does not carry equal weight. For instance, in absolute terms, one per cent of voters in North Eastern equates to approximates 3,100 votes while one per cent in Rift Valley in absolute terms equals to more than 30,000 votes."

Waititu said the accurate thing to do was to establish what exact numbers the percentages represent, add those numbers then divide by the total number of people interviewed in the poll.

Waititu also asked readers to take into account the dates on which a poll is taken.

"This is because public opinion is influenced by political activities at a given point in time. We would therefore like the readers to carefully examine when polls are conducted. Our poll data were collected between November 17 and 19 before finalisation of party nominations."

The Steadman MD maintained that polls "can only go so far in projecting the winner. What ultimately determines victory on election day is the ability of voters to turn out and vote, something that no pollster can pre-determine. But as we’ve suggested, while our polls cannot predict the final tally, we do believe they are an accurate reflection of the voice of the people in the time and manner in which we conduct our surveys."

But the cloud of doubt that hangs over the firm and its findings before they were released has persisted and is getting stronger, despite the management’s stand that it sees no credibility problems.

A day before Steadman released its last poll, ODM issued a statement saying the company planned to release polls whose findings "have already been tainted by discussions held by senior members of President Kibaki’s campaign team and Steadman staff."

Close connection with State House

ODM said such contacts between these two sets of officials undermined the "cardinal principle of impartiality" that lies at the heart of a pollster’s credibility.

"Close connections between President Kibaki’s associates and Steadman officials, as well as persistent reports that one of these associates is a director of the company, have dogged Steadman poll results for some time," ODM said.

The party said suspicions were heightened a fortnight earlier when a poll by the same group pronounced that while Raila was still ahead, the gap between him and President Kibaki had narrowed significantly from the previous poll.

"Independent pollsters immediately raised a series of important concerns about the new poll figures, which Steadman acknowledged were based on a new sampling methodology. One major concern was that this methodology could easily have resulted in more PNU sympathisers having been polled," the party said.

"Even more importantly, unlike other polls, this one did not provide data for provincial and ethnic leanings. The regional data was a significant omission, as it is used to cross check the national figures. To allay fears about the cause of this vital omission, Waititu promised at his press conference to provide the regional data to journalists. But this data has in fact not been provided."

The company is yet to explain why it stopped sampling the likely voting patterns of provinces and ethnic groups.

The province is important in Kenyan polls as the winning presidential candidate has to win 25 per cent of votes cast in five out of eight provinces.

In its statement last week, ODM said the language used in describing the poll results showed how deep the pro-Kibaki bias was at Steadman.

"Among other things, Steadman asserted that the poll showed that the President had "galloped" ahead from the last poll readings. In fact, his rating had risen only three points - which was entirely within the margin of error, and therefore not particularly significant statistically."

But Waititu maintained that the poll released "are statistically correct and above reproach". "We cannot manipulate the poll. This would be tainting the entire science of statistics. But let me tell Kenyans that what they read in the newspapers should not be added up as a percentage of the whole sample."

Waititu said the presidential candidates’ statistics was arrived at upon sampling responses in relation to the proportion of registered voters per province.

The MD also said Mr Joe Wanjui has no position in the company, which belongs to Roger Steadman and Waititu himself.