Why media ignored poll showing Raila leading
A new opinion poll showing Mr Raila Odinga finally overtaking President Uhuru Kenyatta just 20 days to the elections would in most circumstances be treated as a major piece of news.
But the mainstream media generally ignored it despite the best efforts of Nasa to publicise what it considered a significant breakthrough.
The problem was that the survey was paid for by Mr Odinga’s campaign.
But the numbers were still intriguing, especially because celebrated American pollster John Zogby comes with an international brand reputation to protect.
He runs a major operation very different from the myriad briefcase pollsters that crop up every election season in Kenya.
Early last month, for instance, the media also ignored a poll from an outfit calling itself the African Electoral Observers Group showing President Kenyatta winning with 51 per cent of the vote over Mr Odinga’s 39 per cent.
In the first place, the numbers were in marked contrast to other recent polls that all show Mr Odinga significantly closing the gap on President Kenyatta, but still falling short by between five and seven percentage points.
The numbers have all indicated President Kenyatta just short of the 50 per cent-plus needed to avoid a run-off.
The significance is in the 8 per cent undecided voters and lack of any viable third-party candidate, with the other six presidential contenders sharing about 1 per cent of the vote between them.
What this means is that even if the undecided vote was shared equally between President Kenyatta and Mr Odinga, the former would nip past the 50 per cent mark to secure a first-round victory.
But if Mr Odinga maintained his recent surge at Mr Kenyatta’s expense, and then scooped more of the undecided vote, he could be the one to secure a first round victory in the race where a run-off is increasingly unlikely.
The suspect poll was a major departure from the others, but apart from numbers, there were also questions about a group that had suddenly sprung up with no history, no traceable address and no known contact person.
It was in the midst of tightening race indicated by established pollsters and a suspicious poll predicting a Kenyatta victory that the Zogby poll showing Mr Odinga edging past was released.
It significantly showed Mr Odinga leading with 47.42 per cent of the vote to President Kenyatta’s 46.63. It had 5 per cent undecided and 1 per cent shared amongst the minor candidates.
Although the difference between Mr Odinga and President Kenya was less than 1 per cent, the reversal of fortunes was still significant as it indicated where the momentum was in the closing weeks.
Then came the dilemma of how to handle an “internal” poll commissioned by an interested party keen for publicity. The first instinct was to down-play or trash it, but a pollster as established as Zogby is too important to ignore, so Nation decided to at least seek confirmation by contacting him independently.
An e-mail query through his corporate website elicited a ready response: Yes, he was John Zogby, senior partner at John Zogby Strategies, and confirmed participating in the poll commissioned by Nasa.
But on request for more information on the methodology, he responded: “I will leave that to Nasa to release to you with my full blessing.”
The Nation then contacted Mr Odinga’s spokesman Dennis Onyango and Ms Kathleen Openda of the Nasa campaign secretariat, and the latter supplied a more detailed presentation.
A significant fresh detail was that the sampling was not just done in the 47 counties, but in each of 290 sub-counties, presumably corresponding to National Assembly constituencies.
The presentation did not give a breakdown of the sample distribution, but Ms Openda affirmed that the 2,983 respondents were spread across “every sub-county weighted according to its proportion in the IEBC register”.
That might well raise queries on the validity of data where the sample in every sub-county might be too small to draw reliable conclusions from.
But the poll also came with a novel sampling frame where instead of just reporting by region, the demographics are broken down into ethnic groups, weighted by census data.
Those numbers could be significant in a country where the tribe is still the main determinant of voting preferences.
Another novelty is that of the 5 per cent undecided voters, the poll tried to determine where they would eventually vote, and concluded that most, with the biggest majority being Luhya, would lean to Nasa.