Why opinion poll results are questionable

Infotrak chief executive Angela Ambitho speaks during the release of an opinion poll in Lavington, Nairobi, on July 23, 2017. Ms Ambitho was accused by Jubilee campaigners of preparing to release a Raila Odinga-friendly report. PHOTO | DENNIS ONSONGO | NATION MEDIA GROUP

What you need to know:

  • Often, polling results differ dramatically depending on where precisely the respondents are.
  • Voters will be making their choice and also exposing the pollsters who got it wrong.

The two opinion polls released on Sunday may have differed widely on who between President Uhuru Kenyatta and his challenger Raila Odinga is ahead in the presidential election race but they concurred that victory will most likely be in the first round.

There is little chance of a run-off under the scenario presented in the Ipsos poll released by lead researcher Tom Wolf and the Infotrak poll released by chief executive Angela Ambitho.

They concurred that it’s a two-horse race between President Kenyatta and Mr Odinga that will not be upset by any third party candidate securing a vote substantial enough to rob the one who pips the other to the post an absolute majority.

Both polls have the other six presidential candidates sharing about one per cent of the vote, which corroborates all the other surveys published this year.

The other substantial finding was, of course, the Infotrak one that showed Mr Odinga dramatically gaining support since the last poll by the same company in June and overtaking President Kenyatta in the race to State House.

The Infotrak numbers actually mirror quite closely the findings of an internal poll done for Nasa by renowned American numbers cruncher John Zogby and are bound to raise renewed controversy because of persistent accusations by Jubilee Party that he is partial to the opposition leader.

While the Zogby survey could easily be dismissed because it was paid for and publicised by the Nasa campaign, the Infotrak one should attract attention as the first independent poll in the run-up to elections to place Mr Odinga ahead of President Kenyatta.

The survey showed Mr Odinga’s share of the vote growing impressively by four percentage points from 43 per cent to 47 and President Kenyatta’s dropping by two points from 48 per cent to 46 per cent.

The Zogby poll, released just a few days earlier, had Mr Odinga at 47.42 per cent against President Kenyatta’s 46.63 per cent.

However, both are quite a contrast to the Ipsos numbers released on the same day as Infotrak.

Ipsos confirmed the trajectory established by all the other surveys by different Kenyan pollsters over the past month or so — Mr Odinga closing in since he was confirmed the Nasa presidential candidate but not quite threatening to overhaul President Kenyatta’s lead.

It had the race tighten with President Kenyatta down one percentage point from the end of June to stand at 47 per cent and Mr Odinga up by one point to stand at 43 per cent.

It gave five per cent as the number of undecided voters.

It might be significant that the polls were conducted over different periods — Ipsos’s between July 2 and July 12 and Infotrak’s from July 16 to July 22.

Still, that might not be enough to explain Mr Odinga’s dramatic surge in the Infotrak poll.

Ms Ambitho, who the week before the poll was unveiled was accused by Jubilee campaigners of preparing to release a Raila-friendly report, tried to offer an explanation but probably still left many doubts.


One was that undecided voters from the last poll had finally made up their mind.

“More people have decided. Nasa had more ‘undecideds’ in their strongholds than Jubilee did,” Ms Ambitho offered.

But then, that explanation may not be borne out by the numbers.

Infotrak had ‘undecided’ at six per cent — a decrease from the eight per cent in the previous poll but maybe not big enough to explain Mr Odinga’s four per cent gain.

The numbers also do not support the explanation that the ‘undecided’ had substantially fallen in Nasa strongholds.

Western Kenya, with the block of undecided over the year, fell only marginally from 18 per cent to 17.8 per cent.

In Nyanza, the proportion of ‘undecided’ increased from three per cent to 3.8 and in Nairobi from six per cent to 9.5 per cent.

The other notable shifts in the ‘undecideds’ were in the Odinga-leaning Coast, which saw a decrease from nine per cent to 3.6 per cent, and Kenyatta-backing Rift Valley from 10 per cent to 3.1 per cent.

The Infotrak shift probably still needs further interrogation, especially if the poll used the same sampling frame as the previous one.

Often, polling results differ dramatically depending on where precisely the respondents are.

That is why pollsters settle on a sampling frame they are confident is representative in terms of not just demographics but also known political attitudes.

They stick to it for the sake of consistency.

Different sampling frames will often explain variations in findings from different pollsters but, overall, any competent national poll should not differ too markedly unless one of them is totally erroneous.

That’s what remains to be judged on the opinion polls being released in the run-up to elections.

Voters will be making their choice and also exposing the pollsters who got it wrong.


You're all set to enjoy unlimited Prime content.