Grading the pollsters: How accurate are opinion polls?

Tifa Research Director Maggie Ireri. Photo: Jenipher Wachie, Standard

In the run-up to Tuesday’s general election, pollsters released opinion polls on what they felt would be the voting patterns.

But as the results started to trickle in, some of their predictions fell flat, with voters proving the pollsters wrong in several contests.

The main pollsters – Ipsos Synovate, Infotrak Research and Trends and Insights For Africa (Tifa) – have been oscillating in their predictions as to who between President Uhuru Kenyatta and his main challenger Raila Odinga would carry the day.

One week to the polls, Ipsos Synovate put Jubilee’s Uhuru at 47 per cent and NASA’s Raila at 44 per cent. Infotrak put Raila in the lead with 49 per cent, with President Kenyatta at 48 per cent.

Margin of error

An opinion poll for the Meru County gubernatorial race by Tifa showed Governor Peter Munya firmly in the saddle with a 47 per cent lead against Senator Kiraitu Murungi’s 39 per cent.

The polls were conducted less than two weeks to the polls and had an intriguing title, “The Meru County gubernatorial battle is now determined”.

Wednesday, results showed Kiraitu leading with more than 50 per cent of votes tallied in 1,402 out of 1,473 polling stations. Tifa’s opinion poll also had Florence Kajuju leading the county woman representative race with 45 per cent against her closest rival Kawira Mwangaza, who had 38 per cent. Again, Wednesday’s trends showed Mwangaza leading with 52 per cent.

Tifa’s Director Maggie Ireri said there was nothing out of the ordinary in the way elections results pan out vis a vis what various opinion polls stated. Ms Ireri said opinion polls were meant to help politicians gauge their level of success, change strategy and identify areas where they needed to redirect their resources.

“True, we had predicted a Munya win in Meru, but that was before Uhuru Kenyatta implored the electorate to vote ‘six-piece’ for the various positions. It did not help that Munya was in PNU and there was a perception that he was rooting for NASA’s Raila for president,” she said.

While the political class and their minders take the polls seriously as long as they favour them, ordinary Kenyans seem to be least bothered by such predictions.

“The only opinion poll we agree with is the one we gave on August 8. Anything else is a scientific survey that has no basis in reality. In any case, they always tell us that there is a margin of error,” said Peter Makove, a taxi driver in Nairobi.

In an opinion piece published a day before the elections, Masibo Lumala, a senior lecturer at Moi University, said the methodology used by pollsters was based on the old provincial boundaries while the current political dispensation is based on the 47 counties.

In any case, writes Lumala, Kenyans vote purely on tribal lines, regardless of the issues advanced by the pollsters. A poll sample from Central Kenya, he said, would not be indicative of the mindset of the Kenyan voter as most people would ‘naturally’ vote for Uhuru.

Inaccurate predictions

Likewise, a sample from Kisumu, Siaya or Homa Bay would greatly favour Raila.

“The Kenyan scene is quite tricky as voters’ decisions are not necessarily based on issues and manifestos but on tribe and regions. This makes voter representativeness near impossible,” says Lumala.

Angel Ambitho. 

According to Infotrak’s Angel Ambitho, such trends are not unexpected, especially during a tightly contested race. While being careful not to comment on the disputed presidential tally, Ambitho said they got it right on huge voter turnout.

“We had said all along that the elections will not be marked by voter apathy. Kenyans came out in large numbers, especially in the strongholds of the key contenders. We have to wait for the final results from IEBC to see the final voting patterns across the country,” she said.

Opinion polls are the staple of decision making in bigger democracies such as the US and the UK.

However, they have been called out in recent times due to their inaccurate predictions. Cases in point are the recent US elections and the Brexit vote in the UK.

In the US, the pollsters never predicted a Trump win in the November 2016 elections. He went ahead to win and has been berating the pollsters ever since, lumping them in the same category with what he terms as “fake news”.

In the UK, pollsters had predicted a vote to remain in the European Union. Britons opted to leave.

[email protected]