Over the past few weeks, there has been much discussion about the credibility of pollsters, especially when those in the US got it wrong.
Mr Donald Trump, who had been leading in the past 10 polls conducted in Iowa, to the surprise of many, failed to hold his lead. The last opinion poll predicted that Trump would win with 31 per cent but he garnered only 24 per cent of the votes!
At face value, this looks like an error of 6 per cent — but a more critical look at the numbers paints a more damning picture for US pollsters.
Mr Trump underperformed his polls, while his rivals, Mr Cruz and Mr Rubio, outperformed in theirs. The pollsters had predicted that Mr Trump had a 7 per cent point lead over his rivals, but the outcome was a 3 per cent loss to Mr Cruz. If you do the math, this translates to a 10 per cent error by the pollsters.
What happened in the US last week has similarities with Kenya. During the 2013 pre-election period, the polls consistently showed that Mr Odinga was ahead of Mr Kenyatta.
As the country edged closer to the General Election, the last two polls released indicated that Mr Kenyatta had closed the gap. Pollsters also predicted that neither Mr Kenyatta nor Mr Odinga would garner the required 50 per cent +1 in the presidential race and concluded that a run-off was imminent.
Contrary to the pollster’s predictions, Mr Kenyatta was declared the winner of the presidential race after he garnered 50.3 per cent of the votes. This translates to an error of approximately 6 per cent by pollsters.
The two occurrences in the US and Kenya beg the question — are the results published by pollsters reliable? Can pollsters be trusted again to give Kenyans an indication of the election outcome?
As we get closer to the 2017 elections, opinion polls will become a common feature. Kenyans need to be acquainted with some factors that could result in a significant difference between opinion polls and the actual election outcome.
The first factor is the bandwagon effect — it occurs when opinion polls embolden voters to support the candidate who is shown to be winning.
Assuming there are at least three presidential candidates, and a voter supports a candidate who is ranked third in opinion polls, he or she may consider it futile to vote for that candidate, and may opt to vote for either of the top candidates.
In the lead up to the 2013 presidential elections, the last two opinion polls showed that Mr Kenyatta’s and Mr Odinga’s total support was at 88 per cent against the other four candidates, who were expected to garner a total of 12 per cent.
It was evidently a two-horse race, with the last opinion poll showing a difference of 1 per cent between Mr Kenyatta and Mr Odinga. A number of Kenyans have indicated that the last opinion polls could have influenced their voting, as they felt that going for the smaller candidates was futile.
For instance, one voter was passionate about voting for Ms Martha Karua, even up to the Election Day. But as she stood in the queue to vote, she began to think more critically about whether her first choice stood a chance.
By the time she got to the ballot box she had made a decision to vote for her second best candidate (one of the two frontrunners), thereby demonstrating the bandwagon effect.
The second factor that can influence election outcomes is voter turnout. Opinion polls assume equal voter turnout from each candidate’s strongholds. Should there be a drastic shift in the turnout, leaning towards one side, opinion polls could be radically different from the actual election results.
Looking at the last election, a lower voter turn-out was a key factor leading to the variation. Looking at 20 counties with the highest voters, Mr Odinga had a higher number of total voters registered in his strongholds (8.5m), compared with those registered in Mr Kenyatta’s strongholds (8.1m).
Voter turnout in Mr Kenyatta’s strongholds was 88 per cent, compared with 84 per cent in Mr Odinga’s counties. Mr Kenyatta was able to win 75 per cent of the votes cast in his top 20 counties (5.3m votes), but Mr Odinga won only 63 per cent of the votes cast in his top 20 counties (4.5m).
So although Mr Odinga had a higher potential to garner more votes in his strongholds, lower voter turnout affected this potential.
The third factor is the underdog effect. This is the opposite of the bandwagon effect, which occurs when people vote out of sympathy for a candidate predicted to be less popular.
In the 2013 presidential elections, Mr Mohammed Dida charmed the audience during the presidential debates. Then the unexpected happened, Mr Dida, a totally new candidate, garnered more votes than political heavyweights such as Ms Martha Karua and Mr Peter Kenneth.
A substantial number of Kenyans voted for the underdog Dida, hence validating the underdog effect. But it should be noted that there is less empirical evidence for the existence of this effect. Human psychology is such that we like to associate with winners.
The fourth factor considers last minute changes in voter intentions — just before an election, a candidate could lose due to a shift in ideologies when voters begin to have contrary views.
This factor, to some extent, affected Mr Trump’s support in Iowa. An opinion poll done on the election date sought to find out voter intentions a few weeks prior to the election.
This survey revealed that the support for Mr Trump had declined four weeks prior to the election. Four months before the election, 39 per cent of voters had intended to vote for him, but this had declined by 15 per cent.
There were definitely reasons why they changed their mind. The “guestimate” is that Mr Trump’s outrageous comments made voters question his ability to lead.
Despite these dynamics, opinion polls will continue to be an indispensable feature of elections. Polls can be instrumental in assessing how hard a candidate needs to work.
The writer, Ms Maggie Ireri, is the director of local research firm TIFA