The inconsistencies in opinion polling

In ancient Meroe, a young prince was tested to determine his readiness to assume the throne after the death of his father.

He was sent to a nearby bush to bring home and make his wife the only woman he would find.

There was only one woman in the bush, and that was the woman he would take as his wife. One prince went to the bush and the only woman there was his ailing 70-year-old mother.

She was rushed to Meroe’s top-notch healer as he led 10 elders to the neighbouring kingdom where he picked a 21-yearold virgin for a wife. This prince turned out to be Meroe’s celebrated king.

Unlike the prince who would not capture his mother for a wife, some pollsters in Kenya have formed the habit of sharing very unlikely figures.

If a paper report talks of 10 women in a room and you count only one woman, telling people that the room has 10 women is foolhardy. In the Kenyan context, Jubilee and CORD have strongholds with support for each not likely to fall below 30 per cent. This is the same support you can expect for presidential candidate match-ups. If this is true, how then are we supposed to treat figures that award one of the presidential candidates a figure less than 30 per cent?

Known trends in Kenyan elections include voting informed by tribe and that a tribal kingpin can sway an entire tribal vote to a preferred “outsider,” in this case, someone from another tribe.

While we are always viewing tribe in a negative sense and pretending it doesn’t exist, Americans are not afraid of researching and reporting on race-related voting trends such as the over 95 per cent preference of Obama by African-Americans in both terms of his presidency.

Other known factors include the significantly high impact of huge samples and the import of a mixed methods approach.

Unknowns include the methodology Kenyan pollsters use for sampling in such a tribal country with numerically non-homogenous tribes, the explanations behind some of the glaring errors witnessed recently in some opinion polls, and the future of opinion polling in Kenya.

Concerted effort by pollsters, the media, and members of the public in addressing these issues will help make opinion polling in Kenya good enough to be used for planning and strategising, not just by politicians, but sociologists and economists.

In a rather simple manner, this is what we need to do to make the field competitive: First, we need to open the field to competition.

In the US, a country with the best opinion polls that correctly predicts eventual outcomes, there are numerous pollsters. Media houses such as FOX News and The New York Times do polling.

The same is true for institutions of higher learning and polling companies like Gallup. Few polls can lie outside the circle but due to stiff competition, the job these players do is almost always spick and span.

A polls aggregator, Real Clear Politics, collects all these polls and comes up with an average that, most of the time, is quite close to the actual outcome.

The second step is interrogation of the pollsters regarding method.

The media must ask questions Kenyans come up with when these polls are released. For example in a field of seven candidates, if a pollster grades five candidates with the last having 1 per cent and the other two are classified as “others” with 9 per cent, the logical assumption is that each of the two candidates under the tag “others” has 4.5 per cent. The media should ask why a 1 per cent is identified while 4.5 per cent is ignored. Opinion polls can help swing public opinion, meaning that it is imperative that they be reliable and as realistic as possible.

In our corrupt culture, figure cooking in exchange for money is part of the rot that must be cleaned, and the media can play the clean-up role by scrutinising polls, questioning those behind them.

The third action is creating a peer review platform where the pollsters can respectably question each other. For example when Ipsos releases an Opinion Poll stating that President Uhuru can score 51 per cent against Raila’s 28 per cent, the Centre for African Progress and TIFA leaders can get together with Ipsos and seek to understand this outcome.

Such an exercise will be for cultivation of best practices in opinion polling. We are still making the mistakes some of the developed countries made 100 or 200 years ago when they had nothing or no one to learn from.