ADVERT 
ADVERTFrom Kenya’s 2022 presidential polls to the Ugunja by-election, the message is brutally simple: political percentages can inform voters—but they cannot predict them
There is a growing problem with Kenya’s political polling industry.
And it is not simply that some polls get elections wrong.
It is that Kenyans are increasingly unsure what exactly they are being asked to believe when a poll is published.
Is it a scientific measurement?
ADVERTA snapshot?
A projection?
A campaign tool?
A publicity exercise?
Or, in some cases, political propaganda dressed up in percentages?
The public deserves better.
And Siaya, perhaps more than most counties, should be particularly suspicious of anyone arriving with a glossy graphic declaring that Candidate X has 43 per cent, Candidate Y has 31 per cent and Candidate Z has 17 per cent.
Where did those numbers come from?
Who paid for them?
Who was interviewed?
When?
Where?
How were respondents selected?
What questions were asked?
And, most importantly, what happened when the voters finally got to the ballot box?
The 2022 presidential election should have humbled Kenya’s pollsters
The evidence is already sitting in our recent electoral history.
Ahead of the 2022 presidential election, Kenya’s major pollsters produced competing pictures of the race between William Ruto and Raila Odinga.
Africa Check’s comprehensive review of Kenyan polling noted that the surveys were split, with some showing Odinga ahead and others putting Ruto in front. The organisation also warned that polls are snapshots of public opinion and highlighted the importance of sponsorship, sampling, methodology, question wording and the timing of fieldwork.
And then came election day.
The polling industry’s credibility took another blow.
Radio Africa’s polling operation was notably closer to the eventual presidential outcome than several of the more prominent surveys that had received enormous public attention. Ipsos, Infotrak and Media Africa gave it to Raila Amollo Odinga.
That does not prove that Radio Africa possessed some magical ability to see the future.
Nor does it prove that the other pollsters were incompetent.
It proves something much more uncomfortable:
Even reputable pollsters can get it wrong.
And that should be enough to destroy the dangerous habit of treating every published percentage as an electoral prophecy.
Africa Check itself has cautioned that no poll is perfect and that the accuracy of a survey depends on such factors as sampling, weighting, question framing and changes in the political environment.
That is precisely why responsible journalism should treat polls as evidence—not verdicts.
And then there was Ugunja
If anybody still needs a local reminder, Siaya supplied one.
The November 27, 2025 Ugunja by-election should be required reading for anybody who believes political popularity can be neatly captured on a spreadsheet.
The campaign produced competing narratives about who was best placed to succeed James Opiyo Wandayi.
At one stage, Fredrick Owino was being described in political reporting as a leading contender, with his professional credentials and ODM connections receiving considerable attention.
Other surveys and political conversations generated their own narratives.
Mizani Africa, for example, published material ahead of the election indicating that Moses Omondi was leading its survey.
Then voters spoke.
Moses Omondi Okoth won decisively with 9,447 votes.
Wiper’s Orodi Odhiambo finished second with 1,819.
Fredrick Ochiel of UDM finished third with 1,200.
The turnout was only 24.98 per cent, with 15,017 of 60,114 registered voters casting ballots.
And there lies the lesson.
The Ugunja campaign generated expectations, predictions, candidate narratives and polling conversations.
But the ballot box produced the only ranking that mattered.
First was first. Second was second. Third was third.
Everything else was commentary.
That is not an indictment of polling.
It is an indictment of the arrogance with which polling is sometimes consumed.
The Ugunja lesson is bigger than Ugunja
There is a tendency in Kenyan politics to ask:
“Who is leading in the polls?”
The more intelligent question should be:
“What exactly does this poll tell us—and what doesn’t it tell us?”
A poll conducted weeks before an election cannot measure what happens between the survey and election day.
It cannot perfectly measure turnout.
It cannot know the impact of a late political alliance.
It cannot predict a party nomination dispute.
It cannot foresee a scandal that erupts 48 hours before voting.
It cannot fully measure the effect of a funeral, rally, endorsement, defection or last-minute mobilisation.
And it certainly cannot vote.
The Ugunja by-election demonstrated that political campaigns are living organisms.
They move.
They mutate.
They surprise.
That is why Siaya does not need another political oracle
As the 2027 Siaya contests begin taking shape, we should resist the growing temptation to outsource political judgment to pollsters.
Siaya has something else.
It has the sangwenya.
The village political intelligence network.
The people who sit under trees and discuss politics.
The boda boda rider who knows which candidate has been travelling through the wards.
The market trader who knows which politician is being warmly received.
The teacher who hears what parents are saying.
The churchgoer who knows which candidate is appearing at functions.
The youth who know which campaign team is actually spending time on the ground.
The professional who follows policy.
The clan elder who understands local alignments.
The journalist who has spent years following Siaya politics.
The ordinary voter who may never appear in a polling sample but knows exactly what is happening in his or her village.
This is not scientific polling.
Nor should it pretend to be.
But it is political intelligence.
And sometimes political intelligence tells you more about the direction of a race than a percentage printed in Nairobi.
So what does Siaya look like today?
Let us say it plainly.
This newspaper is not conducting a scientific opinion poll.
We are not pretending to have interviewed a statistically representative sample of Siaya voters.
We are not manufacturing percentages.
We are offering an editorial assessment based on political visibility, historical performance, networks, public conversations, candidate activity and the broader political terrain.
If the election were held today, the Senate race would, in our reading, produce a strong showing by Cornell Rasanga Amoth, with Walter Okello and Zack Masiro among the serious names to watch.
That is a political assessment—not a poll.
Rasanga’s return cannot be dismissed as political nostalgia. The former governor possesses name recognition, political experience and a network built over years in Siaya politics.
Anybody writing him off at this stage is taking a considerable political risk.
The governor’s race is even more intriguing
If James Orengo decides to seek another term, every calculation changes.
But if Orengo does not contest, the race becomes a fascinating battlefield.
At this moment, our editorial reading puts Nicholas Odero Gumbo in a particularly strong position, followed by David Ochieng, William Oduol, George Okode and other contenders who could yet disrupt the hierarchy.
But there is no coronation.
Gumbo may be strong today and vulnerable tomorrow.
David Ochieng may gain momentum through alliances.
Oduol has the institutional advantage of being deputy governor but also carries the political baggage that comes with being part of the current administration.
George Okode remains a player.
And the field is nowhere near settled.
That is precisely why anybody publishing a definitive governor’s race ranking today should be prepared to explain not merely the numbers—but the assumptions behind them.
Woman Rep? Good luck to the pollster
Then there is the Woman Representative contest.
This may be the race that exposes the limitations of premature polling most dramatically.
There could be as many as 21 aspirants.
Five already appear capable of mounting credible campaigns.
The field is fluid.
Some candidates will consolidate.
Others will disappear.
Some will secure powerful party tickets.
Others will be stranded at nominations.
Some will discover that popularity does not equal organisation.
Others will discover that money does not equal votes.
And some may emerge late and completely change the equation.
So when somebody comes tomorrow with a poll showing Candidate A at 38 per cent and Candidate B at 22 per cent, our response should be simple:
Interesting. Now show us the methodology.
Kenya’s fake-poll epidemic makes the situation worse
There is another reason voters should be sceptical.
Kenya has developed an alarming ecosystem of fabricated political polls.
Africa Check has repeatedly exposed fake surveys carrying the names and logos of reputable research companies.
In 2022, for example, a fake graphic was circulated purporting to be an Ipsos survey of the Siaya governor race. Ipsos confirmed to Africa Check that it had not conducted political opinion polling during that election period.
More recently, Africa Check has documented fake Infotrak-branded polls circulating in Kenya, including fabricated 2027 presidential rankings and bogus county-level surveys. In May 2026, it also debunked an AI-generated graphic falsely claiming that Infotrak had found William Ruto with 80 per cent popularity.
This is no longer a minor nuisance.
A fake poll can create a political perception.
A political perception can attract money.
Money can attract supporters.
Supporters can create momentum.
And suddenly a completely fabricated number has become a political reality.
That is why fake polls are dangerous.
The pollster’s name is not the same thing as the truth
Kenyans must therefore stop confusing institutional reputation with infallibility.
Infotrak can get something wrong.
Ipsos can get something wrong.
Radio Africa can get something wrong.
Any other polling organisation can get something wrong.
The question is not whether the pollster is famous.
The question is whether the particular poll is credible.
Who commissioned it?
Who conducted it?
What was the sample?
How was it selected?
What was the margin of error?
When was the fieldwork conducted?
What questions were asked?
Were respondents registered voters?
Was the data weighted?
Were the results independently audited?
These are not irritating questions.
They are the minimum requirements of an informed democracy.
Africa Check’s polling guide makes exactly this point: polls cannot simply be “fact-checked” in the same manner as a factual claim, but they can and should be vetted by examining their methodology and assumptions.
Brookings got it right
This is where the wisdom of the Brookings Institution becomes relevant.
Writing about polling and public opinion, E.J. Dionne Jr. and Thomas E. Mann warned against using polls as definitive representations of what the public thinks. They highlighted the risks posed by leading questions, sponsorship, weak methodology and political manipulation.
Their conclusion remains devastatingly relevant:
Polling is a tool, not a principle.
Exactly.
A poll is not democracy.
A poll is not a mandate.
A poll is not a campaign.
A poll is not a ballot.
And a poll certainly is not the voter.
The danger begins when polls become political weapons
A candidate who tops a poll may suddenly behave as though the election is over.
Supporters begin celebrating.
Donors become interested.
Party leaders start making calculations.
Opponents panic.
And undecided voters are subtly encouraged to “join the winning side”.
That is the political power of a poll.
And that is precisely why pollsters must be held to an exceptionally high standard.
A bad poll is not merely bad mathematics.
It can alter political behaviour.
Siaya’s politicians should therefore beware
If a poll places you first, don’t celebrate yet.
If it places you last, don’t panic.
If it says you have 60 per cent, go back to the villages.
If it says you have 5 per cent, go back to the villages.
If your WhatsApp group tells you that you are unstoppable, go back to the villages.
If your opponents’ WhatsApp group tells you that you are finished, go back to the villages.
The village is where the election lives.
And the polling station is where it dies—or survives.
Our present Siaya political reading
So, without pretending to possess a scientific poll, this is where we currently see the major races.
SENATE: Cornell Rasanga Amoth currently looks capable of making the strongest showing, with Walter Okello and Zack Masiro and Oscar Omondi among the principal challengers.
GOVERNOR: If James Orengo does not contest, Nicholas Odero Gumbo currently appears particularly well positioned, followed by David Ochieng, William Oduol and George Okode, with the field still highly fluid.
WOMAN REPRESENTATIVE: Too close and too early to call. At least five candidates have credible claims, while a potentially huge field could fundamentally alter the race.
These are not election results.
They are not poll results.
They are our political reading of Siaya today.
And tomorrow could change everything.
The final warning
The lesson from Kenya’s polling history is not:
“Never trust pollsters.”
That would be intellectually lazy.
The lesson is:
“Never surrender your judgment to a pollster.”
The 2022 presidential election demonstrated that prominent polling houses can produce competing narratives.
The Ugunja by-election demonstrated that campaign expectations and ballot-box outcomes must ultimately be separated.
The fake-poll epidemic demonstrated that even the logos of respected research organisations can be weaponised by people seeking to manufacture political reality.
And Siaya’s sangwenya reminds us every day that politics cannot be reduced to a spreadsheet.
So let the pollsters poll.
Let the sangwenyas talk.
Let the politicians boast.
Let the bloggers predict.
Let WhatsApp experts manufacture their graphics.
But let nobody forget the fundamental truth:
The ballot box is the only poll that cannot be edited after publication.
The people of Siaya will eventually mark one name.
No pollster will mark it for them.
No newspaper will mark it for them.
No sangwenya will mark it for them.
No political consultant will mark it for them.
And when the votes are counted, every manufactured percentage will meet the one number that matters: the actual vote.
That is why, for now, Siaya does not need a political oracle.
It needs informed voters, honest journalism, rigorous scrutiny—and politicians brave enough to face the people without hiding behind polls.
Lawrence Jeffrey
9th August, 2026
Siaya
