The Radio Africa research team was the only research team to correctly predict the result of the August 9 presidential elections in Kenya.
The poll results were published in the Star newspaper, an affiliate of Radio Africa Group which also owns the Kiss, Classic and Radio Jambo radio stations. Radio Africa has a full-time team doing market research.
The Radio Africa poll in early August was in contrast to Tifa, Infotrak and Ipsos who all predicted a landslide win for Azimio candidate Raila Odinga. In the last opinion poll before the election, Tifa gave Raila a lead of 8 per cent, Infotrak 7 per cent and Ipsos 6 per cent.
Radio Africa however gave William Ruto a slight lead of 1.3 per cent, the only poll that predicted a Ruto victory.
The Radio Africa poll grouped the country into 10 polling zones and correctly predicted seven out of 10 and was off slightly in only three areas.
The Radio Africa poll of 3,000 respondents from all 47 counties predicted a very tight race with William Ruto slightly edging out Raila Odinga. The Radio Africa poll almost exactly matched the presidential results as announced by the IEBC. It carried a 1.8 per cent margin of error.
The Radio Africa poll, published in the Star newspaper the week before the election, indicated that 45.5 per cent of respondents supported Ruto compared to 44.2 per cent for Raila.
But when Undecided respondents were eliminated, the Radio Africa poll gave Ruto 50.4 per cent of the vote compared to 49.0 per cent for Raila.
This was remarkably close to the actual results announced by the IEBC where Ruto got 50.5 per cent of the total vote and Raila 48.8 per cent.
The adjusted (removing Undecided respondents) opinion polls for Tifa, Infotrak and Ipsos predicted a Raila win with a gap of 8.8 per cent, 7.6 per cent and 6.8 per cent respectively.
All pollsters erred with over-optimistic turnout predictions compared to the final turnout of 65 per cent.
Radio Africa started conducting polls for the August 9 election in July 2021. During that period, Raila Odinga closed the gap on former Deputy President William Ruto and even overtook him briefly, according to the Radio Africa polls.
Further drama occurred after the elections as media houses were told to stop tallying the results live on air. The tallies, based on Form 34As and Form 34Bs posted on the IEBC portal, showed a very close race with the lead shifting between Ruto and Raila.
Reportedly on the afternoon of Wednesday, August 10, government officials called Citizen TV, owned by SK Macharia, NTV and KTN and asked them to stop announcing publicly their vote allies.
The Kenyan media then went silent on the vote count although the BBC and Reuters subsequently reported that Ruto had a small lead of votes counted by IEBC.
On the morning of Sunday, August 14, the Star website released the Radio Africa tally of results were tallied from a total of 46,136 polling stations out of 46,232 polling stations. The Radio Africa tally showed that Ruto was leading the vote count with 50.5 per cent with less than 100 Form 34As still to be verified and uploaded.
The story headline was 'Ruto leads tallies as IEBC presidential results delay' and effectively it showed that William Ruto had won the presidential election. The Star website had 5.9 million unique visitors in the 30 days over the election period.
The Star website posted a second story at the same time at https://elections.thestar.ke where the data from the tally from all 47 counties was presented. Within 15 minutes of being posted, that story suffered a DOS (denial of service attack) making the data unavailable online. It took several hours for the Star’s technical team to stabilise the service and make the data available.
At 6pm on Monday, August 15, IEBC chairperson Wafula Chebukati finally announced the results at Bomas of Kenya. The results were virtually identical to the Radio Africa tally released the previous day on the Star website and very similar to the Radio Africa opinion poll conducted two weeks earlier.
"Polling is important for Kenya's democracy. It reflects the will of the people and gives our politicians an undiluted and genuine reflection of the concerns and aspirations of our citizens. We take it as a serious responsibility to be as accurate as possible," said Joshua Oluoch, head of research for Radio Africa group.
"Prior to asking the presidential vote question, we asked the respondents whether they were planning to vote and their likelihood to vote. We would then eliminate those unlikely to vote from the presidential poll question," Oluoch said.
"Hopefully polling in Kenya will only get better and less partisan over the next decade," Oluoch added.
Other pollsters were unavailable for comment.