In compliance with the law, a flurry of opinion poll results was released last week by different pollsters including Center for African progress (CAP), Ipsos, Nation Media and Infotrack. The law dictates surveys cannot be published five days before the elections.
In all the polls, Azimio candidate Raila Odinga is leading Kenya Kwanza's William Ruto with a gap ranging between 6 per cent and 8 per cent points.
All the poll results are outside the margins of error which means Raila has opened up an unassailable lead.
However, all the polls, except the one from CAP that predicts Raila would get 52 per cent of the national vote against Ruto’s 45 per cent, show Raila would get between 47 per cent and 49 per cent which is less than the constitutional requirement of 50 per cent plus one vote needed to clinch the presidency.
- Poll official's last moments: CCTV footage offers clues
- 'Rebel' commissioners wanted to force poll re-run, says Chebukati
- Kenya Kwanza and Azimio shift election battle to control of top seats in Parliament
- Ruto: My Cabinet members will answer questions in Parliament
While the undecided voters, who range between 1 per cent for CAP to 9 per cent for IPSOs, are expected to tilt the maths in favour of Raila, all the pollsters have failed to provide for the variation of their results that would be accounted for by voter turnout differences in the regions surveyed.
Most of the pollsters’ assessment of the intention to vote that measure the likely voter turnout shows 94 per cent to 99 per cent of the voters would get out to vote. When this sample statistic is generalised to a population of 22 million voters, it is fair to assume between 20.68 million and 21.78 million will vote, which in all likelihood is false.
It is expected the Mt Kenya voter turnout will drop to around 70 per cent from a high of 86 per cent in 2017 due to the absence of ‘one of their own’ on the ballot and the feeling that whatever way the presidential outcome goes, the mountain people shall at least get a deputy president.
Historically, Raila bases of Western, Coast, Nyanza and Lower Eastern regions have not returned as high voter turnouts as those of the late President Mwai Kibaki and Uhuru Kenyatta since 2007.
This means even if Raila’s performance in those regions was near unanimous, there were fewer votes cast nationally in his favour than those of his opponents.
In a presidential election prediction model that I did in 2017 holding the voter patterns of 2013 constant, data revealed that based on the 19.61 million voting population, Raila would have won the election with 50.24 per cent of the votes cast.
In trying to account for the source of the 1.4 million gap in favour of Uhuru in 2017 using the above model, the elections reported data showed poor voter turnout in Raila bases was the main contributing factor and no significant increases in Uhuru's performance in his bases or in those of Raila or as alleged because of artificial manipulation of the results.
As an example, in 2017, if Raila was to get the projected performance of 50.24 per cent, the voter turnout and his performance in Kisumu County needed to remain at the same level of 2013 which was 90 per cent and 98.5 per cent respectively.
However, the analysis of data showed the above scenario did not occur. Instead what happened in Kisumu County was Raila maintained his performance in the county but the voter turnout reduced from 90 per cent to 70 per cent denying him the 109,000 votes that he needed to get the projected performance of 50.24 per cent, consequently contributing to other counties in his stronghold to the loss of the election in quantitative terms.
The 20 per cent reduction in Kisumu County alone accounted for 8.9 per cent of the 1.4 million gap between Uhuru and Raila and not because the former improved his performance in the county.
Turning to the 2022 elections, a simple scenario analysis seeking to adjust the pollsters’ findings by voter turnout trends by regions shows that if the 2017 voter patterns remain constant in 2022 with the only variable changing being the growth of the voting population from 19.6 million to 22 million, Raila would get 46.16 per cent of the votes cast.
If the voter turnout in Mt Kenya West and East reduces from 85 per cent and 79 per cent to 70 per cent respectively; the simulation model shows Raila will be elected the next president as the net effect of reduced voter turnout moves his performance from 46.16 per cent to 51.42 per cent.
This scenario holds true even after assuming the proportion of the votes he got in 2017 remains unchanged in the Mt Kenya counties and that the voter turnout in his stronghold of Nyanza province does not increase to match those of Ruto in Central Rift counties. Raila’s performance statistics in 2017 were 4.48 per cent and 12.3 per cent in Mount Kenya West and East respectively.
All the pollsters’ results indicate on average Raila has increased his performance in Mt Kenya by an average of 20 per cent points, which is a net loss to Ruto.
Ruto has also improved his performance in Western, Coast and Ukambani region except that the simulation model shows the combined effect of his inroads in Raila's bases as reported by pollsters when adjusted for historical voter turnout trends is not sufficient to deny Raila an outright victory.