My issues with the latest Infotrak Results

There are three Opinion Poll results that have been released today, the one from Infotrak, Ipsos and CTD. The Infotrak results predict a Raila win with 49% of the votes against Uhuru’s 48%, Ipsos Synovate Poll predicts an Uhuru win with 47% of the vote against Raila’s 44%, and the unknown pollster called Centre for Transitional Democracy has predicted a Raila win with 53% against Uhuru’s 39%.

The details of the polls by Ipsos and CTD are yet to be shared, but Infotrak has shared their methodology and the breakdown of the poll results across the former provinces. According to the methodology of Infotrak, some 5000 Kenyans were interviewed in a span of 5 days (July 27th and July 31st) via Computer Assisted Personal Interviews (CAPI) across all the 47 counties. Due to the large sample size, the margin of error in the latest polls is 1.4% compared to the previous standard of 2.2% margin of error when the sample size as around 2,000 respondents.

By margin of error therefore, and if the undecided voters split in the middle, we should expect with 95% confidence a possible result like Raila winning with 51.4% against Uhuru losing by 47.6% , or an Uhuru win by 50.4% against Raila losing by 48.6%. The results between those two extremes are also possible, including the possibility of a tie at figures around 49% for each of the two leading candidates which would lead to a runoff. The Infotrak poll therefore does not tell us who can actually win the forthcoming general elections.

Kenyans supporting either side think otherwise, and this can be seen by how they are both denouncing the polls on Twitter, depending on the side of the political divide they are leaning on. Interestingly, almost all Polls conducted on Twitter on where these Kenyans express their reservations on the Infotrak polls predict that Raila will win by about 60% of the votes against Uhuru’s 35%.

I also have my own reservations on the Infotrak polls, and the reservations are not based on the possible fact that the elections are too close to call, but based on both their reporting and a few interesting findings. On the reporting, Infotrak still groups their findings according to the now defunct provinces.

Before the 2010 general elections, grouping results by provinces could have some sense as these provinces were under the same administrative unit, and the administrator could have some influence on how the province could generally lean. However, the 2010 constitution got rid of the provincial administration to usher in the county governments, which have tended to take individual standpoints that can be so different from the neighboring counties but of similar demographics. For example, since the 2013 general elections, we have had CORD governors changing loyalty from CORD to Jubilee in Kwale, Machakos, Bungoma and a few others. In today’s political alignment, we find the Jubilee leaning Governors of Meru and Uasin Gishu being NASA friendly, and previous Jubilee Governor Isaac Ruto becoming one of the NASA principals.

Lumping Machakos and Meru under the larger Eastern Province should therefore be very misleading. Similarly, lumping Kwale with the larger coastal province can also be very misleading. In Rift Valley, we have counties such as Kajiado, Samburu and Narok that are generally swing counties and shouldn’t be lumped together with the Jubilee stronghold counties of Nandi, Elgeyo Marakwet and Baringo.

Now, if I wanted to do a more accurate prediction based on Infotrak results, I may not be able to do so since the data that I have from IEBC register are grouped by counties. To group these IEBC figures by the former provinces and assuming that on average Uhuru will have a particular percentage from that province would give me totally unreliable results.

The second issue I have with the Infotrak results are two general county findings. Although they have not given specific percentages on the counties, Infotrak has provided the counties that are generally Jubilee strongholds, the ones that are NASA strongholds, and the ones that are battlegrounds. Three counties caught my attention.

The first two counties are Nyamira and Kisii counties. These two counties voted very similarly in the 2013 general elections where Kisii gave Raila 69% of the votes against Uhuru’s 27% of the votes and Nyamira gave Raila 67% of the vote against Uhuru’s 29% of the vote. This similar voting pattern was expected due to similar demographic constitution of the two counties. Today, nothing has changed as far as the demographics of the two counties are concerned. However, Infotrak has reported that Kisii county has remained a NASA stronghold whereas Nyamira has shifted from a stronghold to a battleground.

The third county is Tana River. This county voted for Raila at 62% in 2013, but since then most of the key politicians have decided to support Jubilee. This support however is not expected to translate in that large ground support where the county suddenly changes from being a NASA stronghold to a Jubilee stronghold. I would expect that NASA could have lost ground to make that county a 45/45 battleground – but not for the county to become a strong Jubilee stronghold as implied by the Infotrak results.

The two issues that I have with the Infotrak results simply means I can’t trust their predictions. Let me wait for a complete report from Ipsos and see if their data is more trustworthy, even though last time they had this huge discrepancy with the 61% of Kenyans who thought the country was headed to the wrong direction but only 44% of the very Kenyans would vote to change that direction.