Polling organisations and election analysts have applauded their efforts for exceptional accuracy in projecting the results of the groundbreaking general election.
In the lead-up to the national and provincial elections, organisations including market research company Ipsos, think-tanks such as the Social Research Foundation (SRF) and the Brenthurst Foundation, and Standard Bank and Wits University conducted surveys. These consistently indicated that the ANC’s support would dip below 50%.
In the end, the ANC received 40.18%, DA 21.81%, uMkhonto we Sizwe (MK) 14.58%, EFF 9.52%, IFP 3.85%, Patriotic Alliance 2.06%, Freedom Front Plus (FF+) 1.36% and ActionSA 1.2%.
In March 2023, data by the SRF, a think-tank focusing on public policy issues and the promotion of democracy, showed the ANC’s electoral support at 45.9%, DA at 23.3% and EFF at 8.4%. In November 2023, a survey by the Brenthurst Foundation put ANC support at 47.6% in 2024, while an Ipsos poll commissioned by political think-tank Rivonia Circle showed the party dropping to 41%.
An Ipsos poll in March 2024 showed ANC support falling to 39% nationally on a 66% turnout, DA 27%, MK 13%, EFF 10% and IFP 2%. In April, however, Ipsos released another survey in which the ANC was projected to drop to 40.2%, down from the 40.5% it polled on February 6 and 43% on October 27 2023. The DA was projected to receive 21.9%, EFF 11.5%, MK 8.4%, while the IFP polled at 4.4%, ActionSA 3.4% and FF+ 1.8%.
An eNCA election poll, conducted by MarkData, accurately put the DA at 21% and also got the 58% voter turnout correct.
Elections analyst Wayne Sussman, who was involved in the MarkData poll, told Business Day: “The poll was spot on [for] MK, but we must always remember that polls are snapshots and not predictions of what’s going to happen.
“The SRF were very good with their daily tracker; they caught MK coming in at the end. These are sophisticated organisations with various methodologies.”
Sussman said polling organisations were more reliable indicators of what would transpire than “what was being said on social media”.
“In this election, the by-election trends told us something was coming. I hope there are more polls going forward. This is a victory for polling,” he said.
Ipsos director Mari Harris said Ipsos got its projections as accurate as possible through “tried and tested scientific methodologies and thorough political knowledge and insight”.
Elections analyst Paul Berkowitz, director of The Third Republic, a nonprofit focused on deepening democracy in SA, said the eNCA/MarkData poll was “very accurate”.
“The Ipsos one was good [too] except for MK; they got the EFF and MK [projections] wrong. The SRF poll got better towards the end.”
Berkowitz noted that Ipsos and eNCA/MarkData conducted face-to-face interviews with the people, while the SRF interviews were conducted telephonically.
“Ipsos has traditionally been very good. They said they expected turnout in the high 50s, they were accurate. They were good at modelling the turnout,” Berkowitz said.
SRF associate Gabriel Makin told Business Day: “One of the interesting phenomena to come out of this election was that polls with varying methodologies (face-to-face versus telephonic) provided very similar results. Face-to-face polls undercounted the DA more than telephonic (particularly the MarkData/eNCA poll) but otherwise there were pretty similar results across the different methodologies. So for that reason I would say that wasn’t a defining feature of the polls’ accuracies. Sampling method was essentially the same across the board and so we can discount that too.”
Makin said this year there were a “huge number of high-quality, big surveys that all produced results relatively in line with the election ... While there were less in the lead-up to 2019, it was still the case that the polls done about a month out from the election were pretty accurate last election as well.
“It wasn’t anything special this time around: every polling organisation stuck to its guns and focused on doing research, as we would do in any other year.
“Polls are really only a reflection of voter sentiment and so people were honest with us when we surveyed them and by simply distilling those results and modelling them for turnout we got as close as we could to the election results.”
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.