What the Ipsos Johannesburg election poll found
The most recent Johannesburg election poll puts the African National Congress narrowly ahead of the Democratic Alliance in the city, with the ANC on 31% and the DA on 27%. Ipsos released the figures on Friday, 4 September 2026, two months before South Africans vote in local government elections on 4 November.
For a Johannesburg voter, the practical point is that this result and the one published a month earlier cannot both be close to right. A July survey by the Social Research Foundation put the DA on 42% and the ANC on 18% in the same city.
The Ipsos numbers come from its Pulse of the People survey, conducted between June and July 2026, with 3,600 people interviewed across the metros. The figures are close to the 2021 municipal result in Johannesburg, when the ANC took 33.6% and the DA 26.5%.
Ipsos found no metro heading for a clear single-party majority except one. “There is no single metropolitan story unfolding across South Africa,” said Mari Harris, political analyst and director of public affairs at Ipsos in South Africa. The survey put the DA ahead in Tshwane on 31% to the ANC’s 22%, and the MK Party on 55% in eThekwini.
Nationally, Ipsos put the ANC on 31%, the DA on 25%, the MK Party on 14%, the EFF on 11%, ActionSA on 6%, the IFP on 3% and the Patriotic Alliance on 2%.
The ANC in Johannesburg treated the result as vindication. “The Ipsos research findings reaffirm that the ANC is arresting the decline,” said regional secretary Sasa Manganye, who added that the party should not become complacent.
The other poll said almost the opposite
The July survey was conducted by Victory Research together with the Social Research Foundation and The Common Sense. It covered 504 registered voters in Johannesburg between 8 and 31 July, with a stated margin of error of 4.4 percentage points at 95% confidence.
It put the DA on 42%, the ANC on 18%, the MK Party on 13%, ActionSA on 10% and the EFF on 8%. That is a 24-point DA lead. Ipsos records a four-point ANC lead. The two results are roughly 28 points apart on the same contest.
The obvious explanation, that the ANC recovered over August, does not fit the dates. Ipsos was in the field from June into July, and the SRF poll ran from 8 to 31 July. The two surveys overlapped.
The SRF series has also moved sharply within itself. Its March Johannesburg survey put the ANC on 30% and the DA on 39%, so its own reading of ANC support fell 12 points in four months.
Manganye questioned the SRF work directly, noting its sample of 504 and its turnout assumption, and said the ANC “cautions against the deliberate use of polling as a political propaganda instrument”. He described the SRF as DA-led.
One disclosed fact sits behind that argument. Gareth van Onselen, chief executive of Victory Research, worked for the DA for about 12 years before entering polling. That is a matter of record rather than evidence that the poll is wrong, and the SRF publishes its sample size, fieldwork dates and margin of error, which Ipsos has not done for its Johannesburg subsample.
Where the two surveys actually differ
| Ipsos | SRF, Victory Research and The Common Sense | |
|---|---|---|
| ANC in Johannesburg | 31% | 18% |
| DA in Johannesburg | 27% in most reports, 28% in some | 42% |
| MK Party | Not published for Johannesburg | 13% |
| ActionSA | Double digits, figure not published | 10% |
| EFF | Not published for Johannesburg | 8% |
| Fieldwork | June to July 2026 | 8 to 31 July 2026 |
| Sample | 3,600 across the metros, Johannesburg subsample not published | 504 registered voters in Johannesburg |
| Method | Pulse of the People omnibus survey | Telephone survey, largely urban |
| Turnout treatment | Turnout flagged separately, not modelled into headline figures | 53% turnout model applied |
| Margin of error | Not published for Johannesburg | 4.4 percentage points at 95% confidence |
| Released | 4 September 2026 | Early August 2026 |
The turnout row is the one that does most of the work. A registered-voter figure asks who people support. A likely-voter figure asks who will actually vote, and in South African local elections those are different populations.
Harris made the same point in June, saying the DA relies on its voters being more enthused than average in local government elections and turning out in disproportionate numbers, which gives the party a bump. A model that builds that bump in will produce a higher DA number than one that does not.
Sample size matters too. A 504-person survey with a 4.4-point margin of error carries a wider range than a national study, though the Johannesburg component of the Ipsos survey has not been separately published, so a like-for-like precision comparison is not possible.
What neither poll can tell you
Both organisations have said their figures are not predictions. Harris cautioned that there is still substantial campaigning to do and many voters to persuade.
Ipsos found that in Ekurhuleni, eThekwini, Tshwane and Buffalo City, roughly a fifth to a quarter of registered voters currently say they do not intend to vote at all, and said mobilising even part of that group could shift outcomes in closely contested municipalities. “Party contestation over the next two months in almost every province will be fierce,” Harris said.
Johannesburg has been governed by shifting coalitions since 2021, and neither survey shows any party close to an outright majority in the city. The DA is contesting under Helen Zille as its mayoral candidate, ActionSA under leader Herman Mashaba, and the EFF under Dr Tlaleng Mofokeng.
The issue driving the contest is service delivery. Southafriworld has reported on the national assessment finding that 47% of municipal water systems are in a critical state, up from 39% three years earlier.
What remains unresolved is straightforward. The DA’s Ipsos figure is reported as 27% by some outlets and 28% by others, which changes the ANC’s lead from four points to three. Ipsos has not published a Johannesburg sample size or margin of error. And no poll settles turnout, which is decided on 4 November.
