ElectionsXAwaiting results
← All election insightsELECTIONSX RESEARCH / 2026

Can you trust an election poll?

Before the 4 November local elections, Ipsos asked a few hundred people in each city how they would vote. Ask a different few hundred and the answer changes. Each dot below is one re-run of that poll.

See how all the metros compare

How the metros compare

The latest Ipsos polling figures and their sampling margins. These are survey responses, not counted votes.

Who each poll found, city by city

Each bar is 100% of voters who named a party in that poll.

Johannesburg
Tshwane
Cape Town
eThekwini
National
  • DA
  • ANC
  • EFF
  • ActionSA
  • MK
  • FF+
  • IFP
  • PA
  • Other

Don’t-know answers are removed and shares rescaled to 100%. Colours tell parties apart and are not official branding.

Is the lead bigger than the margin?

The coloured bar is the first party’s lead. The outlined bar is the poll’s maximum margin of error. A lead shorter than the margin is a lead the poll cannot confirm.

JohannesburgDA over ANC
3.1 pts vs ±5.89Within the margin
TshwaneDA over ANC
3.1 pts vs ±7.8Within the margin
Cape TownDA over ANC
30.0 pts vs ±5.95Beyond the margin
eThekwiniMK over DA
26.5 pts vs ±6.42Beyond the margin
NationalANC over DA
2.1 pts vs ±1.95Beyond the margin

Where each party is strong

Darker cells mean more support. Blank cells mean the party was not reported separately.

PartyJohannesburgTshwaneCape TowneThekwiniNational
ANC28%32%21%10%31%
DA31%35%51%21%29%
MK5%2%3%48%12%
EFF13%8%10%4%10%
ActionSA11%8%3%2%4%
IFP2%2%8%4%
PA1%4%2%
FF+3%5%1%1%
Other5%7%7%5%7%

Test a poll yourself

Adjust the sample size to see how much the precision changes.

1006 000

In this poll DA is on 31% and ANC on 28%, a lead of 3 points.

50% (a coin toss)95% sure

75% chance that DA is really ahead

With 277 people, the poll cannot rule out that ANC is level or ahead. To be 95% sure, a poll of this lead would need about 1 643 people.

Sampling error only, assuming a simple random sample. Real polls also face turnout, undecided voters and who agrees to answer.

See one re-run, box by box

Each box is one voter in a re-run of the poll. Choose a city, then re-run the poll and watch the boxes change.

Choose a poll

Simulated re-run of the Cape Town poll (271 people). Each box is one voter.

Running the simulation

Ipsos found DA on 51% and ANC on 21% in Cape Town. The maximum margin of error for this poll is ±5.95 percentage points.

PartyPoll shareRange in 95% of re-runsThis re-run

The gold bar shows the range. The white tick marks the poll share. Don’t-know answers are removed and the other shares rescaled to total 100%, so some figures differ slightly from Ipsos’s headline numbers. Sample size is worked out from the maximum margin of error Ipsos published, assuming a simple random sample. The simulation covers sampling error only. It does not include turnout, undecided voters, late changes of mind or who agrees to answer a survey.

View the published Ipsos poll ledger ↗

Polls against results

Compare survey results with reviewed election counts only when the correct ballot and denominator are available.

Each bar is the poll’s likely range. The result marker appears once verified party-list results are published.

DA
31%
ANC
28%
EFF
13%
ActionSA
11%
MK
5%
FF+
3%
  • Poll range (maximum margin)
  • Poll share
  • Result

Poll shares exclude don’t-know answers. A result inside the range does not prove the poll was right, and one outside it does not prove it was wrong.