South Africa’s local-election polls need more than a party ranking

The headline from Ipsos’s latest Pulse of the People survey, published on 30 October 2026, is that no party commands majority support nationally ahead of South Africa’s local elections. That is consequential, but it is not the whole finding. A party ranking can make electoral uncertainty look like a contest over a few percentage points, when uncertainty about representation and trust may be just as important to understanding what voters want from local government.

Ipsos interviewed 3 600 randomly selected South Africans at home and in their home languages between 12 September and 13 October 2026. Its survey therefore offers more than an election-night forecast: it records views expressed during fieldwork. For researchers and monitoring teams, the first task is to keep those views attached to the population and question that produced them, rather than letting every result become a claim about “the vote.”

Different questions describe different publics

Ipsos reported that the share of South Africans eligible to vote who said no party represented their views rose from 47% in its first wave to 53% in the latest wave. Those are measures of perceived representation, not declarations that those respondents will abstain or reject every party on a ballot.

The distinction matters when these responses sit beside party-preference figures. Ipsos’s published chart put support at 30% for the ANC, 28% for the DA and 12% for MK among registered voters who selected a party. Ipsos said filtering to adults registered to vote produced a sample of 2,533, and that it asked registered voters to mark their preference themselves on an electronic imitation ballot. The chart’s denominator is narrower than the public whose sense of representation the survey also describes. Presenting its percentages without that condition risks making party support appear more settled than the accompanying answers suggest.

A city lead is not a city result

The temptation to overread a ranking grows at metropolitan level. Daily Maverick reported Ipsos’s Johannesburg figures as 30% for the DA and 27% for the ANC. But Ipsos gave the Johannesburg metropolitan sample a maximum margin of error of ±5.89, compared with ±1.95% for its registered-voter sample overall. A city-level lead should not be narrated with the confidence of a national finding simply because both appear in the same survey.

A careful research team would carry the city sample’s uncertainty into every chart caption, briefing and discussion of possible coalitions. It would also resist treating a measured preference as a governing arrangement already chosen by voters. Ipsos found that 65% of South Africans favoured different parties working together at municipal level, while 68% of registered voters said politicians could not be trusted. Support for cooperation and distrust of politicians can coexist; neither response identifies which coalition would work, or whether respondents would welcome its conduct once in office.

Compare polls without flattening their methods

A separate telephone survey published by the Social Research Foundation and The Common Sense on 30 October reported national support of 32% for the ANC, 26% for the DA and 12% for MK. Its figures provide a useful point of comparison, not a reason to collapse the surveys into a single definitive ranking. Ipsos used in-home interviews and an electronic imitation ballot for registered voters’ party preference; the Social Research Foundation and The Common Sense used a telephone survey. Before comparing percentages, an analyst should establish how each survey defines the population behind them and asks people to express a choice.

The practical lesson extends beyond elections. When a finding travels from a questionnaire into a public claim, its denominator, question wording, collection method and uncertainty need to travel with it. Otherwise, precise-looking figures can obscure the more consequential result: many people are not persuaded that any party speaks for them. Ipsos cautioned that its figures describe sentiment during fieldwork, not a prediction of the election result. That is not a footnote to discard. It is the boundary that makes the evidence useful.