Poorest South Africans Turn Away From ANC, New Poll Shows

Polling data released in August 2026 puts African National Congress support at 36% among black voters nationally, a figure that cuts against the assumption that the party’s troubles are concentrated among middle-class defectors. The numbers tell a different story, one rooted in the townships, rural homesteads, and poorest households that have long been the ANC’s most reliable ground.

The survey, conducted between July 8 and 31 by The Common Sense and the Social Research Foundation, reached 2,214 registered voters across South Africa through telephonic interviews. Researchers asked each respondent which party they would back if an election were held immediately. What came back was not a picture of affluent urban drift. It was something broader.

In township areas, where the ANC has historically commanded wide margins, support fell to 32%. Among voters earning less than R2,000 a month, the party held 49%. In traditional rural areas, it registered 39%. These are the communities that have delivered the ANC its electoral foundation across multiple cycles. The contraction in each of them points to a shift that goes well beyond the suburbs.

By contrast, recent political commentary has tended to frame the ANC’s difficulties as a middle-class phenomenon, a story of educated, urban voters moving toward alternatives. These figures complicate that framing considerably.

The methodology behind the poll was designed to reach as wide a cross-section of the electorate as possible. Researchers used single-frame random digit dialling, drawing from all possible South African mobile numbers to give every number an equal chance of selection. The approach is grounded in the country’s mobile infrastructure: more than 90% of adults own phones, and networks cover 99.8% of the population. Samples were then weighted to reflect registered voter populations across language, age, race, gender, education, income, and geographic location.

Turnout modelling assigned each respondent a probability of voting based on their stated likelihood of participation, with the primary model set at 56%. National results carry a margin of error of 2.1% at a 95% confidence level. Metro-level samples were drawn separately, with 504 respondents in Johannesburg, 508 in Tshwane, and 500 in eThekwini, each carrying a margin of error of 4.4%. The survey’s design effect of 3.7938 reflects the layered complexity of that sampling structure, allowing for constituency-level analysis alongside the national figures.

The picture that emerges is of a party facing pressure across demographic segments and geographic areas that it could once count on with confidence. Whether the ANC can reverse those trends before the next electoral test, and among which communities it chooses to concentrate its efforts, may define the shape of South African politics in the years ahead.