This is a paid article which your subscription is allowing you to read.

We asked black voters, township residents, the poorest voters, and rural voters who they would back today. For months, the story has been about the ANC bleeding support among the middle classes. New polling from The Common Sense and the Social Research Foundation* shows the losses run far deeper than that, and much closer to the party's traditional base than it would like to admit.

Polling conducted by The Common Sense and the Social Research Foundation* between 8 and 31 July, asking respondents which party they would vote for if an election were held "today," found the ANC on 36% among black voters, 32% among township voters, 49% among voters earning less than R2 000 a month, and 39% among registered voters in traditional rural areas.

* The Social Research Foundation’s Q2 2026 Market Survey was conducted in conjunction with The Common Sense and carried out by Victory Research among 2 214 registered voters between 8 and 31 July 2026 using telephonic CATI interviews. National results are based on the full sample of 2 214 respondents. Metro results are based on samples of 504 respondents in Johannesburg, 508 in Tshwane, and 500 in eThekwini. A single frame random digit dialling design was used, drawing from all possible South African mobile numbers to ensure that every number had an equal probability of selection. National SIM card penetration exceeds 250.0%, more than 90.0% of adults own a phone, and mobile networks cover 99.8% of the population, providing universal practical coverage. The samples were fully weighted to match the relevant registered voter populations across all key demographics, including language, age, race, gender, education, income, and urban or rural location. Turnout modelling assigned each respondent a probability of voting based on questions measuring their likelihood of participation, with the primary turnout model set at 56.0%. The national results carry a margin of error of 2.1% at a 95.0% confidence level, while the Johannesburg, Tshwane, and eThekwini results each carry a margin of error of 4.4%. The survey has a design effect of 3.7938.