This article is free for everyone to read.

New polling* from The Common Sense and the Social Research Foundation shows the rise in Geordin Hill-Lewis's favourable rating since February last year has come from white and coloured voters. Among white voters the share with a favourable view of the Democratic Alliance (DA) leader has gone from 26% to 44% to 49% across the three rounds in which he was asked about. Among coloured voters it has gone from 18% to 25% to 39%. Among black voters it has not moved, at 2%, 3% and 3%. Half of white voters and four in ten coloured voters now view Hill-Lewis favourably; among black voters the figure is one in thirty. The chart below shows the share of white, coloured and black voters with a favourable view of Hill-Lewis at each of the three rounds.

* The figures come from three Social Research Foundation Market Surveys conducted in conjunction with The Common Sense and carried out by Victory Research using telephonic CATI interviews. A single frame random digit dialling design was used in each round, with respondents screened for voter registration and then interviewed, and the sample was further quota-controlled for key demographic characteristics of South Africa's registered voter population. National SIM card penetration exceeds 250.0%, making telephonic interviewing a reliable method of reaching a representative sample of registered voters. Each sample was fully weighted according to the demographic parameters of the registered voter population. Turnout modelling assigned each respondent a probability of voting in the next election, with the primary turnout model set at 56.0%. Figures in this piece are for white, coloured and black voters, subgroups of each round's registered-voter sample; the margin for a subgroup is wider than the national margin. Each round carries a national margin of error of 4.0% at a 95.0% confidence level.

The rounds were: 4 to 19 February 2025, 1 004 respondents, design effect 2.0127; 20 February to 6 March 2026, 2 222 respondents, design effect 3.2675; and 8 to 31 July 2026, 2 214 respondents, design effect 3.7938.