Poll Shows South Africans United in Hope Behind The GNU

Most South Africans believe the Government of National Unity (GNU) will succeed. This data comes from South Africa’s Social Research Foundation*, which rose to prominence after it called South Africa’s groundbreaking 2024 national election result to within a percentage point.

Eighteen months after the ensuing GNU was formed, the Foundation reports that majorities of all voters, African National Congress (ANC) voters, and Democratic Alliance (DA) voters expect that it will succeed in righting South Africa.

According to that data, gathered during November, 52% of South Africans said the GNU would succeed, with 32% thinking it would fail.

In other words, roughly 5-in-10 South Africans think the GNU will work to make the country a better place, whilst 3-in-10 doubt it will work. In addition to reading the data as a raw percentage, it is also important to keep the 5-to-3 split in mind – as that is a very strong indicator of the balance of public opinion.

Amongst black voters the data split 52% to 30% in favour of the GNU succeeding. For coloured voters the data split 52% to 37%, for Indian South Africans it split 58% to 22%, and for whites it split 47% to 38%.

For ANC supporters the data split 56% to 25% in favour of the GNU succeeding. For DA supporters it split 61% to 24%. For the Economic Freedom Fighters, the split was 56% to 40%, but for uMkhonto weSizwe, a rare exception to the national trend, the data split 30% to 47% – and therefore against the GNU succeeding.

To take full advantage of the 5-to-3 split, and to strengthen it, the GNU will need to start securing material improvements in the lives of people. Progress has been recorded under the GNU in areas ranging from port and rail reforms, to getting off the grey-list, to the currency strengthening, to a credit ratings upgrade. That is all very important and improves the living standards of people and investor sentiment on South Africa.

But the progress needs to go beyond that to hard material improvements in job numbers and living-standards numbers.

When the ANC first came to power in 1994 it did very well at exactly that. Between that year and 2008 the ANC basically doubled the number of people in employment, quadrupled the number receiving welfare payments, cut the share of families without electricity from nearly 50% to below 20%, and presided over the building of almost 10 formal houses for every one shack being erected.

As a consequence of that record, the ANC saw its support lift strongly between 1994 and the 2004 election. As its support lifted, the share of all protest actions in the country that turned to violence was more than halved.

There is little doubt that popular confidence in the GNU will reflect the same patterns. If the 5-to-3 split is to hold, and to improve, then the hard numbers on jobs and living standards must begin to lift, as they did after 1994. Should that not happen, then there is a risk that the current levels of confidence in the GNU may prove difficult to hold.

Pierneef was one of South Africa's greatest artists, known for his paintings of South African vistas. This column named after him aims to do something similar - sketch the broad vistas of South Africa's domestic landscape.

*The Social Research Foundation’s Q4 2025 Market Survey was commissioned by the Foundation and conducted by Victory Research among 1 002 registered voters between 27 October and 14 November 2025 using telephonic CATI interviews. 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, with national sim card penetration exceeding 250%, more than 90% of adults owning a phone, and mobile networks covering 99.8% of the population, giving universal practical coverage. The sample was fully weighted to match the national registered voter population 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 52.8%. The poll carries a 4.0% margin of error at a 95% confidence level, with a design effect of 1.762.