The GNU Pulse Survey: South Africa's Coalition Reality

Catching the headlines, not the details

Half of the panel (50%) said they follow GNU news only at headline level, while a further 22.1% know it exists but not the details. Just 20.8% follow political developments regularly. This is a public that is aware but not obsessive — the coalition is background noise to many, a daily anxiety to a few. The emotional driver here is convenience-seeking: politics competes with the cost of living, and most people ration their attention.

Confidence: a lukewarm 61.6 out of 100

Asked to rate their confidence in the GNU's ability to solve South Africa's problems on a 0–100 scale, the mean response was 61.6, with a median of 59. The single most common score was 50 (18.2%) — the exact midpoint, the answer of someone who has not made up their mind. A hopeful experiment — it could work if they cooperate captured the largest share of feeling (37%), but 18.8% called the GNU an elite pact that serves politicians, not ordinary people, and 9.7% an unholy alliance doomed to fail. The panel is not uniformly cynical; it is cautiously, conditionally open.

The wish list: jobs first, delivery second

Asked for the single biggest problem the GNU should address, 47.4% named unemployment and job creation — more than triple the next issue, crime and safety (12.3%). Corruption and government accountability followed at 9.7%. When asked what the GNU should do differently, the top answer was focus on service delivery instead of internal power struggles (23.4%), closely followed by prioritise economic growth over political deals (21.4%) and listen more to ordinary citizens (20.1%). The through-line is value-consciousness: people want the coalition to stop negotiating and start delivering.

Trust in leadership: a nation without confidence

Asked who should have more power in the GNU, the largest share — 35.1% — said a mix of all parties equally, while 24.7% said ordinary citizens, not politicians. Only 20.1% backed the ANC as the largest party, and 10.4% the DA. When asked whether the GNU should include all major parties or slim down, 34.4% wanted to reduce the number of parties to only the biggest ones, while 24% said keep everyone in. The panel is not demanding a single-party state; it is asking for a leaner, more accountable coalition. The dominant emotion is skepticism — not of democracy, but of the people running it.

News: TV leads, but trust is thin

TV news (SABC, eNCA, Newzroom) is the primary source for 27.3% of respondents, followed by Facebook (22.1%) and TikTok (14.9%). WhatsApp groups account for 7.8% — a meaningful channel, but not the dominant one. Trust in news is fragile: only 19.5% say they trust most of what they see, while 28.6% trust some sources and not others and 24.7% double-check before believing. A further 18.2% trust no political news at all. On social media specifically, trust in information about the GNU averages 62.4 out of 100 — slightly higher than trust in the GNU itself. The panel is not switching off; it is fact-checking as it scrolls.

The future: maybe, just maybe

When asked whether the GNU will survive until the next election, 52.6% said maybe — it depends on what happens next. Only 26.6% believe it will hold together, while 12.3% expect a collapse. This is not a nation predicting doom; it is a nation holding its breath. The panel's hope is real but provisional — a hopeful experiment is the most common feeling, yet it is a hope that could tip either way.

The panel is not asking for a revolution — it is asking for the GNU to work, and it will believe it when it sees it.

The people behind these numbers are not a random sample of South Africa — they are the Dootsa panel, an organic community that grew through referrals and product-led growth, not purchased lists or paid ads. That matters because it means these are real, engaged voices, not rented respondents. Who were they? The age distribution below shows a panel weighted toward the young: 25–34 year-olds make up 43.5% of respondents and 18–24 year-olds 39%, together more than four in five of the room. This is a generation that came of age after 1994, for whom the ANC's loss of its majority is not a rupture but the only political reality they have known as adults. The gender split leans female at 55.8% against 39% male, and the province mix is led by Gauteng at 29.4%, followed by Western Cape at 15% and KwaZulu-Natal at 13.7% — a spread that reaches beyond the economic heartland into the rest of the country. Living-area composition shows 28% in townships, 24.7% in suburban areas, and 22% in urban city centres, with the remainder in semi-rural and rural settings. Household income tells a story of constraint: 36% earn under R5,000 a month, and a further 13.3% fall in the R5,000–R10,000 band, while only 20.7% earn between R20,001 and R40,000. The LSM proxy mirrors this — 37% sit in LSM 1–4, 34.9% in LSM 5–7, and 28.1% in LSM 8–10 — a broad cross-section of living standards, not a privileged slice. Employment status adds texture: 30.7% are full-time employed, but 24% are unemployed and seeking work, and 19.3% are students. These are South Africans who feel the cost-of-living squeeze directly, many without the cushion of a steady salary. This is a panel that carries the weight of the country's economic anxiety in their daily lives. When they tell us the GNU is a 'hopeful experiment' but that they don't trust leaders, they are not speaking from the sidelines — they are speaking from the queue for a job, the taxi rank, the township street. Their caution is earned, and their hope is hard-won.

When South Africans were asked where they get most of their political news, the panel did not default to the platforms that dominate the global conversation. TV news (SABC, eNCA, Newzroom) led with 27.3%, followed by Facebook at 22.1% and TikTok at 14.9%. WhatsApp groups — often cited as a vector for misinformation — drew just 7.8%, roughly the same share as YouTube. This is a public that still leans on broadcasters for the headlines, then layers social feeds on top for commentary and community reaction. The emotional driver here is trust: people are not abandoning traditional gatekeepers; they are supplementing them.

That trust, however, is conditional. When asked whether they trust the news they see about the GNU, only 19.5% said they trust most of what they see. The largest group — 28.6% — said they trust some sources and not others, while a further 24.7% double-check before believing. This suggests a skeptical, verification-minded audience that has learned to triangulate. The paradox is sharp: they choose TV for authority, yet nearly half (47.3% combining 'sometimes' and 'no') withhold full belief from any single outlet.

On social media specifically, the panel's trust score averaged 62.4 out of 100 — slightly higher than the 61.6 confidence they placed in the GNU itself. The most common score was a middling 50 (18.8%), indicating a public that neither embraces nor rejects social information outright. Among the small group who named a news website as their primary source, the follow-up showed News24 dominating the preference, though the sample was too thin to publish a precise share.

The panel is not naive — it is watchful. People are curating their own information diets, mixing legacy media with social feeds, and holding every source at arm's length. For anyone trying to reach this audience, the lesson is not about choosing one channel over another; it is about earning trust through consistency and verifiability, because South Africans are already doing the fact-checking themselves.

For a brand or public-sector reader, this panel is a rare window into how a young, digitally native South Africa is metabolising the country's biggest political story — and where their trust is running thin. The data suggests three questions worth asking.

First, where does attention actually live? Half of the panel (50%) follows GNU news only at headline level, yet TV news (SABC, eNCA, Newzroom) still leads as the primary source at 27.3%, with Facebook (22.1%) and TikTok (14.9%) close behind. For any organisation trying to reach citizens with credible information, the broadcast-to-social handoff is the battleground — not a single platform.

Second, what does trust look like when it's conditional? Only 19.5% trust most of what they see; the largest group (28.6%) trusts some sources and not others, and a further 24.7% double-check before believing. This is a public that has learned to verify — a behaviour brands must respect, not fight.

Third, what would citizens actually change? When asked what the GNU should do differently, 23.4% said focus on service delivery instead of internal power struggles, and 21.4% prioritise economic growth over political deals. The appetite is for outcomes, not optics — a signal for any brand promising tangible value in a cost-of-living crunch.

This is a panel that hopes, but verifies — and it will reward those who deliver proof over promises.

One caveat: this is a 154-person organic panel, not a census. The window (30 July–2 September 2026) captures a single moment in a fast-moving coalition story; sentiment can shift with the next cabinet shuffle. Treat these as directional signals, not a forecast — and as an invitation to ask your own question of the same South African voices.

The panel's verdict is not despair but a weary, watchful hope: 52.6% say the GNU's survival depends on what happens next, and just 12.3% expect it to collapse. South Africans are holding their breath, not writing it off.

If you take one line from this pulse check, let it be this: a nation that mostly catches the headlines still trusts the coalition more than it trusts the news about it. Half the panel follows GNU news only at headline level, yet confidence in the GNU's ability to solve problems sits at a mean of 61.6 out of 100 — higher than the 28.6% who trust only some news sources. They are sceptical of the messenger, not yet the message.

This is one paper on a living Insights desk. Sibling publications keep the South African picture current, and the same organic panel — real voices, no paid social ads, no purchased respondent lists — can answer the questions this study leaves open. What would change if the GNU's next move were measured the same way? Commission your own survey on Dootsa and find out.

This pulse check ran on the Dootsa panel from 30 July to 2 September 2026 — 35 calendar days in field — and took respondents about 4 minutes to complete. Of the 154 completed responses analysed, 150 came from registered panel members and 4 from consented guests; the completion rate was 98.1%. The panel itself is not a purchased or rented sample: people arrive through product-led growth, referrals, and field ambassadors, with no paid social advertising and no bought respondent lists. That means these are real, engaged voices — not rented respondents.

The aggregates here are directional, not a census. The sample skews young (over four in five are under 35) and leans female (55.8%), and while Gauteng leads at 29.4%, the panel spans provinces and living areas from township to suburban. We report only rolled-up, k-anonymised distributions — no open-text quotes, no small cells — and we do not cross-tabulate answers by demographic, so thin slices are suppressed rather than guessed at.

All processing is POPIA-aligned (Dootsa is the Responsible Party for platform personal information), GDPR-aware where EU data is in scope, ISO 27001-aligned in its information security management, and SAMRA-aligned in research conduct. Full methodology, the question instrument, and the field record live under Methodology & Data Collection and the study catalogue.

The survey ran for 35 calendar days (2026-07-30 to 2026-09-02). Starts are the 100% base: 157 started, 154 analysable completes (98.1%), and 3 in progress or abandoned. Completes split as 150 signed-in and 4 public (consented). Paid quota is the signed-in meter only and is not this n=. Analysable completes: 154 (150 signed-in, 4 public/consented). Paid quota is the signed-in meter only — public completes do not fill max_responses. 2 new accounts were created after completing this public survey (not extra responses). 2 public completes still have unclaimed Rocks / no account yet. Full administration, POPIA, and acquisition detail is in How we ran this study.

Completions: 150 registered panel (97.4%) and 4 consented guest (2.6%). Control-tower acquisition mix: Organic n=78 (50.6%); Referral n=55 (35.7%); Direct / untagged n=18 (11.7%); Survey join n=3 (1.9%).

Respondent demographics appear in the dedicated Respondent Demographics panel (not repeated here).

We process personal information lawfully, minimise collection, and protect respondent privacy. Public surveys collect only what is needed for research quality. We use device fingerprint hashing (not raw fingerprint data), Cloudflare Turnstile, IP rate limits (one submission per IP per survey per 24 hours), and hidden honeypot fields to filter automated submissions. Incomplete and abandoned sessions are counted in response totals for transparency. Only verified completions contribute to published insights. Respondents must confirm they are 18 years or older. Minimal optional fields (name and location) may be collected for insight context. Guest responses are screened before inclusion in creator analytics. Suspicious patterns may be flagged or excluded. Age confirmation: I confirm that I am 18 years or older and agree to Dootsa's Terms and Privacy Policy for this survey. Data use consent: I understand that my responses will be screened using anti-bot measures and may be used in aggregated research insights. Guest answers are stored encrypted at rest where required; only anonymised aggregates appear in public findings. Small demographic cells are suppressed (k-anonymity). 2 people created a Dootsa account after this public survey (counted as conversions, not extra completes). 2 public completes have not yet claimed Rocks.

Informational only

This publication and its charts are not legal, financial, medical, or professional advice. Dootsa does not warrant accuracy, completeness, or fitness for a particular purpose. You assume all risk of decisions made in reliance on this content. See Terms.

Research data

Figures are aggregate, anonymised counts from verified Dootsa survey responses between 2026-07-30 and 2026-09-02 (n≈154); values may be rounded. Closed-question shares in the narrative, charts, and appendix use the same public-safe denominator (including any “Undisclosed (small sample)” bucket). This document is for research and planning context only.

Study objective

Responses reflect panel opinions during the fieldwork window (n≈154) and should not be generalised to all South Africans without further sampling design.

Conflict of interest

The author declares no competing interests. Dootsa (Pty) Ltd conducted this research for internal publication and commercial insights purposes.

POPIA

POPIA: Updated consent forms are in place; data collection methods for this fieldwork are designed to be POPIA-compliant. Public outputs are anonymised aggregates only.

Panel and trust posture

No one knows the South African mind and heart like Dootsa — and the public Insights desk is the proof: real voices, organic growth, no purchased respondent lists. The panel is not a purchased or rented sample: no paid social advertising and no bought respondent lists. People arrive through product-led growth, referrals, and field ambassadors. Brands commission fieldwork on the same panel — they do not buy lists of people — and funded studies thank members with Rocks. Personal information is processed under POPIA (Dootsa as Responsible Party), GDPR-aware where EU personal data is in scope, with an ISO 27001-aligned ISMS. Research conduct is SAMRA-aligned. Cite sibling Insights papers for activity; this paper’s findings remain unique. See /trust. Do not invent extra certifications.

Limitations

This is a convenience / organic panel aggregate, not a probability sample of the South African population. Observed completion rate was 98.1% of starts. Treat shares as directional. Small demographic and option cells (under 10) are suppressed on public surfaces as “Undisclosed (small sample)”. “Undisclosed (small sample)” aggregates answer options (or demographic cells) with fewer than the public cell threshold of respondents under k-anonymity. It is not the same as “prefer not to say”, “Other”, or a skipped question.

License

Unless otherwise stated, public findings are licensed CC BY-NC 4.0.

Guest surveys

Guest completions used in-window anti-abuse controls (CAPTCHA, rate limits, 18+ consent). Public outputs remain anonymised aggregates. Guest fields collected are limited to what the survey disclosed at the point of participation.

Topic framing

This survey touched financially or personally sensitive constructs. Answers reflect self-reported attitudes and intent, not observed transactions, diagnoses, or verified behaviour.

Narrative framing

Narrative framing: this report describes shopping preferences, convenience, and local identity cues in comparative terms. It does not judge, stigmatise, or moralise individual consumers or communities.