Why are Liberal polls so consistently overoptimistic?

Most of the work can be attributed to Gareth van Onselen, whose company is rather sophisticated. But in recent years, the opinion polls have repeatedly shown 10% shifts which disappear as the ballot closes.

The Common Sense, using data from the Social Research Foundation (SRF), has showed some stunningly optimistic results for Helen Zille’s mayoralty bid for 2026. She is wiping the floor with the opposition, and has more support even among her opposing parties’ supporters than her opponents do.

I have no doubt she will perform well, but these seem a little too good to be true. Let me state my prediction here, just to keep me honest on this question in the future – the DA will gain a few points, perhaps even achieving their peak 2016 result again, and may even get their favoured coalition with the ANC in Joburg. But after that, I think, the effects on government quality will be inconclusive, and people will go back to the old voting patterns.

In the West, polls are often used as a means for manipulating the public sentiment rather than measuring it. But South Africa has largely been immune to this, at least so far. You can meme the vote as hard as you like, but people are going to vote the same way they always have.
:

And in fact, until recent years, polls have tended to avoid optimistic thinking, and have been pretty much the same as the eventual outcomes, with little deviation from the norm; maybe a 2-3% underestimation on ANC support. But looking back to the year before the last election, we see the polls do something very very odd. In 2023 and 2024, polls for the DA were extremely high, beyond anything they ever received before. And yet when the ballot was finally counted, the DA achieved just the same results they got for every election before – a hair above 20%.

In November of 2023, the SRF gave them 31%, Brenthurst gave them 25-27%, and the IRR gave them 28%. The IRR put them in a neat little graph, alongside the far more realistic 20% result from IPSOS. The SRF got another 31% polling result in April 2024, just a month out from the election.

What’s rather peculiar is that until 2023, the most optimistic results any poll by a South African company got for the DA was 24%, in 2019. That includes the 2016 election, where the DA got a record 26.9%.

IPSOS, a French firm, predicted the 2016 result correctly, with a sample size of 3 681 people. They also, as the graph above shows, got the last election result right – scoring the DA about a point below their actual result with a year to go in both cases.

So why are South African pollsters so far off? Why so optimistic? It isn’t only the vote that is getting distorted these days. Regarding racially divisive policies, you will hear every man and hid dog say that noboy supports EWC or BEE. They tout big poll results which show “jobs” as the highest priority, and a belief in meritocracy in some form. Yet Atlas Intel, a Brazilian firm, blithely asked the population rather directly what they thought of BEE, EWC and Cape independence, and got the following results, which are unsurprising to anyone familiar with the sizes of our populations groups:.

It also made me recall a little question from an IRR poll in 2017:

Gareth van Onselen managed to avoid asking this question of the population for the next six years, until he was commissioned by the CIAG to see whether the Cape population was into secession. 69% of black people responded that they do not believe the country belongs equally to all who live here – only 31% believed in moral racial equality.

But even the remaining third of the black population who believe in the moral equality of the races will not vote against black nationalism. As we wrote about earlier, Gareth van Onselen’s public stats on racial support quietly elide the 4% rate of black support for the DA, riding on the fact that most people are too lazy to do the maths, or to realise that “14% of DA votes are black” is a different fraction than “4% of black votes go to the DA”.

Curiously, it seems the SRF, whose results are so bizarrely optimistic, is still being run by the backbone of Gareth van Onselen’s Victory Research. In fact, so were most of the IRR’s polls. Now just as curiously, despite being wildly off-target right until a month before the election date, giving figures nobody else was giving, Victory Research’s polls zeroed in on the most accurate exit poll figures of any professional polling organisation on the planet (even the foreign ones who took the pulse correctly up to a year before the date) just in time for the vote to line up with it.

To maintain one’s professional reputation, one must always be able to get the exit polls right, and it may be that he revised his methods as the big date approached. It may also be that people just changed their minds last-minute and reverted to the mean. But it does seem very strange that while others who called the basic vibe well ahead of time, van Onselen was still overestimating the DA by 10 whole points.

The real question is, is Gareth van Onselen committed enough to his ideology that he would try to use the polls, not to find out what public sentiment is, but to manipulate it? Would he consider it worth the punt? It’s risky – these 10-point deviations are very strange, and more people remember the exaggerated optimism than those who recall the corrections in the days leading up to the final exit poll.

But equally, is South Africa ready to fall for it, or will it backfire?

I recall finding out from DA representatives that, knowing only 4% of black people vote DA, they stopped campaigning to grow that figure, since whenever they go into certain areas, it just drives ANC and EFF turnout. The thought occurs to me that learning that many many people support Helen Zille will not just have a stimulating effect on minority voters (who do genuinely support her), but also have the opposite effect on black nationalists concerned about losing their city to the woman so many regard as the Evil White Madam of Cape Town. Those who do not vote are a lot blacker on average than those who do.

Despite detailing all sorts of wonderful facts about the ethnic support base of each party, he refuses, under any circumstance, to discuss the existence of that 4% figure I mentioned above. He is a clear supporter of the DA, relentlessly angry and outraged at Solidariteit, and dripping with delight any time something bad happens to their image or support base, putting him very clearly into the mould of a sort of fundamentalist about the national project; the sort who would bend the rules to keep the GNU from falling apart and proving the pessimists right.