James Myburgh: Election’24 – The Pollsters versus the Pundits

In South Africa’s national election, the ruling African National Congress faced the possibility of losing its majority, prompting speculation on coalition scenarios. Polls suggested a decline in ANC support and a surge for the new MK Party, sparking debate and scepticism among pundits. Ultimately, the election results confirmed the polls’ accuracy, highlighting the significance of independent polling in democratic processes. This article was first published by Politicsweb.

Sign up for your early morning brew of the BizNews Insider to keep you up to speed with the content that matters. The newsletter will land in your inbox at 5:30am weekdays. Register here.

Here’s a link to the 12 June Premium newsletter. If you like what you see subscribe here for your afternoon wrap.

By James Myburgh: Editor of Politicsweb

Introduction

The main plot line of South Africa’s national election was whether the ruling African National Congress would finally lose its electoral majority and, if so, by how much. The extent of its fall below the 50% mark, if it happened, would also determine the coalition choices it would be forced to make to secure the support of a majority in parliament.

One of the more nerve-wracking sub plots – at least for political insiders – was who was going to call the election correctly. On the one side were the country’s various pollsters whose surveys were pointing more-or-less to one particular outcome. On the other were a group of pundits who loudly argued that these poll results were not just wrong but nonsensical. There were strident demands made, through the course of the campaign, for a regulatory clampdown on a polling industry alleged to have gone rogue.

As the first results of the election started coming in on 30th May there was a considerable amount of stake for these two groups. Would these results finally expose the “scandalous” behaviour South Africa’s pollsters, as the pundits believed they would? Or would the results vindicate the pollsters and repudiate the pundits?

What the Pollsters were saying

A massive opinion survey conducted late last year by David Everatt on behalf of the ill-fated Change Starts Now initiative reported a substantial drop-in political support for the African National Congress. 39% of registered voters polled said they would vote ANC, versus 19% who said they would vote DA, and 15% who said they backed the EFF. 17% of respondents were undecided or said they weren’t going to vote. In KwaZulu Natal the ANC was polling at only 26% among registered voters, the DA at 15%, IFP at 15% and the EFF at 12%, with a massive 28% of respondents undecided or saying they wouldn’t vote. (An Ipsos poll conducted in around the same period came up with similar results for the ANC’s support nationally, as did one conducted by the Social Research Foundation, headed by Frans Cronje.)

Generally, those falling in the various undecided or won’t vote categories tend not to vote for the ANC but for other parties, if at all. What these results indicated then was that the ANC had lost its majority “base support” nationally and in KwaZulu-Natal and Gauteng but, with the partial exception of the EFF, none of the existing parties had been able to step in and fill the gap.

If this situation held it was likely that the ANC would claw back support through the course of the 2024 election campaign as it mobilised its activist base to go house-to-house campaigning, massively outspent its rivals, and squeezed the support of smaller parties in the run-up to election day. This was, at least, the pattern of previous national elections.

The DA knew from painful past experience that it was vulnerable to being squeezed by the racial messaging of the ANC, as it had always struggled to hold onto the support of sympathetic but skittish black voters the closer it got to the polls. The general expectation then was that the ANC would get itself back to over 45% by election day, after which it would be able to secure 50% of MPs in parliament by bringing on board some of the so-called “rats and mice” parties.

Read more: MK party challenges SA election results, alleges fraud

Zuma’s spanner

On 16th December 2023 Jacob Zuma announced that he would not campaign for the ANC but would, instead, be backing the newly founded MK Party (MKP). It was evident by early the following year that this party, once Zuma had publicly thrown his weight behind it, was going to have a serious impact on the race. This was evident in a poll by the SRF in late January and early February which found that the MK Party already enjoyed similar levels of support to the ANC in KwaZulu-Natal as it re-mobilised disaffected voters and took chunks of support off the ANC, EFF and even IFP in the province.

On 20th February President Cyril Ramaphosa announced that the election would be held on Wednesday 29th May. A Brenthurst Foundation poll conducted following this announcement, released on Friday 8th March, found that MKP’s support was now at 13% nationally, and the ANC’s at just 39%. The EFF was also being hurt badly by the sudden rise of MKP, as its support fell back to just 10% from 17% in October 2023. That Sunday, Rapport newspaper (and City Press) reported on confidential polling data, without disclosing its source, that found that MKP enjoyed the support of 28% of registered voters in KZN, ahead of the ANC at 27%, the DA at 19%, IFP at 15%, and the EFF at 6%.

In mid-March eNCA released the results of its polling, conducted by Markdata, which put the ANC at 41,4%, the DA at 20%, the EFF at 15,5% and MKP at 10,9% nationally. MKP was polling at 35,6% in KZN, 7,7% in Gauteng, and 22,6% in Mpumalanga. (These provincial results had much higher margins of error.)

These poll results meant the ANC was likely to fall far short of 50% nationally and was thus going to need the support of one of the larger parties (the DA, EFF or MKP) to form a government. If the ANC went with the EFF (or MKP) this would be “doomsday” for South Africa, according to DA leader John Steenhuisen. The obvious question for investors, currency traders, and businessmen, was which one would it be?

The Pundits push back

Such reports, and the questions that they raised, were highly unwelcome for the ANC. “A dominant party”, Maurice Duverger noted, is one “which public opinion believesto be dominant”. News reports that ANC’s support was sitting at only 40% seriously weakened that belief and undermined the ANC’s usual election message – that by going out and voting for it black voters were positioning themselves on the right side of the overwhelming majority.

At this point however there was a serious push back launched against the polls and what they were saying. The charge against them was led not by an ANC surrogate but rather ActionSA. In a statement issued on 18th March 2024 the party’s chairperson Michael Beaumont announced that he was launching a campaign not just against many of these particular poll results, but against the entire polling industry, which he said was “unregulated” and a purveyor of “disinformation”, which was being enabled by the media.

The poll mentioned in Rapport, Beaumont stated, came from the Democratic Alliance, which has long run its own sophisticated in-house polling operation, and he was going to lay a charge with the Press Ombudsman over the matter. These poll results, and similar ones from the SRF and Brenthurst Foundation, were it seemed all part of an elaborate liberal komplot to promote a “rooi gevaar” narrative by conjuring up “alarming figures of MKP and EFF support – a favoured campaign approach of the DA to animate its core support base.” Victory Research, which did the SRF’s polls, was headed by Gareth van Onselen a “former DA staffer”, while its Managing Director, Johan van den Berg, continued to serve as DA Head of Research.

Read more: 🔒 ANC faces power-sharing dilemma after election upset

The Brenthurst Foundation poll which painted “the picture of a surging MKP at 13%”, Beaumont added, was “a practical impossibility for a party whose support is almost entirely limited to KZN and parts of Mpumalanga.” These poll results could in other words simply not be true and were thus proof that the polling industry was “open to abuse that appears to be designed to influence electoral trends in South Africa”. ActionSA, he announced, “will begin the long road to fight for regulation of this industry, but we will not stand back and allow the shadowy figures in the polling industry to operate with impunity.”

On 10th April the SRF released the results of a new national poll, conducted over the previous few weeks. This had a margin of error of 2,2% and, modelled for 60% turnout, put the ANC on 38%, the DA on 25%, MKP on 14% and the EFF on 11%. The Foundation also now launched a public daily tracking poll, which ran until two days before the election itself.

The theme that something deeply sinister was afoot was taken up by the academic and writer Jonny Steinberg, who set out to expose what he believed to be scandalous connivance by the country’s pollsters to rig their results in favour of MKP and against the ANC. In his Business Day column on Friday 19th April, he asked: “Is SA getting a dose of fake news in the run-up to the election? And is this fake news coming not from Julius Malema or Jacob Zuma, but from pollsters and academics?”

The answer, he went on to imply, was “yes”. The theory advanced by Steinberg was that poll numbers were being deliberately skewed to construct a narrative that the ANC and South Africa were heading towards an existential choice after the election. Apparently, this was all with the intention of boosting the DA electorally in some way – despite the meteoric rise of MKP sapping even its support in KZN.

Read more: Ramaphosa’s leadership didn’t impress voters, but seeking consensus may be what a GNU needs

Steinberg accused Everatt of fabricating “a large, ghost party” by failing to allocate or exclude undecideds, and he said that the Brenthurst Foundation poll putting MKP at 13% did not make sense arithmetically. The SRF poll putting the ANC and MK at over 50% combined was, he insinuated, more than just a “lucky coincidence” given that it wrote the “doomsday scenario” perfectly into the numbers.

“Maybe these polls are all correct”, Steinberg continued, “and by early June I will be eating humble pie. But if the ANC comes in at, say, 45%, or MK at 6%, the credibility of polling in SA will have been horribly damaged.” He then put a racial spin on the whole matter by concluding: “White pollsters using their hard skills to skew the data to back the white party? That’s supposed to be bullshit sprouted by Jacob Zuma and Julius Malema. It isn’t meant to be true.”

In an article in the Mail & Guardian that appeared on the same day Richard Calland pressed remarkably similar claims. He said he did not buy the poll results suggesting the ANC’s support had fallen to around 40% among registered voters: “Methodological impurity is one problem; a long-standing tendency of such polls to serially under-estimate ANC support is another. Some contain deliberate bias in their set-up, presumably to trigger certain reactions in the political game.”

He dismissed the likely impact of MKP by pointing that if the party won 15% of the vote in KZN this would only take 2-3% off the ANC’s support nationally. “Zuma is a spanner in the works. Not a big spanner, but a spanner nonetheless.” It was unlikely that the ANC would end up having to choose between the DA and EFF as coalition partner as “My base case remains unmoved — the ANC will get between 45% and 49%. Far more comfortable coalition options will be available and Ramaphosa will probably remain in office. Not a great deal will change in government.”

In an article the following week for the Financial Mail on three decades of democracy in South Africa Steinberg again returned to this issue. He dismissed the reports of MKP’s surging support by stating that polling in South Africa is “something of a scandal” and he “would be surprised if the MK Party gets more than 6% or 7% of the national vote.” The reason for this was that its support was “ethnically shackled” to Zulu-speakers and so to get “anywhere near the 13% recently forecast, it would have to score a landslide majority with Zulu speakers, not just in KZN, but in Gauteng and Mpumalanga too. This seems improbable.”

In a poll released at the same time Steinberg’s article appeared Ipsos put ANC support among registered voters at 40,2%, the DA’s support at 21,9%, the EFF’s support at 11,5% and the MK Party’s support at 8,4%. This initial release was not modelled for turnout but before the 2019 elections the same base results had actually better predicted the ANC’s 57% share of the vote than the modelled ones. In a follow up press release, issued a week later, Ipsos put the ANC’s support at 44%, the DA at 20,9%, the EFF at 11%, MKP at 8,7% and the IFP at 3,3%, on a medium turnout scenario of between 57% and 59%.

Given that these results were largely consistent with the results of the other polls (particularly on the ANC’s support) this meant that if there was a conspiracy it had now widened to include Ipsos as well. In his column the following week Steinberg now laid into them too. “Polling in SA is a mess, if not a scandal”, he declared, the results of a “natural experiment” in South Africa of “what happens in a world without regulation.” The results of Everatt’s poll for Change Starts Now were “comic”, the Brenthurst Foundation’s survey was “just bad”, and the Ipsos poll was “meaningless” as the initial press release had not modelled for turnout. In any event, their results did not make any sense, so “one is left with the suspicion that there is something wrong with the data”. Having dismissed these polls out of hand, Steinberg then turned to the vexed question of who was to blame for giving them any credence at all:

Read more: 🔒 FT: Can the party that liberated South Africa still hold it together?

“The media are also to blame. The results of each poll are reported with bluff innocence…. The fault lies with editors, not with reporters…. When what is offered is obviously questionable, and we still insist on discussing it seriously, the whole country begins to look foolish. That is why polling in SA requires regulation. This is a serious country with self-respect. Its national elections are a solemn and sacred business… More than ever, those who work with numbers bear a heavy burden of responsibility.”

Such self-confident rejections of these polls certainly had an impact. They set off a flutter of moral panic in sections of the media, promoted the notion that these results could be ignored, and suggested there was no need to think about or plan for the consequences of the ANC falling to 40% of the vote on election day. Thus, Sam Mkokeli wrote in the Sunday Times that he “would be surprised if the ANC gets close to the 50% mark. My range is 45% to 48%, with a bias right now on the lower side. Also, I do not believe a single one of the polls published in the past 12 months. They have understated the ANC’s reach due to bad design and poor methodologies. Also, the media has misreported the results, failing to emphasise critical elements including the size of the polls and field dates, margin of error, and the non-allocation of undecided voters.”

Business Day also weighed in with an editorial suggesting that there was a problem, as Steinberg suggested, but without explaining exactly what it was. There was a need for “a great deal more transparency and accountability”, a “more vigilant voting public”, and “a more vigilant media.” However, it stopped short of calling for pollsters to be placed under state control suggesting rather that “self-regulation” was the better option.

News24, for the most part, did not buy into this theory – as it was familiar with the methodology of the polls, their track record, and the individuals concerned. In an article in early May Carol Paton wrote that, whatever their limitations, the various different polls had put the ANC’s base support “at around 40%”, the DA’s “at around 23%, and MK – for which there are very few results – between 8% and 14%. The EFF has been sitting at around 11%.” However, she noted that in its behind-the-scenes briefings of journalists the ANC was nonetheless “genuinely bullish about its prospects”. The party’s operatives believed that Premier Panyaza Lesufi’s populism was going to make a big difference in Gauteng, and that the MK Party would have far less of an impact than the polls suggested.

The SRF’s tracking poll, Paton further observed, suggested the ANC was on an upwards trajectory (it was at 43% at the end of April) and the “risk of an ANC-EFF coalition will diminish as the ANC edges higher. Regardless of this, though, ‘the doomsday scenario’ has a low probability of materialising.”

Graph of the SRF’s national tracking poll – on a 58% turnout

Indeed, through the first half of May it appeared that the old laws of South African electoral politics were reasserting themselves. The ANC government was keeping load shedding at bay, there were numerous state funded events and advertisements promoting the government’s achievements as part of the celebrations of 30-years of democracy, and the ruling party suddenly seemed to have plenty of money available again for its campaigning. ANC proxies in the media also put their shoulder to the wheel and started doing their bit for the party’s squeeze message by hammering the DA on racial grounds.

The SRF’s tracking polls showed the ANC’s support edging consistently upwards in the first two weeks of May, to over 45% (and the DA’s steadily downwards). This trend was thrown into reverse, however, after President Cyril Ramaphosa signed the NHI Bill into law, and the Constitutional Court stripped Jacob Zuma of his candidacy for parliament. This suggested that the threat to strip people of their medical aid cover was not the vote winner the ANC expected it to be, and that the ANC (and IEC’s) persistent lawfare against Zuma and the MKP was proving electorally counter-productive.

In its second election poll, released on 20th May 2024, eNCA put the ANC on 43,4% nationally, the DA on 18,6%, MKP on 14,4%, the EFF on 11,4%, and the IFP on 3,1%. It also found that in KwaZulu Natal MKP now enjoyed a substantial plurality of support (at 46,4%). The final SRF tracking poll – covering the period of 22 May to 27 May – put the ANC’s support at 41,6% nationally – on a 58% turnout – the DA at 23,4%, MKP at 12,1%, the EFF at 11,8% and the IFP at 2,1%.

The day before the election the ordinary reader was left with two duelling narratives. The polls stated that the ANC’s base support had dropped to around 40% and the EFF’s to around 10%, as the MK Party’s base support had surged to around 13%. The joker in the pack was the question of differential turnout. The counter to this, promoted by some of South Africa’s leading pundits, was that the pollsters had somehow all connived to cook the books and there was no way the MK Party was going to break out like this. Indeed, the idea was so preposterous that there was clearly a need for a regulatory clampdown on the whole industry.

The verdict comes in…

The answer as to who was right arrived on Thursday morning. With just 13% of voting districts declared the CSIR projected a final outcome of close to 41,65% for the ANC, 21,62% for the DA, 8,98% for the EFF and 13,86% for the MK. News24 also performed modelling on the day, with similar projections.

Once all ballots had been counted the ANC ended up with only 40,18% of the vote, the DA 21,81%, MKP 14,58%, the EFF 9,52% and the IFP 3,85%. As is evident from the IEC election map below MKP devastated ANC and EFF support in Zulu-speaking areas of the country, winning 45,93% of the vote on the national ballot in KZN, 17,25% in Mpumalanga, and 10,65% in Gauteng.

The eNCA’s poll results were closest to MKP’s support nationally and in KZN but undershot the DA’s support both nationally and in Gauteng and the Western Cape. The SRF’s tracking poll correctly called the DA’s support in Gauteng and the Western Cape but slightly overshot the number nationally (though it was still within the margin of error).

If anything, then, the final polls of the election campaign overshot the ANC’s support. Overall, the results vindicated the polls suggesting that the rise of MKP was going to be an electoral and political gamechanger for South Africa, and it left certain publications and opinion-formers with a considerable amount of egg on their faces.

Conclusion

What lessons can be drawn from this story? Proper opinion polling is an important corrective – when it comes to election time – to powerful perceptual distortions brought on both by wishful thinking (wanting ‘your side’ to do well) and the belief that past trends will inexorably continue into the present. In this case the polls were stating that the Ramaphosa ANC was going to do far worse than most would expect or could imagine. This was a particularly unwelcome prospect both for itself and its allies in the commentariat. In addition, Zuma’s MKP was going to grab an unprecedented share of the vote for a newly formed party. (The polling also pointed to ActionSA underperforming hopes and expectations.)

Rather than “updating their priors”, or adopting a wait and see approach, the pundits chose instead to buy into an in-part internet-fuelled conspiracy theory that the opinion polls were being crooked. Dots were joined, anomalies seized upon, questions “just asked”, half-truths circulated unchecked, and various political and personal connections pieced together by online obsessives.

At a certain point the adherents of this conspiracy theory simply lost sight of the fact that its entire premise was ludicrous (and libellous). It is true that those most concerned with politics tend to pay for political polling to be done but, as with a doctor or a lawyer, it is professional malpractice for a pollster not to provide accurate information to their clients. The polling firms concerned are also transparent about their methodology and do generally respond positively to what they regard as bona fide requests from journalists and analysts for clarification.

Read more: NSN: Is Zuma and MK planning another insurrection with Russian support?

When it comes to the question of motive, pollsters are in intense competition with each other for clients, they are in a race to call the election results more accurately than their rivals, and as a result they live in holy terror of misjudging voter sentiment ahead of an election. To concoct poll results, as was being insinuated, would not just be unethical but would be reputational and financial suicide as well.

The best that can be said of the Pundits who promoted this theory is that having seized the wrong end of the stick they ended up giving a painful whack to their own reputations and credibility. It should not be forgotten however that by wildely swinging this stick as they did, they risked inflicting a broader blow against the integrity and legitimacy of South Africa’s electoral democracy.

Opinion polling, along with projection modelling, is a crucial independent check both against systemic vote rigging, on the one side, and false attacks on the integrity of the election results, on the other. Properly conducted opinion polls will give you a rough but good indication of what a fair and credible result should look like. This is particularly important given the growing dysfunctionality of the Electoral Commission, which made itself glaringly evident on election day. The opinion polls showing the MKP enjoyed 10-15% support indicated that it was going to do far better than the ANC believed, but also far worse than the two-thirds majority Zuma and many MKP activists expected.

Demands for “regulation” in the particular South African context, open the door to the neutralisation of the independence of polling firms by bringing them under the control of some-or-other ANC dominated organ. Over the past thirty years our hegemonic liberation movement has invariably used such powers to bring different sectors of broader society to heel politically, not to regulate them in a neutral manner in the public interest. 

Moreover, racially charged claims that the pollsters were skewing polling data to benefit the official opposition, and so they could be confidently disregarded, have hardly done much to hinder bad faith post-election claims that the results themselves were rigged. The MKP’s spokesperson Nhlamulo Ndhlela went so far as to tell City Press that the IEC “can’t come and tell us at a national level that John Steenhuisen is more popular than Zuma or that the DA has a larger constituency than the MKP nationally. Never!”

If you help sow the wind of racial distrust then, you cannot complain when you reap the whirlwind of election denialism.

TABLES

National polls and final results

 Brenthurst FoundationIpsos – medium turnouteNCA(Markdata) SRF track (Victory Research) – 58% turnoutNational results – 58,6% turnout
DatesFeb – March9 March – 15 AprilApril-May22 May to 27 May29-May
Margin of error 1.9%1,80%2,30% 
 TelephonicFace-to-faceFace-to-faceTelephonic 
ANC394443,441,640,18
DA2720,918,623,421,81
MKP138,714,412,114,58
EFF101111,411,89,52
IFP23,33,12,13,85

Provincial polls and final results:

Gauteng
 eNCA(Markdata)SRF track –62% turnoutResults –61,94% turnout
ANC40,940,334,76
DA18,525,527,44
EFF1212,812,92
MK8,67,99,79
ActionSA2,44,74,16
PA20,52,03
Other15,68,38,9
    
KwaZulu-Natal
  eNCA(Markdata)SRF track– 60% turnoutResults– 62,3% turnout
MKP46,429,945,35
IFP14,514,318,07
DA12,218,113,36
ANC11,130,316,99
EFF6,73,52,26
Other9,13,93,97
    
Western Cape
 eNCA(Markdata)SRF track –60% turnoutResults –60,57% turnout
DA44,455,655,3
ANC29,123,419,55
EFF9,37,55,3
PA1,73,97,8
Other15,59,612,05

Read also:

This article was originally published by Politics Web and has been republished with permission.