THE PERILS OF POLLING

Pre-election surveys aren’t predictions of the outcomes of the polls. But they do offer a snapshot of voter sentiment at a particular moment. Two new polls show some convergence, casting light on potential post-election scenarios

Voter turnout will be a key factor in the outcome of the election, with modelling from Citibank suggesting the ANC may be headed for a more favourable outcome than expected, winning 45% of the vote or more.

While scepticism has abounded over election data emerging from market research companies in recent months, two new polls show something of a convergence, casting light on more plausible scenarios.

It’s worth a word of warning: the 2016 US election is a stark reminder that sentiment expressed in polling can miss the mark. Ahead of the election, US media was awash with polling that confidently predicted that the Democratic Party’s Hillary Clinton, wife of former president Bill Clinton, would return to the White House, this time as the country’s first citizen. Most polls placed Clinton’s support at between 70% and 90%, according to the Pew Research Center. But it was not to be a shock win saw Republican Donald Trump take the presidency. Polls had missed the mark in important swing states that tilted the Electoral

College in his favour.

It’s why Victory Research

CEO Gareth van Onselen stresses that polling is not a prediction on the outcome of the election. Polls track sentiment and support around particular parties at particular moments in time.

Things can change as the election draws closer voter sentiment could shift, for example. And much depends on voter turnout the number of registered voters who arrived to cast their ballots on the day.

Turnout is tough to pin down, as Van Onselen points out. For instance, it could rain on the day, prompting people to stay home. Or specific circumstances could see people turn out in droves if, say, “their” party was under particular threat.

This week market research company Ipsos released its polling the largest in-person opinion poll on the election conducted thus far. It polled 2,545 registered voters in face-to-face rather than telephonic interviews.

Ipsos conducted the research between March 9 and April 15, after former president Jacob Zuma’s announcement that he would throw his support behind, or lead the MK Party a clear threat to the governing party’s electoral support. The poll put the ANC’s support at a paltry 40%.

However and critically Ipsos did not model turnout scenarios this time around, which it had done in a February poll, says Citibank’s Gina Schoeman. This means the poll was modelled on the assumption that 100% of those on the roll would turn up to vote.

“We know 100% turnout is impossible given that the previous national election was 65.9% the lowest in South Africa’s democracy,” Schoeman tells the FM. That’s why, modelling for turnout, Citi puts the ANC’s performance at 45%.

And, she adds, “there is very broad consensus that the ANC gets a bump in the end”.

This is due to the party’s election machinery kicking into gear. So far, the ANC’s campaign is more intense and visible than it was in 2019, when a financial crunch and, more significantly, internal division weakened the party’s campaign.

The recent rolling out of former leaders such as ex-president Thabo Mbeki has also intensified the campaign. Mbeki joined the campaign trail

for the first time in a decade, having refused to campaign for the ANC during the Zuma era.

Van Onselen agrees that there is a gradual uptick in ANC support. “It’ sa huge operation that typically climbs as we approach the election; we don’t know when it will plateau, but it usually does,” he says.

In South Africa there’s been some scepticism around polling, amid accusations that certain polling companies are tied to political parties.

One of those under fire is Van Onselen’s Victory Research, which conducts polling for the Social Research Foundation (SRF). It’s been accused of a conflict of interest: one of its founding managing partners, Johan van der Berg, is the DA’s head of research. Van Onselen brushes aside the allegations, saying

Van der Berg has no role in the operations of the company, which has clients across the globe; its success rests on the credibility it has built over the years.

In any event, the SRF has now launched a tracking tool, modelling on a low, high and medium turnout

and its assessment of opinions about political parties isn’t terribly different from that of Ipsos.

The SRF tracker is rather nifty, as it follows opinion daily, discarding the 200 oldest interviewees and adding 200 new potential voters. The drawback is that the polls are conducted telephonically.

So far, both the SRF and Ipsos polls indicate a shift in voters between the EFF and Zuma’s MK Party.

Schoeman offers a word of caution. “We do have to be careful in making the assumption that the EFF is losing votes to MK because none of us knows how undecided voters are shifting support,” she says.

It’s a salient point, and another difficult one for pollsters: allocating undecided voters. For example, the Pew Research Center partly attributed the off-base polling in the US in 2016 to a failure by pollsters to correctly place undecided voters.

Still, Van Onselen says Zuma’s MK Party seems to be making progress in KwaZulu-Natal and, surprisingly, in Gauteng.

Gauteng was the province in which Zuma’s popularity in the ANC was severely tested. In 2013, during the memorial service for South Africa’s first democratic president, Nelson Mandela, Gautengers converged at the FNB Stadium. When Zuma took to the podium, he was booed the first sign of cracks in his iron hold over the ANC and the country. It culminated in the “Zuma must fall” protests centred in Gauteng.

The SRF poll has Zuma’s party at about 10% in the province and the EFF at 16%, which could make the pair a considerable force in Gauteng if the voter sentiment holds until election day.

In Gauteng the DA continues to post its most disappointing result a paltry 25.8% was tracked by the SRF on Friday. It is a poor indicator for the party, which is seeking to grow its support from 2019’s disappointing 20% nationally. Its campaign in Gauteng has been uninspiring, though it should have good potential in the province, which is highly urbanised with a large percentage of middle-class voters.

“Gauteng is going to be a bun fight,” Van Onselen says. The province holds the potential for coalitions between the ANC, EFF and MK, but also the ANC, DA and the multiparty charter. It is, once again, the province where anything is possible, given that polling places the ANC consistently at the lower to mid-30s.

The DA in the Western Cape is being tracked at just 50.9% of the vote in the province, down from 55% in 2019 at a 60% turnout. However, in a lower turnout scenario, the DA could increase its support to about 59%.

Public scepticism aside, what opinion polls can do is ignite party loyalists into action if they feel under threat or take the wind out of their sails if they feel there is little hope of improving their lot. Polling, with all its perils, remains a useful marker of voter sentiment in an environment where political showmanship and noise muzzles the only voice that matters: that of ordinary citizens.