If you Google “election poll” in South Africa right now, you’ll find headlines like: “DA leads ANC by 13% in Johannesburg as urban politics fragment, poll finds.” That was Business Day reporting on the latest eNCA/MarkData poll. Its headline finding is significant: the poll puts the DA on course to become the largest party in Johannesburg and retain its majority in Cape Town.

But this is just one poll.

Over the past few months, we’ve had at least four major polls trying to give us some sense of what could happen when South Africans vote in the municipal elections on 4 November. There’s been Ipsos, SRF/Common Sense, News24/SABI and, most recently, eNCA/MarkData.And they’ve generated plenty of headlines.

As someone who is deeply interested in politics, and who spends far too much of her life talking to political analysts, I find these numbers fascinating.

I want to know where the ANC is losing support. Whether the DA is growing. What MK is doing in KwaZulu-Natal. What is happening in Johannesburg. And, perhaps most importantly in our era of coalition government, where the smaller parties are sitting.

I’ve always thought polls gave us a useful snapshot of where things might be heading politically.

But the more I spoke to people about them, the more I realised that many of us don’t really understand what a political poll is supposed to do. One of the biggest misconceptions is that polls are predictions of election results. They aren’t.

Then there’s the belief that polls can somehow influence voters: if you tell people one party is winning, perhaps voters will simply follow the crowd and vote for that party.

And then there are those who believe polling is essentially a sophisticated plot designed to favour certain political parties and embarrass those projected to perform badly.

So which is it?

Leave a comment

With just over two weeks to go before we vote, my team and I decided to properly delve into the world of election polling.

I called Gareth van Onselen, a well-known political analyst and pollster, and asked him to come to Debrief Labs and explain it to us.

There is an important caveat here. Gareth’s firm, Victory Research, conducted the polling for the Social Research Foundation’s survey, and he previously worked for the DA. That context matters and you should know it. But Gareth also understands polling, the methodology behind it and, importantly, how to make sense of the numbers.

And there is a lot to make sense of. Because the four major polls don’t all ask exactly the same people the same questions in the same way. Some interview people face-to-face. Others use telephone interviews. Their sample sizes differ. Some focus on registered voters. And, crucially, they make different assumptions about who is actually going to turn up and vote on election day.

That last point matters enormously. You can ask 100 people which party they support, but if only 50 of them actually arrive at a voting station, the election is decided by those 50. That is one reason two credible polls can look at the same city and produce surprisingly different numbers.

And there are some fascinating differences this time around. The broad trends are fairly clear. The polls point to continued pressure on the ANC’s support. The DA has opportunities in several metros. MK appears particularly strong in eThekwini.

But once you dig beneath those broad trends, things get much murkier. How big could the DA’s lead be in Johannesburg? Could MK actually win an outright majority in eThekwini, or will it need coalition partners? What happens in Tshwane? And which of the smaller parties could end up holding the balance of power?

Those questions matter because South Africa’s big metros are increasingly fragmented. Coming first does not necessarily mean governing.

A party can win the largest share of the vote and still find itself sitting in opposition if other parties manage to put together a coalition with more than 50% of the seats.

And this is where I think polls become useful. Not because they can tell us what will happen on 4 November. They can’t. But they can show us where the political battlefield is.

In our latest episode of Let’s Debrief, Gareth takes us through how these polls are constructed, why their results can differ so dramatically and what we should and should not read into them.

Remember, ultimately there is only one poll that matters. The election itself.

On 4 November, voters get to give us the real numbers.

So perhaps that’s the best way to think about political polling: a poll doesn’t tell us what is going to happen. It helps us understand what could happen and why.

Thanks for reading! This post is public so feel free to share it.

So, can we actually trust election polls?