Is the ANC doing as badly as the Ipsos poll suggests? Kinda. Sorta. Maybe

15 Aug 2022

SA is abominably served by its polling. It’s just awful and we have to do better than this. Polling is skewed anyway by the fact that when you ask people questions, they fib. They just do.

This weekend, polling company Ipsos set the cat among the pigeons by publishing a poll that appears to show that the ANC currently has about 42% support. This got the ANC very upset, the other parties prematurely happy, and political analysts very upset. Understandably.

Polling is a fascinating topic and there is a link here to how markets work, but we’ll get to that in a second. The big problem with this poll is that Ipsos, as it always does, simply asked people: “If the election were to happen today, who would you vote for?” Then it tabulated the results.

Sounds reasonable, right? The problem is that in this particular poll a huge proportion of people refused to answer, said they were not registered or said they would not vote. That took 26% of the respondents out of the equation.

What Ipsos needs is an election turnout algorithm that allows it to predict for whom the group that refused to answer would ultimately vote. And, of course, that’s not a simple thing because it would shift over time.

What we can do in the meantime is look at the proportions of the vote the parties get if we exclude the “not registered”, the “don’t know” and the “will not vote” respondents. I did a thumbnail version of this calculation and what I got is that the ANC goes up to 56%, the DA from 11% to 14.6%, the EFF from 9% to 12% and ActionSA from 3% to 4%, with the rest taking the remaining 13%. Those begin to look closer to the numbers we would expect in an actual election.

Read more on Daily Maverick: ANC's collapse as South Africa's majority party is foretold in new poll

Let’s compare that with the previous Ipsos poll taken before the local government elections in November 2021, just as a methodological check. The poll was produced, in roughly the same format, in September 2021 — not that long ago. At that point, according to the Ipsos methodology, the ANC was tracking at 50%, the DA at 17.9%, the EFF at 14.5% and the rest were under 2%. At that point, the imponderables were much, much lower, at 7.4% of the tally.

Exclude them from the proportional figures, and recalculate, and the results of that poll and the subsequent one start looking similar. The ANC jumps to 53%, the DA to 19%, the EFF to 15.5% and the rest take around 13%. Notice how close these numbers are to the reconfigured poll Ipsos has just released.

Visit Daily Maverick's home page for more news, analysis and investigations

So how did those numbers compare with the actual results of the election? Well, sadly, even the reconfigured numbers are way, way off. I mean really badly off, and in directions that are very difficult to fathom. The ANC got 45.6% of the vote, the DA 22% and the EFF 10.3%, just to look at the top three. In other words, the poll substantially overestimated the ANC and the EFF, which it got wrong by 50%! And the DA was mis-polled the other way, getting more than the poll suggested, but not by a huge proportion.

How does one read these figures with this gloss? Honestly, it is very difficult to draw any meaningful conclusions. For the record, here is mine; by all means disagree, for far from being in a terrible state, the ANC might actually be holding its much-reduced position. And that makes sense if you think about it, since Covid-19 is largely behind us, and the economy is holding, more or less. The DA really needs to worry because, even given the poll’s historical underestimation, this is a big drop. That too makes sense, because the DA’s perceived rejection of the goal of winning the black vote would naturally mean a fall-off. The EFF is also in trouble, on these numbers.

The winners are the small parties, which have all grown. The exact proportions might be in doubt, but one thing is not: this is the winter of our discontent. Party support has spread, and the number of people who say they are not going to vote or refuse to say has increased substantially, suggesting voters are looking for options and not finding them. Not surprisingly, another part of the Ipsos poll found that very specifically.

There is one more thing that needs to be said. SA is abominably served by its polling. It’s just awful and we have to do better than this. Polling is skewed anyway by the fact that when you ask people questions, they fib. They just do.

You can see that in these results. My guess is that DA supporters don’t like telling pollsters who they are voting for, which accounts for the undercount. And voters do like saying very volubly that they support the EFF, but when they actually cast their vote, they go for the more staid option. People tend to respond to polling questions not only with what they think but with what they think the pollster would like them to think, or what their friends would like them to think.

And if you think about it, this is what makes financial markets so powerful: when you buy shares, you are awarded only to the extent you are actually correct; how you look to outsiders doesn’t matter a smidgen. DM/BM

Categories:

Tags:

DM is much better if you are signed in...

Business Maverick