Election predictions do not stand up to scrutiny

Rhodes>Perspective>2014 Archive

If numbers can’t bring realism to election predictions, perhaps common sense might.

A weekend newspaper report, which breathlessly revealed that the Democratic Alliance (DA) and the Economic Freedom Fighters (EFF) had informally discussed a coalition in Gauteng, typifies much of the hype that has surrounded this election.

Media reports, some polls and back-of-the-envelope guesses by analysts create the impression of an African National Congress (ANC) under severe pressure from the DA and the EFF, and so on the brink of a huge setback. (Agang SA was also meant to eat into the ANC vote but these claims have subsided). A look at what evidence we have suggests that these claims rest on very flimsy grounds.

Hard evidence of how the parties will fare is thin on the ground. The opinion polls that are publicly available have been inaccurate in the past, so we cannot rely on them.

In previous elections, predictions based on (claimed) attendance at rallies have been wide of the mark.

Municipal by-election results are far more reliable, but most voters can’t vote in them and those who can often don’t bother.

But we can still use what information we have and some common sense to get an idea of how things are likely to turn out. This method doesn’t say who will get what — but it does help separate reality from fantasy. Here is how the logic works.

A strong reason for wondering why DA leader Helen Zille still says the DA can win Gauteng is that this would contradict what she has been saying about the election. She has said that she expects the ANC to win 60%. Then, in an exchange with this newspaper’s editor, she retreated from the DA’s 30% target, implying that it would get less, although she didn’t say how much less.

Of course, Zille could be talking down her party’s prospects so that a moderate result will look good. But then why does she continue to insist that it is in the running to win Gauteng? And talking down your party’s prospects is not how the political game is played here. Another reason for taking her seriously is that the larger parties tend to commission competent pollsters and she was no doubt relying on the DA’s polls.

So we can assume the ANC will get at least 60%. This is probably an underestimate if it was based on polling a few months before the election — in all recent elections, competent polls conducted a couple of months before have underestimated the ANC vote. This is so because some voters, who traditionally support it but are unhappy with it, insist months before the poll that they will vote against it but then give it another chance.

But, for the sake of argument, let us stick with the low figure and assume that the ANC will get 60%, losing about 9% of the votes it won in 2009.

The DA won 24% in the 2011 local elections. Since its vote has increased in every election, we can assume that it will gain. But, if its leader says it won’t get 30%, a generous estimate gives it, say, 28%.

It is hard to see how any of this means an ANC loss in Gauteng. It received 64% in the province last time and would have to drop by more than 14 percentage points. So it must lose 22% of its Gauteng votes — double the loss Zille predicts for it countrywide. There seems no reason why it should fare doubly as badly in Gauteng as Zille thinks it will do everywhere else.

What about the other "coalition partner", the EFF? Or Agang? We have given the ANC and DA combined 88%. To that we need to add the two parties that appeal to the KwaZulu-Natal heartland, the Inkatha Freedom Party (IFP) and the National Freedom Party (NFP). In the local elections, they received 6% between them, an increase on the IFP’s 2009 performance (presumably because the NFP brought waverers back into the fold). They might well repeat this in this election but, again for the sake of argument, let us assume that they lose as much as one-third of their vote and come in at 4%.

To that, we need to add the African Christian Democratic Party, Freedom Front Plus, the Congress of the People (COPE) and the United Democratic Movement: all have fallen on hard times recently but all retain enough loyalty to win at least a couple of seats. On the basis of recent results, they are surely entitled to expect at least on average 0.5% each. With the two KwaZulu-Natal parties, this pushes the vote accounted for to 94%. And this assumes that parties such as the Azanian People’s Organisation and the Minority Front don’t even appear on the radar screen.

So, even if we make very generous assumptions, the EFF and Agang are left with 6% of the vote. While the lion’s share would go to the EFF, it would still underperform the previous ANC breakaway, COPE. This logic does not tell us exactly how the parties will do. But what it does say is that much of what we are being told about this election does not make sense.

Friedman is director of the Centre for the Study of Democracy.


Article Source: Business Day