Just before the queues started at Joubert Park, Hotazel and Bishopscourt, the polling company Ipsos released its final poll. It indicated that the ANC would get 63%, the DA 22% and the non-tie wearing beret bearers 5%. In the end, once the counting was over, Gwede Mantashe and Helen Zille had got some sleep and Malusi Gigaba had done his last interview, the final results were the ANC 62%, the DA 22% and the EFF 6%.

In other words, well within the margin of error the company itself had indicated existed.

The company’s Mari Harris says that while they were extremely pleased with their prediction results this time around (they weren’t the only ones we understand….Ed) they weren’t that far wrong in 2009 either. She says what really made it tough then was that voters kept changing their minds between the DA and Cope, particularly as the in-fighting within Cope fed into the election cycle. Picky things, voters.

This time around, it seems, it was easier, which probably indicates that not many people were choosing between the EFF and the DA. As Harris points out, it’s always easier to poll for a bigger party than a smaller one. Of course anyone who’s ever dealt with sampling will know that when it comes to smaller numbers, the margin of error starts to really increase.

One of the reasons they were able to get it so close was that we have a system of proportional representation. In other words, you do a sample, send your people out, and come back with a result. As Harris puts it, it would be much harder in the UK with its system of constituency democracy, “It’s very difficult if it has anything to do with constituencies”, because you then need to poll each one, and comprehensively, and only then start to do the maths.

It is for this reason that the Local Government Elections in 2016 are going to be so difficult to predict. “There you have wards”, she says, “and then you have to do several thousands of interviews just to predict each ward”. Harris says it’s much easier to simply use the national poll from these last elections and try to extrapolate from there.

Which is a pity really because it does seem that if you look at the numbers from last week’s elections, Jo’burg, Tshwane and the Nelson Mandela Bay Metro Municipality (PE) are all in play in two years time.

Anyone working on the political scene over the last few weeks will have known that the political parties themselves were never going to leave this sort of thing to someone else. If they can, they run what they call “tracking polls”. There are various types, but basically after a sample is worked out, a series of households get called every few days and asked, usually by someone speaking their language, which way they are going to vote.

These are also hugely useful for political parties because they allow a message and a strategy to be tweaked as you go along. So, if a certain well known Harvard student is coming along as too strong or too weak on say Nkandla, the strategy can be changed. And of course, should a poll put you in a particularly good light, should it shine some particularly good news in your direction, then you could always make it public. Selectively, of course. If you are an opposition party wanting to show that you “can govern” then this would be a useful strategy.

Harris makes the point that one of the good things about making polls public is that “in a democracy, voters should know how other people are feeling”. That does make complete sense, and would certainly be a good argument for making polls public.

But there several others that go to the heart of a democracy.

Speaking hypothetically, if someone wanted to try and cook the results of an election, it would be far harder to do that, if everyone knew that several polls, conducted by different companies, had come to a completely different result. So, if a whole clutch of polls consistently predicted that Party X would win by a landslide, and Party Y is ruled the winner by the electoral authority, then it would force that authority to explain itself. It would simply be harder to get away with any kind of rampant vote-rigging.

In a democracy like ours, with its combustable personalities, it is also not a bad thing to be able to predict election results with some certainty. If, for example, you can imagine someone angrier, more radical, and more anti-establishment than Julius Malema who honestly believes they are going to win an election, and they also happen to have, at their beck and call, a few thousand people willing to wreak havoc, it is only a good thing for that person to know in advance what is going to happen.

Think of it like this. What would have happened if back in 1994 Mangosuthu Buthelezi had really believed he would win the election? He had control of large numbers of people who had already been involved as perpetrators and victims in violence just months before the elections (The Shell House Massacre etc) and could have reacted unpredictably if he truly believed he should have won. These polls take away the shock value in situations like that, and make it less likely for people to behave emotionally as a result.

All in all, this increase in the quality of polling can only be A Very Good Thing. But it doesn’t come cheap. Harris reckons you’re looking at around a million and a half just to do something half-decent once. But for the strength it gives our democracy, it has to be worth it. DM

Grootes is an EWN reporter.