What are the opinion polls really telling us about party support?

ANC headquarters in Port Nolloth, circa 2008

Photo Credit: Flickr/James Castleden

Lately, there seems to be a consensus amongst many analysts that ANC support will fall well below 40% in the next election, thus making a minority government  - where the opposition parties form a coalition in order to govern - a real possibility.

They base this scenario on the outcome of by-elections as well as opinion polls. Although it is true that the ANC has lost some seats in by-elections it is not true all round. It is also important to note that the turnout in most of these by-elections has been abysmal.

Also, people often vote differently at local level than they do in national elections. Local issues and personalities play a far bigger role in voter choice during local elections or by-elections – which explains why the ANC has consistently done better at every national election compared to the preceding local elections. Under Mandela and Mbeki, it was between 8% and 10% better; under Zuma it fell to 1% and under Ramaphosa, it increased again to 4%. 

So, if this trend continues, the ANC could get close to the 50% mark or even above it at the next election.

My point is that by-elections are not a foolproof indicator of national support. 

Apart from by-elections, many analysts base their assumptions on various recent opinion polls. Yet, these too need to be treated with caution. As Tim Cohen lamented recently, we have very little reliable polling in South Africa. In fact, around the world opinion polls have proven to be very inaccurate over the last few years.

It is therefore important to analyse these polls carefully, since behind the headlines are often far more nuanced scenarios.

For example, Ipsos recently released a poll that suggested that if an election were to be held now, the ANC would fall below 42% - a dramatic drop from the current 57,5% in the National Assembly. Their polling further suggested that the DA would get 11%  (down by 9%) and  the EFF would remain stable at around 9%. 

However, as Dawie Scholtz pointed out, Ipsos polled eligible voters (people over 18), rather than only those who are registered and intend to vote. This means that a sizeable percentage (some argue up to 26%) of those polled would not actually vote.

If you adjust the poll to eliminate the non-registered and non-voting participants, the ANC would then receive between 47% and 52% depending on the turnout. (The DA support would be between 13% and 18%)

Contrast that with a “confidential country wide poll” reported on by Rapport newspaper which suggested that the ANC would get 38% and the DA 27%. In their front page report, they didn’t disclose how the poll was done or who did it, so it is not clear what the sample size, geographical distribution or voter profile were.

Another poll at around the same time by the Social Research Foundation found that ANC would get roughly 50% of the votes with a 66% voter turnout, or 52% with a 56% voter turnout.

According to this poll, the DA would get 25% with both a 56% and 66% voter turnout, while the EFF would get between 11% and 12%.

So, from all the polls we can only really conclude that support for the ANC is on the decline, but how much exactly, only time will tell.

We also know that it will be extremely dependent on turnout. Research proved that - with the exception of a few pockets in Gauteng – ANC voters did not vote for other parties in the last local government election. Instead, they chose not to vote at all. It therefore stands to reason that if the ANC can convince their erstwhile voters to turn up on the day that they would continue to vote ANC, regaining for that party some of the support lost during the last few years.

Of course, that will rely heavily on effective organisational capacity as well as money, both of which are in short supply at the moment.

Finally, it is important to note that we are still far away from the actual election. As the saying goes: “a week is a long time in politics”.

So, for now these polls make for interesting reading, but very little else.