IEC counts the votes. Picture: THE TIMES

IEC counts the votes. Picture: THE TIMES

Related Articles

To those with spines

Food for Thought

Both sides must be bold

From the Front
Nkandla. Picture: TEBOGO LETSIE

Nkandla definitely an election issue

No fire in the thatch

Mentioned in this Article

Organisations: People: FM Edition:

ASSORTED poll predictions about how political parties will do in the May election is fuelling a public spat between academics and market research companies.

Africa Check, a nonprofit organisation aimed at promoting accuracy in public debate, and the Centre for the Study of Democracy have accused market research companies of producing polls that are more wishful thinking than actual science.

Their main criticism is that these firms don't randomly sample opinion and don't check that respondents are who they say they are. Africa Check has warned that some polls are "designed to mislead" while others are little more than conjecture. It points to a recent poll by financial services group Nomura as an example.

This poll predicted that the ANC would get 56% and the DA 27% but then also acknowledged that "given a total lack of publicly available, detailed polling data at this stage, the margin for error around these forecasts is clearly quite large".

The poll by market research company Ipsos in January also caused a stir when it came to the dramatic conclusion that the ruling party was 10% less popular among eligible voters compared with a similar poll conducted six months before the 2009 election. The poll showed ANC support among a randomly selected group at 53%. Ipsos attributed this to political uncertainty, leadership issues, the aftermath of the Marikana shootings, issues about Nkandla, service delivery protests and the forming of new political parties. Ipsos pegged DA support at 18% and the EFF at 4%.

The March Ipsos poll of registered voters tells what appears to be a completely different story. It pegs ANC support at 66% regardless of high or low voter turnout, which is a higher result than the 65% the party secured in 2009. It pegs the DA at 22,9% and the EFF at 3,7%. There is a caveat to these polls. The first was conducted before Nelson Mandela died and the second before public protector Thuli Madonsela released her Nkandla findings.

Ipsos director of public affairs Mari Harris stresses that the closer a poll is done to an election the more accurate it is because opinion is affected by different events. She also objects to questions about the two different pictures painted by the polls, saying that they can't be compared. One was a sample of randomly selected adults and the other was of registered voters.

While Ipsos stands by its methodology of 3564 face-to-face interviews by trained fieldworkers, Butch Price of market research company Pondering Panda says assertions by African Check and the Centre for the Study of Democracy's Steven Friedman are tantamount to "intellectual terrorism".

Rice does concede that the methodology used by Ipsos could have caused the ANC vote to be overstated. Respondents who didn't express an opinion were assigned to the party that they were "most likely to vote for". Rice also believes that the Ipsos survey vastly underestimated the EFF's vote. He acknowledges that Pondering Panda's research also has limitations because it's done mainly through WAP-enabled phones. This means that poor people in deep rural areas are undersampled.

Though Rice says all pollsters in SA face the same problems unless they have significant funds he dismisses critics who believe that telephonic surveys are more flawed than face-to-face interviews.

"Do you really think you get valid results when a respondent is being quizzed in their homes about their political views, by an interviewer they've never met before, and who might well belong to another ethnic group? In this country, it doesn't pay to make your true political views known in many areas. You could pay for it with your life. Don't you think that a short interview on a cellphone screen might just elicit more honest responses?" asks Rice, who stresses that Ipsos telephonic surveys poll weaker support for Zuma compared with the results of face-to-face surveys.

Pondering Panda's youth surveys found that 25% of young voters between 18 and 24 support the EFF. The company predicts that the new party will secure at least 10% of the overall 2014 election vote. This isn't far off the DA's internal polling, which is done in collaboration with American pollster Stan Greenberg who worked on Bill Clinton's campaign as well as the ANC's in 1994. The DA predicts 8% for the EFF, 59% for the ANC and 26% for itself, which is a 2% improvement on overall percentage of the vote it received in the 2011 municipal election.

DA CEO Jonathan Moakes won't comment on the methodology used by market research companies, but he says: "Ipsos has a record of consistently understating DA support and overstating ANC support. In 2009, they predicted that the DA would get 13% and that the ANC would get 67%. The 2009 election result gave the DA 16,7% and the ANC 65,9%. Our internal polling reflects that today's poll is continuing this trend. We're confident that the DA will succeed in reducing the ANC's share of the vote to well below a two-thirds majority."

Several new polls will be released before the May election and it is only these results that will count in the end. So what will happen if the ANC drops below 60%? Is it a psychological ceiling that, if broken, will unleash panic and a search for scapegoats in the ruling party?

"People are miscalculating the 60%. If the ANC gets below this there won't be a shake-up in the party. Zuma won't lose his job. It will be explained as a sign that the party needs to consolidate and regain ground. But, if the ANC gets 53%, Zuma will lose his job because this means that the party is only a few percent away from losing power," says Wits University vice-chancellor Adam Habib, who doesn't believe that the ANC is at risk of dropping into the low fifties.

For the DA, the real game changing number would be a 35% share of the vote. It's close enough to 50% to give the official opposition a real smell of blood. But DA MPs think a more realistic number for this election is 23%, which is not far off the Ipsos, Pondering Panda numbers. Anything between 23% and 25% will give the party much more traction to build on ahead of the 2016 municipal election, where it hopes to win another metropolitan council as a shoo-in to winning its second province.