How the survey was made as accurate as possible

THE random survey sample was demographically representative and representative of rural and urban voters.

It comprised 2 219 face-to-face interviews in which about 40 questions were posed, including a national and provincial ballot question.

The margin of error was between 1.25% and 2.8%, at a 95% confidence level.

Respondents were disqualified if they indicated they were not registered voters.

The official Independent Electoral Commission registration figures were used for weighting. To determine the different turnout scenarios, the response to a question about an intention to vote was used. To obtain a more accurate reflection of the final numbers, undecided voters (those who did not indicate a party preference, either because they were undecided or refused the question) were assigned to a party based on their answers to other key indicators, including trust, partiality towards party leaders and policy approval.

Without assigning undecided voters or catering for turnout, the percentages for the top three parties were ANC 56.6%; Democratic Alliance 18% and Economic Freedom Fighters 4%.