JOHANNESBURG - In any survey, the method used to reach the results is key.
This survey was conducted by prominent market research company MarkData.
To maximise the response rate, personal interviews were used in township, lower-income and middle-income areas. Telephonic interviews were used in higher-income, gated communities and unsafe areas.
The survey was conducted in six metros: Cape Town, Ekurhuleni, eThekwini, Johannesburg, Nelson Mandela Bay and Tshwane.
The people polled were registered voters 18 years and older, residing in South Africa, who intended to vote in this election.
A multistage random probability sample design was used to draw the sample. A probability sample means that every person in the target population had an equal chance of being selected. Multistage refers to three stages of selection to ensure random, non-biased selection. The three stages of selection were the enumerator area, the household within the enumerator area and the respondent within a household.
The sampling frame was generated from a top-down and bottom-up approach, using the latest released StatsSA census and mid-year estimates, and the GeoTerraImage national Building Census (structure/dwelling unit type) dataset, respectively – an updated version of the SA population.
The sampling frame information included, among others, the number of households, number of people, numbers per population group, gender, age, and neighbourhood lifestyle index (NLI™©).
Regarding telephone interviews, a database of more than 14 million contact numbers from the credit bureau is linked to the sampling frame already referred to was used as sampling basis. Higher-income respondents were randomly selected from the linked telephone database.
The realised sample and margin of error by metro were:
| Realised sample | Margin of error at a 95% confidence level | |
| Cape Town | 1004 | ±3.09 |
| Ekurhuleni | 1019 | ±3.07 |
| eThekwini | 1039 | ±3.04 |
| Johannesburg | 1011 | ±3.08 |
Nelson Mandela Bay | 994 | ±3.11 |
| Tshwane | 1046 | ±3.03 |
| Total | 6113 |
The sample design was disproportionate to ensure that smaller groups and areas were sufficiently represented.
To correct these disproportions, the results were weighted back to the correct population proportions (as per StatsSA Midyear estimates), IEC data on registration, and data on registration vs intention to vote as collected during the survey.
Interviews were conducted at respondents’ homes or telephonically in their language of choice by supervised experienced interviewers from 10 August to 14 September 2026.
At least 20 percent back checks were conducted in the field and at the office to ensure correct completion of the questionnaires.