Cape Town Household Survey 2024
Round 1
South Africa, 2024
Get MicrodataIdentification
| Name | Country code |
|---|---|
| South Africa | zaf |
| Type | Identifier |
|---|---|
| doi | https://doi.org/10.25828/r2bg-fw79 |
| doi | 10.25828/r2bg-fw79 |
The City of Cape Town conducted the first Cape Town Household Survey between February and October 2024. The survey covered over 8,300 randomly selected households across around 150 suburbs, representing a mix of living arrangements and levels of formality. The survey was designed to obtain accurate, frequent, sub-metro data on households in formal dwellings, informal settlements, and additional dwellings informal (ADI). National surveys and City administrative records leave gaps, especially for households living in informality. The survey informs metropolitan planning, policy, and decision-making on socio-economic conditions, service delivery, and urban dynamics. Its aims are to bridge the municipal level data gap in national surveys, establish longitudinal data at sub-metro level, disaggregate service-delivery variables, group results into different spatial boundaries, compare results with Statistics South Africa census data, and integrate with city-level administrative data. The next round of the survey is expected in 2027.
Version
Scope
Questionnaire modules cover assigned geographic identifiers, household composition, demographics, education, transport, disability, migration, employment, dwelling and tenure, assets, basic services, community amenities, burial preferences, negative events, spending changes, hunger, home business, income and grants, expenditure, interviewer observations, and device-captured fields including duration and GPS. Some modules (health diagnoses, safety, protection measures, rental, and precise GPS) are withheld from the Open Data release.
Coverage
The data covers the City of Cape Town Metropolitan Municipality in the Western Cape Province of South Africa. Orphaned sub-places too small for a suburb profile were sampled for metro totals only.
The lowest level of geographic aggregation of the data is suburb.
Private households in residential dwellings in the City of Cape Town Metropolitan Municipality. A household is a group of people who live together at least four nights a week, eat together, and share resources, or a single person who lives alone, occupying the same dwelling unit. Collective living quarters (student residences, prisons, military barracks, long-term care), industrial sub-places, and nature-reserve sub-places were excluded. Employees of the household (such as domestic workers) and tenants or lodgers paying to live with the household were treated as separate households.
Producers and sponsors
| Name |
|---|
| City of Cape Town, Policy and Strategy Department, Research Analytics Unit |
| Name | Abbreviation | Role |
|---|---|---|
| City of Cape Town | CCT |
Funder |
| Name | Affiliation | Role |
|---|---|---|
| Kantar | Field research company contracted by the City of Cape Town |
Fieldwork and data collection |
Sampling
Sample frame
GTI updates BBLU through the New Development System, using satellite imagery on a weekly basis. The City updates population and household counts by sub-place annually with a growth model that uses building applications, aerial photography, Human Settlements data, and Statistics South Africa Mid-Year Population Estimates.
A stratified random sample was drawn with suburbs as the primary reporting unit. Strata combined 2011 Census sub-places with dwelling type: formal, informal settlement, and additional dwellings informal (ADI). After grouping sub-places and splitting large suburbs (including Delft, Gugulethu, Khayelitsha, Kraaifontein, Mitchell's Plain, and Philippi), the design used about 211 suburbs. The sample size targeted a 95% confidence interval per suburb and a 15% margin of error, with at least two households in each stratum. Homogeneous sub-places (at least 85% of one dwelling type) were sampled at random within the sub-place. Heterogeneous sub-places were stratified by dwelling type, with a minimum of two dwellings for each type that accounted for more than 10% of households; ADI and informal settlement were combined when either fell below 10%. Remaining sample was allocated by a power rule between proportionate and equal allocation. Visiting points came from the GeoTerraImage Building Based Land Use (BBLU) dataset through Survey Sampler, spatially sorted and drawn systematically, with twice as many substitute points as sample points. Collective living quarters, industrial areas, and nature reserves were excluded. Some orphaned sub-places were sampled for metro totals only and were not assigned to a suburb profile.
After the sample was drawn, City officials reviewed geographic spread. Some informal-settlement and ADI boundaries were redrawn where dwellings lay just outside the intended sub-place or official property boundary.
The overall response rate was 84.0%: 80.2% in additional dwellings informal (ADI), 86.6% in formal dwellings, and 74.0% in informal settlements. Some suburbs recorded a 0% response rate (including Acacia Park, Foreshore/V&A, Pollsmoor, and Stonehurst).
The household design weight is the estimated number of households in the stratum divided by the selected sample in the stratum: W_HH = HH_stratum / n_stratum. Household weights were then adjusted for non-response and for over- or under-coverage of dwelling types within the suburb, calibrated to the number of households by dwelling type in each suburb, and scaled to the Census 2022 household total of 1,452,845 for the City of Cape Town. Because every member of a selected household was listed, the person design weight equals the household weight. Integrated weighting (Lemaitre and Dufour 1987; Nieuwenbroek 1993), applied with a SAS macro, produced a single set of weights for household and person estimates. Auxiliary totals were suburb household counts and dwelling type from the City sampling frame, and Census 2022 age group (0-4 / 5-14 / 15-24 / 25-59 / 60+), gender, and population group. Use hhwght for household analyses. Person weights were available only at metro level at the time of release; unweighted person figures may be more stable for small geographic areas until person weights are updated.
Survey instrument
The survey used a structured CAPI questionnaire on SurveyToGo, with mainly precoded answers and some open text (for example relationship-other and place names). Interviewer showcards appear in Appendix 1 of the source metadata document provided with the data. A full alphabetic index of variables is in Appendix 2.
The questionnaire was piloted in October and November 2023 (n = 150) and revised before main fieldwork. Translations were prepared in Afrikaans, English, and isiXhosa.
Data collection
| Start | End | Cycle |
|---|---|---|
| 2024-02-17 | 2024-09-30 | Main fieldwork |
| 2024-10-05 | 2024-10-06 | Additional interviews in selected suburbs |
Every 2-3 years
| Name | Affiliation | Role |
|---|---|---|
| Kantar | Contracted by the City of Cape Town |
Fieldwork contractor |
| Response Rate | Training |
|---|---|
| Interviewer briefing |
A face-to-face briefing was held on 13 February 2024 in Cape Town and was attended by 99 interviewers and fieldwork supervisors. Kantar project managers led the session; City officials attended. Interviewers then worked through the project pack and a test-mode questionnaire for two days. A Microsoft Teams follow-up on 15 February 2024 answered remaining questions. |
The interviewed household had to live in a dwelling within 60 m of the sampled point, confirmed by GPS at the end of the interview or by a geolocated address. In informal settlements, the interview had to take place in the same informal settlement or Area of Informality (AOI), with GPS, address, or settlement-name evidence. Dwelling type (formal, ADI, or informal settlement) and census sub-place had to match the sample. Kantar conducted back-checks up to six weeks after the interview. Names of household members and contact details collected for quality control were deleted after that period; the City did not receive names.
On arrival at the sampled dwelling, the interviewer listed adult household members and identified those able to answer for the household and its members. Any such adult who was available could complete the interview; selection within the household was not randomised, and appointments and return visits were common. A substitute dwelling could be used after three unsuccessful visits at different times of day (including one outside working hours), after a refusal, or when the structure was vacant or not residential. Two replacements were drawn in the same stratum; if those also failed, a new dwelling was sampled. Interviews used Computer Assisted Personal Interviewing (CAPI) on SurveyToGo devices running Android 11 or higher. The questionnaire was available in Afrikaans, English, and isiXhosa; most interviews were in English. A pilot in October and November 2023 tested dwelling types, gated communities, translations, foreign respondents, and high-risk areas in five parts of the city (n = 150).
Data processing
Interviews that failed location, dwelling-type, or sub-place checks were excluded. Household members who did not stay at least four nights a week were removed, as were persons employed by the household or paying to live with the household where those cases had been listed in error. Open-text relationship and other-specify fields were categorised during cleaning where possible. Date of birth was used to validate age and is not released.
Study activities
A pilot in October and November 2023 tested all aspects of the survey, with emphasis on the three main dwelling types, gated communities, questionnaire translations, response of foreigners, and high-risk areas. A representative sample was drawn in Hout Bay, Philippi, and Vrygrond; a purposive sample was drawn in Bellville and Strand. The pilot achieved 150 households.
| Name | Role |
|---|---|
| Kantar |
Fieldwork |
| City of Cape Town |
Survey sponsor |
Pilot sample of 150 households; questionnaire and field procedures revised for the 2024 main survey.
Data appraisal
The sample was designed for a 95% confidence interval per suburb and a 15% margin of error, subject to a minimum of two households per stratum.
Interviewers sometimes interviewed a nearby household rather than the sampled dwelling. A 60 m proximity rule (or same AOI in informal settlements) and a requirement that dwelling type and sub-place match the sample were used to limit selection bias. GPS captured on the device was not always accurate; geolocated address was used as a check. Some special- request modules (health diagnoses, safety, protection measures) have known limitations and are released only to subject-matter experts.
Metadata production
| Name | Affiliation |
|---|---|
| DataFirst | University of Cape Town |