In the year 2000 a small team of social scientists from the Universities of Cape Town and Michigan collaborated on designing a survey with a special focus on labour market issues as a precursor to a Cape Area Panel Study with a special focus on youth planned for the year 2002. After much debate and taking due cognisance of time and budget constraints the team decided to target the magisterial district of Mitchell’s Plain within the Cape Metropole for the survey.
This decision was informed by data gleaned from the 1996 census which revealed that Mitchell’s Plain – demarcated a magisterial district in 1986 – contained almost thirty percent of the population in the Cape Metropolitan Council area. It straddled the two cities of Cape Town and Tygerberg and housed nearly 74% of the African and over 20% of the ‘coloured’ metropolitan population. It included the three established African townships of Langa, Gugulethu and Nyanga as well as informal settlements such as Crossroads and Browns Farm. It also included Khayelitsha an African township proclaimed in the early 1980s with the first houses being built in 1986. The 1996 census had recorded high unemployment rates of over 44%, for Africans and over 20% for Coloured people.
Kind of Data
Sample survey data
Unit of Analysis
Households and individuals
The survey covers the magisterial district of Mitchell's Plain within the Cape Metropolitan area in South Africa. This includes the townships of Khayelitsha, Langa, Gugulethu and Nyanga and the informal settlements of Crossroads and Browns Farm.
The lowest level of geographic aggregation covered by the data is township.
The survey population of interest isl adults aged 18 years of age and older in households in every Enumeration Area in the Mitchell's Plain magisterial district. Before selecting Enumerator Areas, the survey excluded all non-residential and institutional Enumerator Areas (except for hostels) from the sample frame. Enumerator Areas were selected systematically to ensure that their probability of selection was proportionate to their population size. The Mitchell's Plain magisterial district consist of 1,486 populated Enumerator Areas (as defined in the 1996 Population Census). Dividing the target number of questionnaires (2,875) by the average number of adults per household (2.66), the survey determined to select 1,081 households.
A more detailed description of the universe for this survey can be found in the sampling method document available in the zipped folder under 'documentation'.
Producers and sponsors
Southern Africa Labour and Development Research Unit
University of Cape Town
University of Michigan. Population Studies Centre
Andrew W. Mellon Foundation
The sample is based on the 1996 Population Census which recorded a total population of 728 916 people in the Mitchell's Plain Magisterial district. The survey sampled households in the designated geographic area using a two-stage cluster sample. The first stage of this sample entails selecting clusters of households and the second stage entails the selection of the households themselves. For the clusters of households, the survey relied on the Enumerator Areas as defined by Statistics South Africa for the 1996 Population Census. Enumerator Areas are neighbourhoods of roughly 50 to 200 households and drawn up by the Chief Directorate of Demography at Statistics South Africa. This directorate is responsible for developing and maintaining a GIS system that provides the maps that are used for conducting the five-yearly national population census (Statistics South Africa, 2001:42-44). Although Enumerator Area boundaries do not cross municipal boundaries, they do not correspond to any other administrative demarcations such as voting wards. Enumerator Areas are designed to be homogeneous with respect to housing type and size. For example, Enumerator Area boundaries within the Mitchell's Plain Magisterial District do not usually cut across different types of settlements such as squatter camps, site and service settlements, hostels, formal council estates or privately built estates. Instead, each Enumerator Area is homogeneous with respect to any one of these housing types.
The method of selection used was that of Probability Proportional to Size (PPS), where size is measured as the number of households in each Enumerator Area (as defined by the 1996 Population Census). This method provides the most efficient way to obtain equal sub-sample sizes across two stages of selection, i.e. to select the Enumerator Areas and then select from each Enumerator Area a constant number of households for all Enumerator Areas in the sample. The sample is implicitly stratified by location and by housing type.
A more detailed description of the sampling method and procedure for this survey can be found in the sampling method document available in the zipped folder under 'documentation'.
Post Stratification and weighting for non-response:
Non-response can lead to an increase in sample errors and bias in estimates. The weight pweight2, included in the data, is designed to correct for households who were selected into the sample, but were not interviewed. This could have been due to refusals, failure to locate the household or not being able to make contact with the residents.
The weight pweight2, included in the data, adjusts for this type of nonresponse. Pweight2 was constructed by adjusting the original weight, pweight1 by the inverse of the response rate in each Enumeration area. For example, if the response rate is 80 percent then a suitable weight would be 1/0.80 =1.25. This was done for each numerator Area and applied to each responding household. (Under such a method of dealing with non-response it is assumed that all households selected into the sample have the same probability of responding.)
Non-response in the survey also occurs at the individual level i.e. there are cases where not all adults within a household were interviewed. Nonresponse at the individual level is corrected for by post stratifying the data according to known age, gender and race proportions as reflected in the the 1996 population census. The raking ratio method of post stratifying weight adjustment was used to calculate and make adjustments to the pweight2 variable in the data. This results in the variable, adultrakingweight, in the data. When analysing data in the adult file of the data the post stratified weight, adultrakingweight, should be used to adjust for adult non-response in the survey.
All three weights are available to the analyst:
• pweight1:-the original weight with no adjustments for nonreponse.
• pweight2:- pweight1, adjusted for household non-response.
• Adultrakingweight:-post stratified weights to adjust for adult nonresponse.
It is suggested that the post stratified weights be applied when doing most analysis.
Attention is drawn to the fact that the survey was completed in the year 2000, whereas the auxiliary information used to make the post stratification adjustments come from the 1996 population census. It is likely that some change in population proportions may have occurred between 1996 and 2001. It is our intention to use the 2001 census data to further adjust the weights, when this information becomes available.
Dates of Data Collection
Data Collection Mode
The household questionnaire was aimed at establishing the household roster with the usual questions on age, gender and relationships. It was divided into two sections covering those aged 18 and older and those younger than 18. For the latter, a separate set of questions covering education, health and work status was included.
The adult questionnaire was aimed to fit the international standard approach on the labour force by allocating the labour market status of ‘employee’ to all those ‘at work’ (for profit or family gain, in cash or in kind). One of the innovative aspects of the survey was that respondents were asked about all income-earning activities. In other words, they were not allocated into particular labour market categories during the process of the interview.
The adult questionnaire was divided into 13 sections:
• Section A on education and other characteristics covered age, racial classification, educational attainment, language, religion and health.
• Section B on migration covered place of origin, relocation and destination.
• Section C on intergenerational mobility aimed at capturing parental influence on the respondent.
• Section D on employment history aimed at capturing the respondent’s work history.
• Section E on wage employment attempted to capture respondents working for a wage or salary whether full-time, part-time, in the formal sector or the informal sector including those who had more than one job.
• Section F on unemployment included questions on job search
• Section G on self-employment included a question on more than one economic activity and the frequency of self-employment.
• Section H on non-labour force participants was aimed at refining work status.
• Section I on casual work aimed to capture not only those in irregular/short term employment but also people who might have more than one job.
• Section J on helping other people with their business for gain was aimed at identifying respondents who assist others from time to time but who might not regard themselves as ‘working’.
• Section K on reservation wages attempted to establish the lowest wage at which a respondent would accept work.
• Section L on savings, borrowing and grants and investment income attempted to capture income derived from sources other than work
• Section M on perceptions of distributive justice posed a number of attitudinal questions.
Southern Africa Labour and Development Research Unit (SALDRU). Khayelitsha Mitchell's Plain Survey 2000 [dataset]. Version 7. Cape Town: SALDU [producer], 2000. Cape Town: DataFirst [distributor], 2008. DOI: https://doi.org/10.25828/6hsm-5v93