From 2002-2005 the Agincourt Integrated Family Survey project collected data in Limpopo Province, at the Agincourt Demographic Surveillance Site through the auspices of Philani Nutrition and Development Project. The work was funded by the National Institute on Aging under grant numbers R01 AG20275-01, P01 AG05842-14, and P30 AG024361. The project used integrated health and economic surveys in South Africa to investigate the links between health status and economic status. Our survey instruments collected data on a range of traditional and non-traditional measures of well-being including income and consumption, measures of health status (including mental health), morbidity, crime, social connectedness, intra-household relationships, and direct hedonic measures of well-being.
In 2004, the households who had been interviewed in 2002 were re-interviewed (if they were willing and if they could be found), for part 1 of the second wave. In 2005, the households who were interviewed in 2003 were re-interviewed (if they were willing and if they could be found), for part 2 of the second wave.
For all of the studies, the methodology for conducting the surveys was the same. The questions varied some from year to year. The crosswalk (see table of contents) identifies these variations.
In study years 2003, 2004, and 2005, detailed questions in the household questionnaire about the impact of the most recent death in the household were asked of the most knowledgeable household member.
Kind of Data
Sample survey data
Unit of Analysis
Households and individuals
v1: Edited, anonymised dataset for public distribution
The survey collected data on income and consumption, measures of health status (including mental health), morbidity, crime, social connectedness, intra-household relationships, and direct hedonic measures of well-being
The survey covered a rural sub-district in Mpumalanga Province, South Africa
The lowest geographical unit is village.
The Agincourt Integrated Family Survey universe included all household residents, Woman aged 60 and above and men aged 65 and above in the household
Producers and sponsors
Professor Anne Case
National Institute on Aging
In January 2004, using Agincourt Health and Population Unit (AHPU) census information, the Agincourt Integrated Family Survey project team drew a stratified random sample of 475 households across all villages in the Agincourt Demographic Surveillance Site, with stratification on both citizenship (South African versus Mozambican) and on whether the household had lost a member to death in the period from June 1, 2002 to May 31, 2003. By "Mozambican household" in the sample design we mean that the nationality of the head of household is Mozambican.
Deviations from the Sample Design
The sample was designed to be 60 percent South African, and 40 percent Mozambican. In execution, slightly fewer South African households without a death were interviewed (187 instead of 190), and one extra South African household with a death was interviewed (96 instead of 95).
These discrepancies were the result of confusion over which households were considered to have a complete interview in cases where the head of household refused to be interviewed. The survey is composed of a household module, to be completed by a knowledgeable household member; an adult module, to be completed by each member aged 18 or older; and a child module, to be completed for each child aged 12 or younger. Some adult household members were migrants who were not in the field site to be interviewed (although the field team made a great effort to make appointments with the household to return at month-end, or at Easter, to interview returning migrants). In addition, some adult members refused to be interviewed. It was decided that if the household module was completed, and at least one adult was interviewed, the household had a complete interview.
In the South African-Death Stratum, an extra household was interviewed because the household head came home for Easter, after the rest of the household had been interviewed, and refused to participate. The field team then interviewed a replacement household, but need not have as a decision had previously been made that if the head refused to participate, but did not stop other members from doing so, then that household's information would be used. However, if a returning head refused to let any members participate (even if they had already been interviewed),that household was not used in the analysis.
We calculate household inflation factors as the inverse of the product of the probability that an individual household i in strata s was chosen times the number of households surveyed in strata s: (See Angus Deaton, Analysis of Household Surveys, 1997, page 50, for details.)
The inflation factors, calculated using actual sample sizes from Box 1, and the number of households of each type in the field site, are given in Box 2. These inflation factors are also included in the household level data file under the name "Hweight."
Dates of Data Collection
Data Collection Mode
The most knowledgeable household member (khhm) was the initial person interviewed within the household. He or she would list all of the members of the household. This list of household members was then used as a guide for the entire interview process. The khhm first answered questions about the individual members of the household: age, gender, education, marital status, is that person?s partner in the household, is that person?s parent in the household. Additionally, the khhm was asked about the source and amount of income of each household member.
Summary information was gathered from the khhm about the household members who had died and the household members who had moved. In study years starting in 2003, detailed questions were posed about the effect the most recent death had on the household.
Detailed questions were asked of the khhm about the living conditions: access to toilet facilities and running water, a stove, a phone. Questions about household expenditure were asked
In 2002 there were two separate adult questionnaires. One questionnaire was for adults whose ages fell between 18 and 54; the second questionnaire was for adults 55 years old and up. In subsequent study years there was one questionnaire for adults 18 and older. Every adult from the household, who was available and willing to, answered these questions. Questions about age, marital status, number of living children, and number of children who have died were included. Detailed questions were asked about their sources of income and their expenditures. For older adults, there were questions about pensions and grants, for mothers there were questions about childcare grants. Individuals described the type of jobs they held over the years, how much money they earned and how they spent that money. Detailed health questions were posed; both physical and mental health issues were covered. Physical measurements were taken of the individuals interviewed: their height, weight, waist size; blood pressure and pulse.
The parent or guardian of each child was questioned about the child. Questions included those on birth weight, history of breastfeeding and health of the child. With the parent's or guardian's permission the child?s height and weight were measured. Detailed information was recorded about the child?s immunization history.
Case, A. 2005. Agincourt Integrated Family Survey 2004 [dataset]. Version 1. Princeton: Princeton University [producer], 2003. Version 1. Cape Town: DataFirst [distributor], 2011. DOI: https://doi.org/10.25828/s4x8-da41