In 2003, the Institute for Security Studies (ISS) in South Africa conducted a national victims of crime survey with the aim of measuring crime trends in the country, public perceptions about crime and safety, as well as confidence in the criminal justice system. The study was planned and carried out to allow direct comparisons with a victim of crime survey conducted in 1998 by Statistics SA for the Department of Safety and Security and the United Nations Interregional Crime and Justice Research Institute (UNICRI). The survey was conducted between September and October 2003. Households were randomly selected across the country based on the 2001 Census, and a national sample of 4,860 people, over the age of 16 years, was realised. The sample was stratified by province and urban/rural areas, and the data was weighted to reflect the actual composition of the population.
Kind of Data
Sample survey data
Unit of Analysis
Households and individuals
v1: Edited, anonymised dataset for licensed access
The survey collected household data, including demographic data and data on services and facilities available to the household, and knowledge of loss to HIV/AIDS within the household. Victim data collected included data on general thinking/ beliefs about crime and personal safety, individual and community response to crime, victim support and other interventions, citizen interaction / community cohesion, contact with and attitudes towards police services and the courts, and experience of and attitudes to corruption, and experience of different forms of crime.
The lowest level of geographic aggregation covered by the data is suburb
The survey covered all households in South Africa
Producers and sponsors
Institute for Security Studies
Danish Embassy in Pretoria
International Development Research Centre
Multi-stage cluster sampling was utilised, with Enumerator Areas (EAs) from the 2001 Census selected at the first stage of the sampling, households within the EAs at the second stage, and individuals within the household at the third stage. Based on the total number of households in South Africa (identified by the 2001 Census as 11,205,705), a total of 80,787 EAs were allocated. The total sample size was determined to be 4,050 households. The sample was calculated at a 95% confidence interval, and with a design effect of two. Ten interviews were collected from each EA. The distribution of the sample through the provinces is reflected in the table below. In total, a sample of 4,860 was realised.
Households were selected randomly in the following manner: Maps of the EAs were obtained prior to entry of the EA, and random starting points selected. The household nearest to the starting point was selected, and a household interval randomly selected by the supervisor using the day, week and month of the interview. Every nth household was then completed until the target number of interviews was obtained. Respondents over the age of 16 years were then randomly selected for interview using a KISH grid.
In order to correct slight discrepancies between census data (based on the 2001 SSA Census) and the demographics of the sample achieved, the data was weighted by province, race, gender, age and employment status. This is consistent with the process followed in the 1998 Victims of Crime survey and ensured that the findings are truly representative of the South African population as a whole.
Dates of Data Collection
Data Collection Mode
Data Collection Notes
An independent research company, Development Research Africa (DRA), with prior experience conducting victim and crime related surveys, was appointed to undertake the data collection, quality control, coding and capture process. Teams of four enumerators were supervised by one field supervisor. Enumerators from each of the nine provinces were selected to ensure that all vernacular languages were represented in the teams. Fieldwork took place between 1st September and 30th October 2003. Enumerators were equipped with DRA t-shirts and identity cards to ensure that potential respondents could validate their credentials, as well as to differentiate them from any political or other party or organisation that might canvas or undertake household visits.
In order to ensure comparability with the 1998 national Victims of Crime survey, as well as ISS surveys in other African countries that were being conducted at the same time as the national survey, specific aspects of the survey design were predetermined. These included certain components of the questionnaire as well as the sample design. A basic questionnaire, based on those used in the ICVS, previous ISS studies, concurrent regional studies, and the 1998 Victims of Crime survey was used as a starting point. To enable comparison of the 1998 and 2003 data sets, the definitions used in the study were kept identical to those used in the 1998 survey, with the exception of attempted car hijackings and attempted housebreakings. In this study, car hijacks specifically only included successful incidents, rather than including attempts as was the case in the 1998 study, while attempted housebreakings were excluded from this study. For methodological purposes, the types of crimes were divided into those that affect the household, and those that affect individuals.
In order to minimise fieldworker error, and coding and capturing error, as well as any ambiguities in responses, the instrument was largely pre-coded. A number of drafts were prepared and presented internally to the ISS, and then to a stakeholder committee for input, before finalisation. The questionnaire was not translated into any vernacular languages. However, the introduction and definitions of crimes used by enumerators were translated and back-translated from English into Afrikaans, Zulu, and Sotho. Although every precaution was taken during fieldworker training to ensure that interpretations were correct, it is not inconceivable that errors may have occurred. Cross-checks were carried out on 20% of the questionnaires to ensure that the responses given were an accurate reflection of respondents’ opinions. These cross-checks did not reveal any consistent error from the latitude ordinarily accorded interviewers in translating the original questionnaire
Capture was undertaken on Epi-Info. A process of double capture was undertaken in order to eliminate capture error.
Licensed use files, available for non-commercial use only
Institute for Security Studies. National Victims of Crime Survey 2003 [dataset]. Version 1. Pretoria: Institute for Security Studies [producer],2004. Cape Town: DataFirst [distributor], 2012. DOI: https://doi.org/10.25828/jgz6-4721