VICTIM SURVEYS1

Over the past thirty years, countries across the world have begun conducting victim surveys to fill the gaps left by official crime statistics.2 By asking representative samples of the public about the types of crimes they have experienced over a particular period, a more accurate picture of the levels and categories of crime can be compiled than that provided by the police. However, the value of victim surveys goes beyond simply compiling better statistics about incidents of crime. These surveys collect information from the perspective of the public rather than the police and courts, which means that experiences of crime and violence are not limited by the legal definitions of these acts.3

The surveys also canvass the views of the public and crime victims about their experiences with the police and the legal system, which enables better evaluation of these agencies. In addition, by determining which crime types are perceived to be the most serious and occur the most frequently, survey material facilitates the prioritisation of preventive measures by the police and other agencies. Further, by shifting the focus of the inquiry from the offender (traditionally the pre-occupation of the criminal justice system) to the victims of crime, these surveys can provide information which enables victims themselves to take preventive action against further victimisation.

In South Africa, several national public opinion surveys about crime have been conducted, primarily by the Human Sciences Research Council (HSRC) over a period of more than two decades.4 Some of these tended to focus on attitudes towards safety and the fear of crime rather than the actual extent of victimisation. Those surveys which did measure crime levels were conducted as part of broader public opinion assessments rather than as focused crime surveys. One exception is the national crime survey conducted by the Nedcor Project in 1995, which measured crime levels and perceptions of safety.5

The National Crime Prevention Strategy, released in May 1996, calls for a more accurate picture of crime and more information on the impact of crime on the lives of ordinary citizens. Recently, the Department of Safety and Security has commissioned a countrywide survey to be conducted during March 1998 in all nine provinces. This national victimisation survey ever to be conducted in South Africa and run by Central Statistics (CSS) will be the first official attempt to probe unreported crime levels. During the survey, personal interviews will be conducted with randomly selected persons in 4 000 households. Respondents will be asked to provide information on their experiences as victims of crime during the past five years, with a specific focus on crimes that occurred during 1997. The survey will provide baseline data for future surveys, which will initially be repeated annually, and at a later date, every two years. By repeating the survey, a more accurate assessment of increases or decreases in crime, than that possible using police crime statistics, will be achieved.6

During recent months, the Institute for Security Studies with its research partner DRA Development, have conducted city surveys in Johannesburg7 (July 1997), Durban (December 1997) and Cape Town (February 1998). While based on questionnaires similar to that used by the International (Crime) Victim Survey (ICVS), direct comparisons between the data sets of these surveys and the ISS studies need to be conducted with caution given their methodological differences. The ICVS surveys were administered to people in their households, while the ISS carried out a street survey, sampling the views of people in various public places across metropolitan areas. Despite this difference, as well as the expectation that a street survey would deliver abnormally high victimisation levels, results in terms of the incidence of crime do not differ dramatically.

While victim surveys provide rich information to complement that collected by official criminal justice agencies, there are limitations regarding the type of data collected and their application. Victims' responses to the surveys are affected by their ability to recall events and when they happened, their reluctance to discuss their experiences, particularly in the case of sexual crimes and domestic violence, and their failure to recognise that some incidents are relevant to the survey. These factors are likely to result in an undercount of crime and suggest that, to some extent, surveys measure public perceptions of crime as expressed to the interviewers, rather than actual experiences.8

Variations in how incidents are understood by respondents, particularly where the sample is not homogeneous, may also affect the consistency and comparability of data. If victims have had contact with the criminal justice system, their interpretation of incidents for the survey may be confused by the official definitions used. For example, the difference between burglary and robbery, particularly if the victim was present when the burglary occurred, may not be immediately apparent. Problems of definition and interpretation especially affect the documentation of sexual incidents.

In the case of sexual incidents, the quality and quantity of information reported to victim surveys is generally likely to be limited. Estimating the level of gender-based violence, utilising both police reports and through victimisation surveys, can be intensely problematic. While gender-based crimes recorded by the police are widely accepted as reflecting a small percentage of the actual incidence of sexual victimisation (i.e. it is estimated that only one out of every 35 rapes is reported to the police), victimisation surveys can also yield underestimations and inaccurate results for a variety of reasons.

Firstly, women who are or have been victims of gender-based crimes, particularly sexual assault, are rarely in a position to discuss this violation in public. For the very reasons that survivors of sexual assault do not report incidents to the police (embarrassment, self-blame, fear of not being believed, trauma of official action, secondary victimisation by state officials or fear of retaliation), survivors of sexual assaults may also be extremely reluctant to confide in unknown interviewers about questions of a sexual nature. Rape is not only a traumatic experience for the victim, but also the only crime for which the victim is often socially stigmatised. More so than any other crime, there are strong pressures on the victim not to report the incident to a complete stranger.9 Research has also shown that many women are reluctant to report sexual victimisation when they feel that they should have predicted when a man would act violently toward them.10 Feeling responsible for the attack is unfortunately a very common reaction for women. Clearly, victimisation surveys need more anonymous and confidential modes of acquiring sensitive victim information.

Secondly, while criminal justice definitions of sexual offences are narrow, the respondent of the victimisation survey may also perceive the crime within this legal framework. For example, current rape law is gender, object and orifice specific (for rape to have occurred the woman's vagina must have been penetrated by the man's penis) and the victim, based on her knowledge or understanding of sexual assault, may likely disregard sexual violations that do not fit the prevailing legal or social definitions of sexual assault. In other words, the survey may not capture the broad range of violations experienced by women identified by researchers and service providers. Some women may be reluctant to admit a violation even took place at all.

Further, like legal and social definitions of rape and sexual assault, victimisation surveys may impose predetermined definitions of sexual victimisation, also narrowing the scope of responses. A review of twenty studies of rape prevalence showed a wide variety of estimates, in large part due to the kinds of questions asked, the way they were asked and whether the respondents were probed for further information.11

In principle, victimisation surveys offer an alternative approach to the measurement of sexual victimisation by taking into account women's experiences of sexual crimes, and whether they have reported the crime to the police or not. In practice, however, the reliability of survey measurements of victimisation is still highly questioned by feminist researchers who argue that victimisation surveys need to be radically reworked so that disclosure of sexual violations are facilitated with far more sensitive and informed methodologies.

It can safely be argued that there will be substantial attrition between assaults that are reported to the police, assaults that are revealed in victimisation surveys and the actual number of gender-based offences that occur. There is considerable doubt whether the number of affirmative answers about sexual victimisation represents a reliable account. Criticisms of conventional survey approaches in measuring sexual victimisation abound: "Errors which apply generally to measuring victimisation also apply to sexual victimisation ... the principle ones being sampling error, memory problems, the tendency to pull forward relevant incidents in time and 'routine', non-response owing to concerns about the validity of the survey [and] lack of time."12

The combination of these factors may indeed underplay the frequency and extent of sexual victimisation recorded by the survey in the Greater Cape Metropolitan area. If anything, this survey of victims on the issues of sexual assault and harassment, reveals the reluctance of women to disclose these violations. Recognising the limitations of general victimisation surveys in capturing information about sexual offences in an accurate and sensitive manner, the Institute for Security Studies will be conducting a focused Women's Survey in Johannesburg during 1998.

Similarly, crimes committed against children are poorly covered, largely because parental consent and supervision are required when surveying children. Specialised surveys have been conducted in recent years to cover the experiences of these and other less sensitive groups, such as tourists and business.

In developing countries like South Africa, the application of crime survey data faces difficulties: police crime statistics are often regarded with scepticism and lack detail, and thus, expectations of victim surveys are high. In attempting to meet these expectations, questionnaires may become too long, adding to the interview time and thus potentially reducing the accuracy of the data.

That the public, policy-makers and criminal justice officials - given the paucity of useful crime information - often expect victim surveys to provide answers on how crime can be controlled and prevented, is of more concern. This is not the case: apart from identifying crime trends, victim surveys indicate broad areas for further debate, investigation and intervention. Bearing these difficulties in mind, the methodology of the Cape Town survey, in the context of financial and time constraints, required careful consideration.

SURVEY METHODOLOGY

The Cape Town victim survey was initiated to gain an understanding of the levels of crime and violence in the different communities within the Cape Town Metropolitan Council. This entailed:

the collection, collation and synthesising of all existing and available research and information on generic types of communities (including demographic statistics) and residential areas located within the Cape Town metropolis;



the development of broad profiles of the defined constituents, including some select demographic statistics such as age, gender, vocation, and place of work and study;



the identification of the ratio of victims to non-victims of crime by race and residential area; and



the development of the necessary sampling formulae to accommodate this.

The administration of interviews in an uncontrolled environment, such as on the street, at nodal interchanges, and in other public spaces such as places of welfare, is unique to the street survey methodology. This is a significant departure from typical crime surveys that rely either on face-to-face interviews in a controlled environment, or on postal surveys. Such surveys have advantages, but are both expensive and time consuming to administer. Given the financial constraints of the project, it was decided to conduct a survey based on a street sample instead. This methodology had been applied in similar studies conducted by the ISS both in Durban and in Johannesburg.13 This resulted in the calculated application of a more refined research methodology in Cape Town. Careful planning was required to establish the sample size and sample points. This was done in a number of stages.

Stage 1

From the outset, it was decided that the survey results should be analysed categorically by race, gender, age and type of residential settlement. A departure from the Durban and Johannesburg surveys included the separation of the grid questionnaire (which provided a profile of both victims and non-victims in the metropolis by age, race, residential area, crime type, victimisation level, as well as repeat victimisation and multiple victimisation) for the determination of victimisation ratios (no sexual crimes were asked in the grid). In this scan survey, 3 839 people were approached in 54 areas across the metropolis, plotting the incidence of crime. In each sampling node, 72 respondents were approached in four age categories, half of which were male and half female. This provided a minimum subsample of eighteen respondents in each age and gender category. Prior to approaching the respondents, no distinction was made between victims and non-victims. The only screening criterion was age.

In the more detailed experience and perception survey, 2 000 people (1 000 victims and 1 000 non-victims) were approached. In this survey, sixty sampling nodes were selected. Variables for analysis were applied similar to those in the grid survey, namely race, gender, age and type of residential settlement, with a minimum subsample of eight in each age and gender category.

Stage 2

It was decided that, although the research could not be representative of individual substructures within the Cape Town Metropolitan Council, it should represent the various area types within the metropolis. It was also accepted that there are very few areas in Cape Town that lack a residential population, thus all geographic localities were included in the sample framework.

In order to target the correct types of sample areas, the selection had to be undertaken in such a manner that each different residential generic typology was represented by at least one sample point. In order to ensure that the selection of the sample points covered the entire metropolis, and that all residential generic typologies were represented, a total of 54 primary sample points were identified for the grid survey. Another sixty primary sample points were identified for the detailed survey. Primary sample points were suburbs, townships, inner city areas or informal settlements.

Within each primary sample point it was important to select a number of secondary sample points where the interviews were to be administered. It was decided that at least five different secondary sample points were to be identified within each primary sample point. Secondary sample points included:

shopping and recreation centres (shopping malls, flea markets, corner cafés and stores, plazas, spazas, shebeens, sports grounds, public parks and major streets);



transport nodes (taxi ranks, bus stops, railway stations and parking lots);



education centres (schools, universities and technikons);



health and welfare centres (pension pay-out points, civic buildings, hospitals); and



residential areas (private homes, old-age centres, apartment blocks and informal settlements).

Within each secondary sample point, an interview referral point had to be identified. This achieved two objectives:

to allow the identification and selection of respondents; and



to provide an appropriate interview environment.

Finally, in order to limit any selection bias, the number of interviews conducted at each secondary sample point was restricted, by both time-delay and locality factors.

The questionnaire was refined through the experiences and results of the Durban and Johannesburg victim surveys. The pilot survey was thus restricted to being a training tool for refining interview techniques, measuring the length of the questionnaire and establishing a respondent sampling technique. During the pilot survey, 128 questionnaires were administered.14 The length of the Cape Town survey was marginally longer at 15 to 20 minutes than those in Johannesburg and Durban, due to the addition of gang-related questions. Clearly, respondents wanted to speak about their experiences and the longer than standard questionnaire did not reduce the response rate. On the contrary, fieldworkers often had difficulty in terminating interviews.

Coinciding with previous findings, it was confirmed that male enumerators could only interview men, while female enumerators could be utilised to interview both men and women. Of interest was the fact that cross-race interviews were undertaken with relative ease. This suggests that the issue of crime and violence transcends racial inhibitions. Despite this finding, and as a precaution, the race of the interviewer in the final survey matched that of the respondent.

Linked to the finding that enumerators dressed in an identifiable 'uniform', as opposed to ordinary casual attire, have a higher success rate both in terms of selecting respondents and the time taken to execute each subsample, it was decided to dress the field teams accordingly. Each fieldworker wore an identifiable T-shirt, cap and bag bearing the logo of DRA Development, the survey company, and the Institute for Security Studies.

An important component of the pilot study was the construction of a respondent selection technique that would allow the enumerators to obtain their necessary quotas, while ensuring that the selection process was both random and rigorously implemented. During the actual fieldwork stage, dummy respondents were used to ensure that these standards were being adhered to. As a further check, enumerators had to calculate the rate of flow of respondents (who roughly fell in the sample unit that was being targeted) at that specific referral point.

Stage 4

Fieldworkers were selected from a variety of sources, but were all residents of metropolitan Cape Town. The fieldworkers attended a half-day training course that was followed by a number of training interviews. These interviews were undertaken both in a controlled environment to check for knowledge and technique, and in the field to test interviewers' actual abilities. In addition to the training, the field teams were subjected to a half-day workshop on how to empathise with respondents who were victims and how to cope with the potential stress of being involved in such a study. A similar exercise was undertaken during the debriefing session that followed the research process.

In addition to the sensitivity counselling, fieldworkers were carefully selected from community-based structures, specifically from previously disadvantaged communities. People involved at the grassroots level largely had a prior awareness of the type of circumstances that existed within certain communities, which promoted a greater level of empathy with respondents.

In total, fourteen enumerators and two supervisors were divided into three teams that were used for the fieldwork process.

Stage 5

In the scan survey, a total of 3 839 interviews were undertaken during a seven-day period in January 1998 (as shown in the table below).