: This section of the manual draws extensively on the United Nations publication Guidelines for Producing Statistics on Violence against Women: Statistical Surveys
(2014). Readers are strongly encouraged to consult the above-mentioned publication, as more details as well as additional information are covered at length.
- + Uses of surveys on violence against women for gender statistics
- Surveys on violence against women have a unique role in measuring the extent, nature and consequences of all types of violence against women. Data collected in these surveys are necessary (a) to estimate the prevalence of violence against women; (b) to identify the groups of women most at risk of violence; (c) to identify the characteristics of perpetrators; (d) to estimate the impact of violence on women’s physical and mental health; (e) to identify the barriers faced by victims of violence in seeking support and services; and (f) to understand attitudes towards violence against women. This information is crucial to efforts of prevention and response to violence against women.
Surveys on violence against women are an effective tool to measure the prevalence of physical, sexual, psychological and economic violence. Data are usually collected for four types of violence against women: physical, sexual, psychological and economic abuse. In dedicated surveys, these types of violence can be examined in depth, through detailed questions. Some other surveys that include modules on violence against women may not be able to include all types of violence. When surveys have limited space to accommodate questions on violence, psychological and economic abuses are often left out. On the other hand, administrative records of data on violence against women are characterized by massive underreporting of violence, and therefore cannot be used as a reliable source of data for measuring the prevalence of violence against women.
Surveys on violence against women collect data necessary to identify the groups of women most at risk of violence. In particular, dedicated surveys have the advantage of using samples designed to represent various groups of women in sufficient numbers to allow the calculation of specific prevalence rates for each group. In all societies, women and girls from various social groups can be victims of violence. In many countries, there is even a cultural acceptance of violence. Not all groups of women have an equal risk of being abused, however. Surveys on violence against women can identify the groups most at risk by collecting information on the characteristics of women who experience violence and those who do not. Characteristics usually measured include age, age at first marriage, marital status, educational attainment or literacy, economic activity, place of residence, ethnicity, language and religion. Other data on women’s awareness of legal rights or on characteristics showing women’s ability to access resources and live independently (such as regular and stable cash income or the ownership of property) may also be useful to understand women’s vulnerability to violence by intimate partners or other perpetrators.
Detailed information on perpetrators covering all types of violence and all types of victims are often collected only in dedicated surveys focused on violence against women. Data on characteristics of victims and perpetrators serve as the basis for designing strategies for the prevention of violence and early intervention. Data on the relationship to the perpetrator is a key element of any survey on violence against women. This characteristic serves as the basis for identifying perpetrators who are intimate partners (current or former) and other perpetrators (such as various types of relatives, acquaintances, friends, co-workers, local authorities or others). Slightly different data are usually collected on intimate partners and on other perpetrators. For intimate partners, a minimum recommended list of characteristics includes age, educational attainment and literacy, economic activity, substance abuse and witnessing partner violence in the family of origin. These characteristics should be collected for the current partner as well as for the most recent previous partner. Characteristics of non-partner perpetrators that could be accurately recalled by respondents include the sex and relationship of the perpetrator and the location where the violence occurred.
Surveys on violence against women are a source of representative and comprehensive data on the impact of violence against women. Data collected may refer to frequency of violence by type of violence, types of physical injuries and the need to seek medical attention. Additional data on mental health impacts, relying on self-evaluation, may refer to being afraid or depressed, experiencing anxiety and suicidal thoughts, being unable to perform paid or unpaid work or turning to alcohol or drugs. It is also important to identify violence that took place at specific life stages, such as during pregnancy, and whether the violence resulted in miscarriage. All this information is crucial in order to plan appropriate services for victims and effective responses to the violence.
Surveys on violence against women may be the only source of data on the proportion of victims of violence seeking help and receiving informal and institutional support and services. Barriers faced by victims of violence in accessing support and services can be identified on the basis of data on victims seeking help from family, friends or institutions and data on using medical or social services, combined with data on individual and household characteristics. The information obtained is crucial to the development of policy and programmes that more effectively respond to the needs of the victims of violence, such as emergency housing, counselling, medical assistance, the prevention of sexually transmitted diseases and unwanted pregnancy.
Surveys and survey modules on violence against women can collect data on the extent to which violence is tolerated in the wider community. The information obtained can be used to design strategies of zero tolerance towards violence against women directed at the general public or targeted at groups of women most at risk of violence.
Additional sources can be used by national statistical offices to collect data on violence against women. Data on the prevalence of violence and additional related topics can be collected in modules attached to more general household sample surveys focused on health, such as DHS and MICS, or crime, such as victimization surveys. Considering the sensitivity of the topics addressed, health surveys in particular provide a feasible vehicle for a module of questions measuring women’s experiences of violence. Although these surveys may not be able to accommodate the detailed range of questions needed to study violence against women in all its complexity, they may be a practical alternative for measuring violence against women when the budget is limited.
Other sources of data regarding violence against women are administrative and include but are not limited to records from health and medical services, agencies of the criminal and civil justice systems, social services, legal aid services, research and documentation centres, and services designed specifically to respond to women who have experienced violence, such as shelters or refuges, rape/sexual assault centres, crisis telephone lines, women’s groups and advocacy organizations and women lawyers’ associations. These agencies and organizations collect data about women experiencing violence as part of their daily recording practices. The data collected provide an indication of societal responses to reported cases of violence against women and available services for victims. Acts of violence against women go highly unreported, however, and those that are brought forward to agencies and services are often among the most severe and affect the most disadvantaged women. Therefore, data collected are not representative of the full extent and nature of the problem and cannot be used to measure the prevalence of violence against women taking place within a population.
- + Avoiding gender bias in data collection
- The following paragraphs focus on data collection issues specific to surveys on violence against women, organized into four categories: concepts and measurement; questionnaire design; sample coverage; and the selection and training of interviewers.
- + Concepts and measurement
- Prevalence of violence against women should be measured using questions listing different acts of violence for each type of violence. Detailed, behaviour-specific questions are needed to ensure that respondents consider a wide range of acts in their responses that they might otherwise omit if they are left to speculate about the meaning of a more general term. For example, physical violence encompasses all acts of intentional force that have the potential to cause injury or death (not including forced sexual contact). Questions measuring the prevalence of physical violence should lists acts such as hitting with a fist, open hand or hard object; kicking; biting; pushing; squeezing; pulling hair; repeated hits that result in beating; choking; burning; holding the person down; threatening to use a gun, knife or other object used as a weapon; and actually using a weapon. This list and other suggested lists of behaviours needed to measure each other type of violence (sexual, psychological and economic) as well as the definitions of all concepts used are presented in the United Nations publication Guidelines for Producing Statistics on Violence against Women: Statistical Surveys (2015).
Severity of violence should not be assumed based on specific acts of violence. Similar acts of physical violence can have very different consequences owing to a number of circumstances (prolonged violence as opposed to a single incident, for example). Severity should be assessed separately according to the consequences suffered by the victim, including frequency, injury and other impacts that make it difficult for the woman to carry out usual daily activities. Emotional consequences, such as fear of the perpetrator, for example, could also be incorporated as indicators of the severity of intimate partner violence.
Broad concepts of marital status and intimate partners should be used in order to adequately estimate the prevalence of violence by intimate partner. The recording of marital status should extend beyond relationships that are socially or legally sanctioned and cover forms of informal unions specific to each country. In addition, the definition of intimate partner should not be limited to legally married partners, but kept as broad as possible and include, for example, boyfriends and dating partners. Testing of the questions and training of interviewers is essential to ensure that responses are recorded accurately. Probing questions should be used if necessary.
- + Questionnaire design
- The questionnaire design is instrumental in ensuring participation in the survey; establishing rapport and trust; ensuring the disclosure of sensitive information and the safety of the respondent; and reducing the risk of emotional trauma. Of particular importance are the presentation of the survey, the introductory sections of the questionnaire, the sequence of questions in the questionnaire and the wording used.
The presentation of the survey in the initial approach to a selected household and subsequently to a selected respondent should not put the respondent’s safety in jeopardy. It is common practice to introduce the survey to the household not as a survey on violence against women but as a study on women’s health, safety or other issues of particular relevance to women, without specifically mentioning violence. This is done to avoid raising suspicion from other family or community members and to encourage participation in the survey. Interviewers will not know in advance if the woman has experienced violence or is currently living in an unsafe situation; therefore precautions must be taken at all stages not to disclose the specific objectives of the survey to anyone in the household or community.
Introductions to various sections of the questionnaire are essential for building rapport and trust, ensuring disclosure and setting the context. At the beginning of the questionnaire, the respondent should be provided with information about measures that have been put in place to ensure her safety while she responds to sensitive questions, such as rescheduling to another time or switching to a neutral questionnaire if a family member comes on the scene. Careful introductions should be designed for the sensitive sections. For example, sensitive questions on intimate partner violence should be introduced by first assuring respondents that relationships have both good and bad moments in order to avoid the risk that only socially desirable responses are given. At the same time, the respondent should be reassured that their answers will remain confidential and that participation in the survey remains voluntary. Introductions are also important for setting the context of a particular section in the questionnaire. For example, an introduction to questions about violence involving individuals other than intimate partners should help the respondent consider a wide range of actors, including violence by male or female relatives, friends and acquaintances or strangers.
The flow of the questionnaire should be from topics and questions that are non-threatening and less sensitive to topics and questions that are more personal and sensitive. This gradual transition is critical to the disclosure of experiences of violence, reducing the risk of a participant prematurely terminating the interview and reducing the risk of emotional trauma. Questions about family composition, health, social networks, personal characteristics and partner characteristics can be situated prior to modules about experiences of violence. Situating questions about psychological and economic abuse by partners prior to questions about physical and sexual violence is important for providing a gradual lead-up to these more sensitive questions. It is good practice to place questions about forced sexual acts after questions about physical assaults because of the difficulty many women will have disclosing these very personal experiences to an interviewer. Questions concerning childhood experiences of violence may also be placed towards the end of the questionnaire. Within each section covering a type of violence, the behaviour-specific questions should also range from less serious to more serious acts of violence.
Careful wording is essential to reducing underreporting of experiences of violence. As a general principle, value-laden and stigmatizing terms such as “rape” and “violence” must be avoided in the questions or introductions as they lead to underreporting of experiences of violence. Instead, detailed, behaviour-specific terms, which yield much higher disclosures of identical experiences, should be used.
Survey questions and introductions to various sections of the questionnaire should cue respondents to consider a variety of different settings (for example, home, work, school and outside locations) and specific categories of perpetrators (for example, current partners, former partners, other male relatives and males in positions of authority).
The drafting of the introductory sections of the questionnaire and the development of the right wording of the questions can greatly benefit from focus group discussions engaging local women. Preferably, focus groups should include women identified as experiencing partner violence and sexual violence. These women can be recruited from shelters and counselling groups, for example. It is also important that women participants in the focus group represent various social groups, including immigrants and other minority groups.
- + Sample coverage
- All significant subgroups of population should be represented in the sample in numbers large enough to allow analysis at their levels. The expected prevalence of types of violence in the sample population and the requirement to conduct analysis for subgroups of population, such as different age groups, geographic areas or ethnic minority groups, are important factors in deciding the sample size and sample design. In order to produce prevalence estimates of specified precision for subgroups, the sample size will have to be increased considerably or higher sampling fractions may need to be applied to the smaller subgroups.
The method of data collection should not prevent the possibility of covering specific groups of women. Women living in violent situations may be prevented from using the telephone or may not take the risk of undergoing detailed questioning over the telephone if their movements are monitored by a violent partner. As another example, web-based and disk-by-mail surveys require wide-ranging computer ownership, Internet access and literacy.
It is recommended that only one woman be interviewed in each household, for safety reasons. Surveys on violence against women that interview all women in the household should take extra precautions to ensure the safety and privacy of all respondents.
- + Selection and training of interviewers
- All interviewers for a survey on violence against women must be female. Using female interviewers increases disclosure of sensitive information, particularly experiences involving sexual victimization and violence perpetrated by male partners. Field supervisors must be female in face-to-face interviewing situations since they must travel with interviewers periodically to oversee their work. When forming teams of field workers, however, project managers must take into account local norms prohibiting women from working in public spaces and other barriers female interviewers may face when approaching households to obtain an interview. In these situations, teams of male and female interviewers working in tandem have been shown to improve household contact, lower refusal rates and be beneficial in ensuring the safety of female interviewers. In telephone situations, it is possible to have male supervisors if the supervisors do not have contact with respondents and if the supervisors have the required skills and training to further train interviewers and support them through regular debriefings.
Age is an important factor to consider when selecting interviewers. Interviewers that are seen as too young may experience distrust from respondents, which can lead to outright refusal to participate in the survey or reluctance to disclose personal information. In some locations, it would be considered inappropriate for a young woman to pose questions related to violence to an older woman. While there is no recommended specific age limit or range applicable across all settings, many respondents perceive older female interviewers as warmer and more reassuring than younger women.
Other characteristics of interviewers that may affect participation and disclosure rates should be taken into account when selecting interviewers and during training. Particularly important to obtaining accurate information from the respondents are non-judgemental and empathetic attitudes, good interpersonal skills and sensitivity and knowledge of gender issues and issues related to violence against women. These skills should be further developed through extensive training.
Training of interviewers is a crucial step in reducing bias in data collection. Training is essential to ensure that interviewers understand the purpose of the survey and know the structure of the data collection instrument; are skilled in conducting the interview and developing a rapport with the respondent; understand the ethical requirements of the survey, including confidentiality, safety and support for respondents; are aware of the possible dangers women face when responding to questions concerning their experiences of violence; are able to ensure the safety and emotional well-being of respondents; and guard the confidentiality of the information collected. All interview training should be in accordance with the guidelines for addressing ethical and safety issues in research on violence against women established by the World Health Organization.
Sensitivity training is an important component of interviewer training. Sensitivity refers to an interviewer’s ability to pose very delicate questions about experiences of violence in a respectful manner, accurately assess the feelings or reactions of respondents in a variety of situations and respond appropriately. Sensitivity training must facilitate the understanding of gender issues related to violence, the dynamics and causes of male violence against women and the impacts of violence on women’s health and well-being. Training techniques can be used to reduce the chances that interviewers will respond with judgmental or blaming comments when interacting with respondents. Such techniques include listening to and discussing in-depth testimonies of abused women and advocates for abused women; the discussion of general or local stereotypes, prejudices and myths regarding female victims of violence; the discussion of an interviewer’s own biases and stereotypes concerning women who have experienced violence; and role playing. Role-play scenarios are particularly important in preparing the interviewers for situations in which women refuse to participate in the survey, or situations that may compromise the safety and well-being of respondents and the confidentiality of information (such as interruptions of the interview by other household members, including threatening husbands, or emotional reactions by respondents to the disclosure of experiences of violence).
Interviewers should be trained to control the interviewing environment in order to ensure the safety and privacy of respondents and the confidentiality of the information disclosed. The potential that women may be put at risk of violence for participating in a survey cannot be underestimated. It is essential that respondents be in a position to answer freely and without fear of repercussions. As mentioned before, selecting only one female member of the household to be a respondent is one of the mechanisms used to ensure that women who have experienced violence from intimate partners or other family members are able to respond in a manner that will not jeopardize their safety. Other strategies oriented towards the safety and privacy of respondents should be developed by survey designers and covered in interviewer training. It is recommended that interviews should take place at a time when other household members, particularly male partners, are not present. Interviewers should be prepared to switch to a neutral questionnaire if a household member comes on the scene. In some contexts, it may be advisable for escorts accompanying the primary interviewer to conduct interviews with other family members to distract them from the main focus of the survey. When the privacy of the respondent cannot be ensured, the interview should be rescheduled. In telephone surveys, respondents should be offered a toll-free telephone number they can use if they have to hang up suddenly or want to continue the interview at another time. Interviewers can establish a safety plan with respondents so that respondents can stop an interview at any time if they feel unsafe. Interviewers should also check in with respondents periodically during the interview to confirm that they are able to proceed.
Interviewers should be trained to identify and respond properly to respondents’ emotional trauma. Given the personal and delicate nature of the information requested in surveys on violence against women, respondents can be expected to react in a wide range of ways. Some respondents may be open to disclosing their experiences and may view the survey as an opportunity to allow them to make their experiences known. Others, however, may be fearful that a violent partner might learn of their participation in the survey, feel disturbed by the content of the interview, be traumatized by recent experiences of violence or feel embarrassed or stigmatized when disclosing their experiences. Interviewers should be trained in ways to properly react to this multitude of possible reactions. It is important that the reaction to emotional distress is in a warm, empathetic but neutral manner. Interviewers must be instructed not to counsel respondents themselves. They should be able to refer respondents to a list prepared in advance consisting of agencies in the local community who can provide assistance. Where few resources exist locally, the survey design may need to take into account the development of short-term support mechanisms.
Interviewers should be trained to identify their own emotional reactions and reduce their own stress. In surveys on violence against women, interviewers will be engaged in emotionally draining work. Interviewers should be trained to identify their own emotional reactions to the numerous disclosures about violence and helped to develop skills that manage and reduce their stress. Supervisors should also be trained to recognize emotional trauma among interviewers during fieldwork. Strategies to prevent burn-out should be considered, such as regular debriefings by supervisors or counsellors specially engaged as part of the project team, or offering interviewers the chance to participate in less taxing administrative tasks between interviews.
- + References
- Andersson, Neil, and others (2009). Collecting reliable information about violence against women safely in household interviews: experience from a large-scale national survey in South Asia. Violence against Women, Vol. 15, No. 4 (April), pp. 482–496.
Ellsberg, Mary, and Heise, Lori (2005). Researching Violence against Women: A Practical Guide for Researchers and Activists. Geneva: World Health Organization; Washington, D.C.: Program for Appropriate Technology in Health.
Ellsberg, Mary, and others (2001). Researching domestic violence against women: methodological and ethical considerations. Studies in Family Planning, Vol. 32, No. 1 (March), pp. 1-16.
___ (2007). WHO Multi-country Study on Women's Health and Domestic Violence: facilitator's manual - workshop for training fieldworkers. Geneva: World Health Organization.
Jansen, Henrica A. F. M., and others (2004). Interviewer training in the WHO Multi-country Study on Women’s Health and Domestic Violence. Violence against Women, Vol. 10, No. 7 (July), pp. 831-849.
__ (2007). WHO Multi-country Study on Women's Health and Domestic Violence: Supervisor's and field editor's manual, revised 27th ed. Geneva: World Health Organization.
Johnson, Holly, Natalia Ollus and Nevala, Sami (2008). Violence against Women: An International Perspective. New York: Springer.
Paletta, Anna and Karen Mihorean (1998). Cognitive testing of questions to measure family violence. Ottawa: Statistics Canada.
United Nations (2005). Report of the Expert Group Meeting on the theme “Violence against women: a statistical overview, challenges and gaps in data collection and methodology and approaches for overcoming them.” Geneva, 11-14 April.
___ (2015). Guidelines for Producing Statistics on Violence against Women: Statistical Surveys. Studies in Methods, Series F, No. 110. Sales No. E.13.XVII.7.
United Nations Office on Drugs and Crime and United Nations, Economic Commission for Europe (2010). Manual on victimization surveys. Geneva. ECE/CES/4.
Walby, Sylvia and Andrew Myhill (2001). New survey methodologies in researching violence against women. British Journal of Criminology, Vo. 41 No. 3 (June), pp. 502–522.
Watts, Charlotte, and others (2007a). WHO Multi-country Study on Women's Health and Domestic Violence: Interviewer's manual, revised 18th ed. Geneva: World Health Organization.
___ (2007b). WHO Multi-country Study on Women's Health and Domestic Violence: Study protocol. Geneva: World Health Organization.