Skip to content
Back to search results

Gender-Based Violence Survey

Introduced by
WHO (García-Moreno, Jansen, Ellsberg, Heise & Watts) and the DHS Program
Year
2005
Type
Population-based prevalence survey
Family
Violence measurement
Font size

What does it do?

Gender-based violence (GBV) and violence-against-women prevalence surveys are standardized, population-based instruments designed to estimate how many people — overwhelmingly women — experience physical, sexual, and emotional violence by intimate partners and others.

The dominant designs are the WHO Multi-country Study on Women's Health and Domestic Violence and the Demographic and Health Surveys (DHS) domestic-violence module. They measure violence through behaviorally specific acts rather than the word 'violence', and they are conducted under strict ethical and safety protocols that govern who is interviewed, how, and where.

The intuition behind the method

Asking 'have you experienced violence?' fails for the same reason asking 'were you raped?' fails: it depends on respondents labeling and disclosing stigmatized, often normalized experiences. GBV prevalence surveys instead ask about concrete acts — being slapped, pushed, forced to have sex, threatened, humiliated — and let the researcher decide what counts as physical, sexual, or emotional violence. The harder problem these surveys solve is that the act of measuring can itself endanger the respondent, so the methodology is built as much around safety as around statistics: only one woman per household is interviewed, privacy is guaranteed, and interviewers are trained to respond to distress and never to raise the topic where a partner might overhear.

History

The modern template was set by the WHO Multi-country Study on Women's Health and Domestic Violence, whose initial results were published in 2005 and whose intimate-partner-violence findings appeared in The Lancet in 2006, led by Claudia García-Moreno and colleagues. It established behaviorally specific act measurement, the single-respondent-per-household rule, and a formal ethical and safety protocol as the field standard. The DHS Program incorporated a closely aligned domestic-violence module into its nationally representative surveys, extending comparable measurement to dozens of countries. Together these instruments underpin most global statistics on violence against women and the indicators used for Sustainable Development Goal 5.

When to use it

Use a GBV prevalence survey when you need population-representative estimates of the extent of violence against women for monitoring, advocacy, or policy, especially where international comparability or Sustainable Development Goal reporting is required. The WHO and DHS designs are appropriate for national and subnational prevalence, trend monitoring, and linking violence to health outcomes. They are not appropriate for case identification of individuals, for settings where the ethical and safety protocol cannot be guaranteed, or where the goal is to study a perpetrator population or detailed psychological mechanisms rather than population prevalence.

Limitations and considerations

  • Even with best-practice methods, stigma and fear cause substantial underreporting, so estimates are lower bounds.
  • Most large studies interview only women, leaving male and other-gender victimization and perpetration largely unmeasured.
  • Differences in act lists, reference periods, and partner definitions limit comparability across non-standardized instruments.
  • The required safety protocol is resource-intensive and difficult to maintain in conflict, displacement, or low-privacy settings.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Behaviorally specific, act-based items yield far higher and more valid disclosure than asking about 'violence' directly.
  • Standardized WHO and DHS instruments allow comparison of prevalence across countries and over time.
  • Probability sampling with weights produces estimates that generalize to the whole population.
  • Built-in ethical and safety protocols protect respondents and improve data quality simultaneously.

Applications

  • Producing national and global prevalence estimates of intimate-partner and sexual violence against women.
  • Monitoring progress on Sustainable Development Goal 5 indicators on violence against women.
  • Linking experience of violence to health, reproductive, and mental-health outcomes in epidemiological analysis.
  • Informing the design and evaluation of prevention programs and survivor support services.
  • Supporting advocacy and legal reform by quantifying a previously hidden problem with credible data.

How it works

How it works

1

The questionnaire is built from behaviorally specific act items grouped into types of violence: physical (slapped, pushed, hit, choked, threatened or hurt with a weapon), sexual (physically forced sex, coerced sex through fear, degrading or humiliating sexual acts), and emotional or psychological (insults, belittling, intimidation, threats). Items distinguish acts by an intimate partner from those by non-partners, and ask about both lifetime and a recent reference period (commonly the past 12 months). The WHO study standardized these items so estimates would be comparable across very different countries, and the DHS domestic-violence module embeds a closely related set within a larger health survey.

2

A probability sample of households is drawn so that estimates can be generalized to the population, typically through a stratified multi-stage cluster design with sampling weights. Crucially, only one eligible woman is randomly selected and interviewed per household. This single-respondent rule is both a statistical and a safety measure: it prevents within-household clustering on the violence question and ensures that no two women in the same home are seen to be discussing violence, which could expose them to retaliation.

3

Administration follows formal ethical and safety guidelines. Interviewers are extensively trained, are themselves women, and conduct the interview in complete privacy; the survey is presented with a neutral cover topic (women's health) so that neither the partner nor the community knows its true subject. The interview is suspended if privacy is lost, respondents can decline any question, and interviewers carry referral information for support services and are trained to manage distress. These protocols are treated as a precondition for the survey, not an add-on.

4

Responses are coded into the violence-type categories and combined with the sampling weights to estimate population prevalence — the proportion of women who experienced at least one qualifying act over the lifetime or the past 12 months — usually reported separately for intimate-partner and non-partner violence and disaggregated by age, region, and other characteristics. Because the design is complex, variance estimation must account for stratification, clustering, and weights to produce correct confidence intervals.

Common pitfalls

  • Comparing prevalence across surveys that used different act items, reference periods, or partner definitions as if equivalent.
  • Running a GBV survey without the full ethical and safety protocol, endangering respondents and biasing disclosure.
  • Ignoring the complex sampling design when computing standard errors, producing falsely narrow confidence intervals.
  • Reading a low reported rate as low true prevalence rather than as suppressed disclosure under unfavorable conditions.

Frequently asked questions

Why is only one woman per household interviewed?

It serves both safety and statistics. Interviewing a single randomly selected eligible woman ensures that no two women in the same home are visibly drawn into a conversation about violence, which could provoke suspicion or retaliation from a partner. Statistically, it also avoids clustering of the violence outcome within households, keeping the sample design clean. The selection is randomized so the chosen respondent represents eligible women in that household.

Why use behaviorally specific acts instead of asking about 'violence'?

Many respondents do not label coercion, slapping, or forced sex as 'violence', especially where such behavior is normalized, and stigma further suppresses disclosure of the labeled concept. Asking about concrete acts — being hit, threatened, forced to have sex — lets respondents report what actually happened and lets the researcher classify it. This consistently yields higher, more valid prevalence estimates and makes results comparable across cultures and languages.

Are these survey estimates the true rate of violence?

No; they are best understood as lower bounds. Even with behaviorally specific items, trained female interviewers, and guaranteed privacy, fear, shame, and recall limits cause underreporting. The methodology is designed to minimize that gap, not eliminate it. Estimates should therefore be read as conservative, and any apparent decline must be checked against changes in willingness to disclose, not assumed to reflect a real fall in violence.

Related methods

Sources

  1. García-Moreno, C., Jansen, H. A. F. M., Ellsberg, M., Heise, L., & Watts, C. H. (2006). Prevalence of intimate partner violence: Findings from the WHO multi-country study on women's health and domestic violence. The Lancet, 368(9543), 1260–1269.

    Open DOI reference
    Copied.
  2. World Health Organization (2005). WHO Multi-country Study on Women's Health and Domestic Violence against Women: Initial Results on Prevalence, Health Outcomes and Women's Responses. WHO, Geneva.

    Open publication reference
    Copied.