A/HRC/7/6 page 14 route to track trends in reporting (are women, and younger women in particular, more likely to report?), recording (what is the gap between what women say and institutional data?) and the impact of policy changes. 49. Larger representative samples increase both the number of cases detected and the strength of claims to accuracy, especially where accepted standards for random selection are used. The latter are, however, difficult to implement in societies where the tools for selection - census data, voting registration and/or high telephone ownership - are unavailable or inaccurate. While geographic spread across a country is the most preferable, funds, transport and political contexts may make this impossible. Decisions such as that made in the WHO multi-country study to randomly sample in two areas - one urban and one provincial/rural - represent the best possible compromise where resources are limited. 50. How surveys are administered has implications for disclosure. The early orthodoxy was that face-to-face contact, with interviewers matched for sex, age and ethnicity, increased disclosure. Subsequently, other methods have produced high disclosure rates, and confidentiality appears to be as, if not more, important, as rapport.19 Both specialist surveys on violence against women and carefully designed modules in health or crime victimization surveys have proved effective. While surveys raise awareness about violence against women and produce methodological innovation, they are expensive, and no State has yet committed itself to conducting them on a regular basis, which is necessary for indicators. At the same time, modules have drawbacks, especially the danger of progressive reduction across two dimensions of what we currently know to be good practice - multiple questions and covering a range of forms of violence against women.20 51. An example of a module that combines five key outcome and process indicators and addresses trends is the one on female genital mutilation, used in rapid assessment tools (the Demographic and Health Survey and the Multiple-Cluster Indicator Survey) in 14 countries.21 52. The module includes: • Prevalence of female genital mutilation The key outcome measure, calculated across five-year age bands for women aged 15-49 • Genital mutilation status of all daughters A second outcome measure, tracking recent change, including the current age of daughters, whether and at what age they were cut 19 Walby, S. (2005) “Improving the statistics on violence against women”, Statistical Journal of the United Nations, UNECE, 22, 4: 193-216. 20 UNECE Task Force on Measurement of Violence Against Women, Analysis of national surveys (ECE/CES/GE.30/2006/6). 21 See www.measuredhs.com/topics/gender/FGC-CD/start.cfm; African Child Policy Forum (2006), Born to High Risk, Violence Against Girls in Africa.