Background
This paper reports on the use of a “neighborhood method” to measure the nature and incidence of violence against women and girls in post-conflict Liberia.

Methods
The study population comprised females in Montserrado and Nimba counties. Study participants were randomly selected for interviews using multi-stage cluster sampling. 30 clusters of households were sampled in each county. Information on incidents of domestic violence and rape within the preceding 18 months was collected with regard to females of all ages in the respondent’s household, and those of her four closest neighbors to make up the full sample.

Findings
Households in the sample contained 7015 females (1687 girls, 4586 women, 742 age missing) in Montserrado and 6632 (2070 girls, 4167 women, 95 age missing) in Nimba. In the previous 18 months 54.1% (CI 53.1-55.1) and 55.8% (CI 54.8-56.8) of females in Montserrado and Nimba respectively were indicated to have experienced non-sexual domestic abuse; 19.4% (CI 18.6-20.2) and 26.0% (CI 25.1-26.9) of females in Montserrado and Nimba respectively to have been raped outside of marriage; and 72.3% (CI 70.7-73.9) and 73.8% (CI 72.0-75.7) of married or separated women in Montserrado and Nimba respectively to have experienced marital rape. Husbands and boyfriends were reported as the perpetrators of the vast majority of reported violence. Strangers were reported to account for less than 2% of the perpetrators of rape in either county. Incidents were most commonly disclosed to other family members or to friends and neighbors, and less often to formal authorities such as the police, court or community leaders. Incidents were approaching fifty times more likely to be reported to police if perpetrated by strangers rather than intimate partners.

Conclusions
Violence against women and girls is widespread in the areas studied. Programming needs to address the fact that this violence is primarily occurring in the household, where most incidents go unreported outside the immediate family or social circle. Police and hospital reports severely under-represent these known perpetrators. Inter-interviewer variance and differences in reports for self and neighbors for some outcomes caution the precision and validity of some estimates. However, the potential utility of the neighborhood method for estimating prevalence rates with an accuracy suitable for programmatic purposes in conflict-affected and post-conflict settings is noted.