This paper analyzes the magnitude andpredictors of misreporting on intimate partner and sexualviolence in Nigeria and Rwanda. Respondents were randomlyassigned to answer questions using one of three surveymethods: an indirect method (list experiment) that givesrespondents anonymity; a direct, self-administered methodthat increases privacy; and the standard, directface-to-face method. In Rwanda, intimate partner violencerates increase by 100 percent, and in Nigeria, they increaseby up to 39 percent when measured using the list method,compared with direct methods. Misreporting was associatedwith indicators often targeted in women's empowermentprograms, such as gender norms and female employment andeducation. These results suggest that standard surveymethods may generate significant underestimates of theprevalence of intimate partner violence and biasedcorrelations and treatment effect estimates.