Measuring Violence Against Women with Experimental Methods

The prevalence of intimate partner violence is a central indicator of the Sustainable Development Goals for women's agency. However, measuring this indicator largely relies on self-reports that could suffer from severe misreporting if women face high costs of revealing their victim status. We study the degree of misreporting in surveys that have been identified as the best source of data, such as the widely used Demographic and Health Surveys (DHS). Focusing on a sample of women in impoverished urban areas of Lima, Peru, we conduct an experiment that replicates direct measures from these surveys and compares them against list experiments, a method that provides greater privacy to respondents. We find no significant differences across direct and indirect methods in any of the seven acts of physical and sexual violence considered. This result largely persists when testing across sixteen different subgroups and accounting for multiple hypothesis testing.