Using a list experiment to measure intimate partner violence: Cautionary evidence from Ethiopia

While indirect methods are increasingly widely used to measure sensitive behaviors such as intimate partner violence in order to minimize social desirability biases in responses, in developing countries the use of more complex indirect questioning methods raises important questions around how individuals will react to the use of a more unusual and complex question structure. This paper presents evidence from a list experiment measuring multiple forms of intimate partner violence within an extremely poor sample of women in rural Ethiopia. We find that the list experiment does not generate estimates of intimate partner violence that are higher than direct response questions; rather, prevalence estimates using the list experiment are lower vis-à -vis prevalence estimates using the direct reports, and sometimes even negative. We interpret this finding as consistent with “fleeing†behavior by respondents who do not wish to be associated with statements associated with intimate partner violence.