The effect of interview method on self-reported sexual behavior and perceptions of community norms in Botswana

Since self-reports of sensitive behaviors play an important role in HIV/AIDS research, the accuracy of these measures has often been examined. In this paper we (1) examine the effect of three survey interview methods on self-reported sexual behavior and perceptions of community sexual norms in Botswana, and (2) introduce an interview method to research on self-reported sexual behavior in sub-Saharan Africa. Comparing across these three survey methods (face-to-face, ballot box, and randomized response), we find that ballot box and randomized response surveys both provide higher reports of sensitive behaviors; the results for randomized response are particularly strong. Within these overall patterns, however, there is variation by question type; additionally the effect of interview method differs by sex. We also examine interviewer effects to gain insight into the effectiveness of these interview methods, and our results suggest that caution be used when interpreting the differences between survey methods.