Do Third-Party Reports Improve the Measurement of Sensitive Behaviors? A Validation Study Using Social Network and HIV Biomarker Data

James Mkandawire, University of Malawi
Stephane Helleringer, Columbia University

Third-party reports (e.g., best friends or confidants) have recently been suggested as an alternative approach to measuring sensitive behaviors such as abortions or multiple sex partnerships (MSP). The validity of this method has however not been systematically assessed. During the Likoma Network Study (Malawi), 2,014 respondents were asked whether their best friend had MSP. We then traced the nominated best friend in population rosters and linked all reports. For a large fraction of study participants (>75%), we thus have data on MSP from both one’s own self-reports and their best friend’s report. We use these data to 1) determine whether best friends constitute a representative sample of the population and 2) compare self-reports to best-friend reports of the MSP of a given participant. We test whether HIV/STI infections are more common among participants who did not self-report MSP, but who were classified as having MSP by their best friend.