Assessing Potential Bias in Respondent-driven Incident Based Data from a Web Survey of College Students
Marcus Berzofsky
RTI International
Chris Krebs
RTI International
Christine Lindquist
RTI International
Incident-level data collection is a useful approach when measuring events that can occur multiple times within a survey's reference period. Incident-based data allow survey researchers to analyze not just characteristics of persons but also characteristics of incidents. Survey practitioners often cap the number of incident reports required for each respondent to reduce burden. If reported incidents differ from those not covered in the survey instrument, then bias potentially exists. The Campus Climate Survey Validation Study (CCSVS) was a web based survey administered at nine colleges that collected prevalence and incident-based information on unwanted sexual contact. The CCSVS capped the number of incident reports at three and allowed respondents to determine the order in which incident reports were completed. To assess the potential for bias, we determine whether respondents systematically ordered the reported incidents. We consider incident ordering based on the chronological order and severity of incidents. Also, we assess whether respondents who were unsure of the month in which one of their incidents occurred reported those incidents in a systematic way.