How do interviewers and respondents navigate sexual identity questions in a CATI survey?

Abstract

Accurate measurement of sexual identity is needed in surveys to explore sexual minorities’ unique experiences in health, crime, politics, and education-related outcomes, among others. Previous research on interviewer-administered sexual identity questions (SIQs) has focused on data quality indicators derived from responses (e.g., “Don’t Know” or refusal answers) or has used lab-based cognitive interviews. However, these studies ignore interactional difficulties between interviewers and respondents that may occur prior to negotiating an acceptable answer, and lab studies lack generalizability. In this study, we code behaviors of interviewers and respondents during administration of an SIQ on Work and Leisure Today 2, a nationally representative telephone survey of US adults. We use these behavior codes to examine deviations from a paradigmatic “question asked/answered” interaction between the interviewer and respondent that arise because the SIQ is sensitive and/or confusing. We find evidence that the SIQ is sensitive: some respondents express concerns over the intrusiveness of the SIQ, while interviewers occasionally avoid probing unacceptable answers and sometimes skip administering the SIQ altogether. We also find that indicators of problems with the SIQ’s construction are more prevalent than sensitivity concerns. Respondents express confusion with the words used to categorize sexual identity, and mistake the SIQ’s nominal construction for a yes/no question. Interviewers sometimes deviate from reading the SIQ exactly as worded, perhaps in an attempt to repair the SIQ’s construction. Together, these issues highlight problems with interviewer-administration of an SIQ that may affect accurate classification of sexual identity.