Two researchers conducted a total of 27 semi-structured interviews with trained interviewers. Each interview lasted 60 minutes. All interviewers were from a single organization, with representation from regional offices across the United States. All interviewers were trained in the standardized interviewing approach. All interviewers were notified that their responses would be completely anonymous and confidential, and that only the research team would have access to their responses. Interviewers were paid their usual hourly rate for participating in this research. The interviewers had a wide range of experience interviewing, from under 1 year to over 15 years. Most had experience across multiple surveys, spanning topics such as employment, health, housing, crime, and household expenditures.
The protocol (see Online Appendix 5A) started with introductions and an explanation of the purpose of the interviews. Afterward, topics ranged from how interviewers approach asking sensitive and difficult questions, deciding whether and how to probe, and general aspects of interviewer-respondent interactions. Due to time constraints, not all questions in the protocol were asked. Afterward, the researchers administered fictional vignettes (see Online Appendix 5B) depicting common interviewer-respondent interactions. The goal of using these vignettes was to get feedback on how interviewers would handle scenarios like these in the field. The vignettes varied in their level of sensitivity and difficulty (e.g., asking a respondent who was recently laid off about their employment status). Interviewers rated the sensitivity and difficulty of each vignette and then provided feedback on how they would approach each scenario in the field. The resulting qualitative data were analyzed with a grounded theory approach to identify themes and patterns and to gain a deeper understanding of the range of interviewer responses (Glaser and Strauss
Interviewers perceived questions about income, children, housing/assets, disability, taxes, where respondents were born, and other demographics as the most sensitive topics from the respondent's point of view. Interviewers reported that they believed respondents seem to find these questions sensitive because they are highly personal or because respondents believe they are none of the government's business. Interviewers reported using techniques such as distancing, apologizing, and repeating the question to help administer questions that they anticipate respondents will find sensitive. Several interviewers reported being able to anticipate which respondents will find income questions sensitive. In those cases, some interviewers reported skipping the scripted question and asking immediately for an income range (e.g., "I don't need to know an amount, just give me a category").
Methods for administering sensitive questions. Distancing themselves from the survey was common as many interviewers reported showing respondents their CAPI screen during in-person interviews. This seemed to accomplish two goals: (1) proving that the question is real and that they are required to ask it; and (2) allowing respondents to select a category without having to tell the interviewer the answer, thereby turning the question into a quasi-self-administered item. For example: "I ask for their best estimate, or let them pick a range from the [CAPI] screen. Once they start, they're usually pretty cooperative" and "I show the screen to respondents to show them what the question and response options are. For income questions, the screen is evidence that we aren't looking for a precise dollar value, just categories."
Interviewers also distanced themselves from the survey by simply informing respondents that they are required to read all questions as worded as mandated by the survey organization:
"I'm required to ask all questions but you aren't required to answer;" and "[I will say to respondents] the question is written like this and it's not up to me; I am low on the totem pole. It's okay to refuse to answer, but your response is helpful data."
Several interviewers mentioned that respondents find questions sensitive because they have confidentiality concerns. Once the interviewers explain how the data are used and the penalties for breaches, more respondents become assured. For example: "Most people are concerned about their confidentiality. We explain their names are stripped from the data and they are just a statistic. Once you explain how it's used, most people don't care;" and "I tell respondents that data are confidential and represent many other households in the aggregate. I mention that there is a fine of $250,000 for breaking confidentiality."
Other interviewers make sure respondents know that they do not have to answer every question. For instance: "I say to all respondents before the interview starts that if there's something you don't want to tell me, we can just skip over it. I am the only one who will know the response, no one else will."
Other interviewers used lead-ins to questions to help ease respondents' observed hesitation: "I understand you might be uncomfortable..." and "If there are any questions that make you uncomfortable, just say 'No thank you.'"
Questions perceived as sensitive to ask. Interviewers identified questions related to sexual orientation, gender identity, disability, and religion as being sensitive for them to administer. These questions or topics can be awkward for interviewers to ask, so they sometimes modify those questions, again using distancing and apologizing lead-ins: "[For in-person interviews] it's awkward to ask 'Are you male or female?' in front of them. I add, 'I'm sorry I have to ask everything as worded.' Then they laugh and answer. It's easier to do over the phone;" and "The biological children question. Some people get offended you'd even ask it. I wouldn't answer such a question either; it's too personal. I always stress the confidentiality of the question."
Questions may also be sensitive given the context of the survey interview. There are situations where interviewers may encounter unanticipated sensitivity, and have to think in real-time about how to handle the question-asking process, as several interviewers explained: "Asking about sexual orientation and gender identity was recently added to the survey. This is sensitive when asking people from other cultures or countries, or they have religious memorabilia in the house;" "Asking about biological, adopted, stepchildren when maybe the kids are within earshot. Kids may not know they are adopted; stepchildren may be considered their own;" and "I despise asking the question about whether anyone in the home has difficulty dressing or bathing. It's awkward to ask this question to a 30-year-old woman."
Interviewers reported that they did not receive much training in asking sensitive questions. In general, their training focused on using a standardized interviewing approach, reading each question as scripted, keeping their voice modulated, not showing emotion, not leading respondents, and not giving one question any more weight than another. While interviewers reported that they are trained to make respondents feel comfortable, generally they are not given much guidance on how to do so when administering sensitive questions.