Conversational Interviewing and Data Quality

摘要

Standardized interviewing is widely practiced because it promises to reduce interviewer related error and because it is cheap. Yet the technique cannot guarantee uniform understanding of questions and, thus, may reduce data comparability. Conversational interviewing may standardize the meaning of questions by allowing interviewers to clarify survey concepts, but it cannot guarantee uniform interviewer behavior. We discuss four experiments (three in the laboratory where respondents answer on the basis of fictional scenarios allowing us to directly assess accuracy) and one in the field (in which respondents are interviewed twice so that response change and explanation allow us to indirectly assess accuracy). We have found that conversational interviewing improves response accuracy when respondents' circumstances are atypical (e.g. Should a lamp purchase be counted as a furniture purchase) but requires additional time to clarify concepts. The more ways in which clarification can be provided the more accurate are responses and the longer interviews last. Respondents are not always willing to seek clarification when they need it, though we were able to overcome this in one experiment by reducing the social and cognitive demands of asking for clarification. Instead of interacting with an interviewer respondents interacted with a computer and clicked a mouse on highlighted text to obtain clarification; when respondents were told that clarification was essential for accurate responding they frequently obtained it. We conclude that researchers have the option of keeping interviews short and risking some misunderstanding or investing more time to be sure questions are understood as intended.