Interviewer and respondent behaviours when measuring change with dependent interviewing
Jul 26, 09:00
Longitudinal surveys can take advantage of data collected in previous waves to make later-wave interviews shorter and easier. A respondent can be asked simply: “Last time we spoke to you, you were employed full-time. Has that changed?” If there has been no change, there is no need to repeat the long employment battery. However, such proactive dependent interviewing questions can be asked in several ways. For example, one seemingly equivalent wording is: “Last time we spoke to you, you were employed full-time. Has that changed?” Yet research has shown that these two wordings lead to different estimates of change in employment status. In this paper we seek to understand why these differences arise.
We report on an experiment in the UK Understanding Society Innovation Panel where the wording of these questions was varied. For example, respondents were reminded of their previous health status and randomly asked one of the following four questions: “Is this still the case?”; “Has this changed?”; “Is this still the case, or has this changed?”; “Has this changed, or is this still the same?”. We coded the respondent-interviewer interactions from audio-recordings of 1,187 face-to-face interviews from the third and seventh waves of the Innovation Panel. The codes capture several aspects of both respondent and interviewer behavior.
With these data, we address three research questions:
- Which wording is associated with desirable interviewer behaviour? With desirable respondent behaviour?
- Why are reports of change more likely in some formats and less likely in others?
- What behaviours are associated with respondents changing their minds – that is, reporting a change at first, but then reporting no change in the follow up questions?
Our results show that the “Is this still the case?” wording works best for respondents and interviewers. The two balanced formats (“Is this still the case, or has this changed?”; “Has this changed, or is this still the same?”) do not work well, despite our expectation. The “Has this changed?” format seems to encourage respondents to think about changes that have occurred and thus to overreport change.