Reducing Burden in a Web Survey through Dependent Interviewing

Longitudinal surveys provide valuable data for tracking change in a cohort of individuals over time. Respondents are often asked to provide similar, if not the same, data at multiple time points. One could argue that this unnecessarily increases respondent burden, especially for information that does not change frequently. One way to reduce burden while still capturing up-to-date information may be to implement dependent interviewing (DI), where the respondent is provided information from the last data collection to aid in answering the current survey. If the information is still correct, then no change is needed, but if incorrect, the respondent has the option to change the response. To test this, we implemented two different versions of DI in a self-administered web survey and compared these against a traditional version of the web survey. We examined respondent burden by analyzing timing data and respondent enjoyment by analyzing debriefing questions. To assess the success of the implementation, we looked at timing data and undesirable behavior (missing data and backtracking). Finally, to evaluate measurement error, we looked at the number of meaningful changes. We found that DI is faster, more enjoyable, easily executed by the respondent (more so in one of our experimental formats), and significant measurement error was not introduced. In addition, DI provided consistency in the data, minimizing the noise introduced by nonmeaningful changes. The findings have significant implications for implementing DI in self-administered modes without an interviewer present.

Automated summary

Longitudinal surveys provide valuable data for tracking change in a cohort of individuals over time. Implementing dependent interviewing (DI) can reduce respondent burden and improve respondent enjoyment without introducing significant measurement error. DI provides consistency in data and minimizes nonmeaningful changes.