Conference Paper European Association for Survey Research Conference
Dependent Interviewing: A Framework and Review of Current Practices
20 Jul 2005
Many panel surveys have introduced dependent interviewing (DI), the feeding forward of substantive information from previous waves of data collection to formulate questions or determine routing (proactive DI) or prompt post-response edit checks (reactive DI). Proponents of dependent interviewing claim that it reduces respondent burden, increases the efficiency of data collection and improves data quality. The main purpose, however, varies across survey organisations and across surveys resulting in a variety of designs and applications. As a consequence it is not straightforward to evaluate the effects of different design features or compare the use of dependent interviewing in different surveys.
This article reviews current applications and develops a conceptual framework of dependent interviewing, in an attempt to disentangle design features, channels through which they take effect, and implications for burden, efficiency and data quality. The scope of design features is broad. With proactive DI respondents can be reminded of previous answers (1) to aid their memory and provide a boundary before asking the standard independent question, (2) to ask respondents to check and confirm previously recorded answers, or (3) explicitly to ask about changes. If circumstances have not changed, proactive DI may in addition provide the means to route around follow-up questions (perhaps with subsequent imputation of previous data). Reactive DI can be used to follow-up item non-response (don't knows or refusals), or as a corrective edit which can be displayed (1) always, for example to check consistency of verbatim answers with previous reports, (2) to clarify reports that are inconsistent with previous reports, or (3) selectively, if reports differ from previous reports beyond a defined threshold. The aim of the conceptual framework is to aid the understanding of the effects of dependent interviewing, about which empirical evidence has been very limited until recently.