Conferences

Dependent Interviewing is the process of reminding respondents of their earlier answer in panel surveys. In this paper we examine two Dependent Interviewing designs that are frequently used and show how the relative stability of the variable and the reliability of its measurement both contribute to the choice of the best Dependent Interviewing technique possible. Our data consist of income reports of respondents from a four-wave panel study conducted in the Netherlands. We manipulate the income questions using a 3×2 factorial design (three Dependent Interviewing conditions with two types of questions) The effects of our experiment on data quality are modelled with a quasi-simplex structure to enable the division of variances into measurement error and real change. Our main conclusion is that the use of Proactive Dependent Interviewing is problematic when the reliability of the variable of interest is low. We finally discuss our results and suggest guidelines for the design of Dependent Interviewing procedures.