Notice: the WebSM website has not been updated since the beginning of 2018.

Web Survey Bibliography

Title Correlates of Attrition in the German Internet Panel: Drop-Outs and Sleepers
Year 2014
Access date 28.08.2014
Which characteristics are associated with continued participation in a longitudinal survey? Answering this question is essential in any longitudinal survey and has been well-researched for face-to-face and telephone surveys. However, in the context of probability-based online surveys little is currently known about the correlates of attrition. This paper examines the correlates of attrition and nonresponse in the German Internet Panel (GIP) using longitudinal and multilevel data analysis methods for clustered and repeated
measures data. The GIP is based on a true probability sample of individuals and was the first of its kind to be conducted in Germany. In 2012 the recruitment of the GIP was first carried out offline through face-to-face interviews. To minimize coverage- and nonresponse errors households without access to the internet were equipped with the necessary hardware and/or a broadband internet connection. After online registration interviewing takes place bi-monthly on topics of political and economic behavior and attitudes. By May 2014 the GIP will have conducted ten waves of data collection. We first report results from the sequence analysis to better understand attrition patterns across all waves. Then we model participation at the first wave (recruitment stage) and nonresponse and attrition in subsequent waves. Our modelling approach aims to take account of the clustering of individuals within households and interviewers (from the initial recruitment stage) and of responses across waves (repeated measures). A range of covariates are available including both time-invariant and time-varying factors. We specifically focus on what we may learn from the paradata collected online at each wave, such as response times, break-offs and (mobile) device used. Initial analyses show various patterns of participation across waves, including respondents who drop out completely and sleepers who return to the panel. Our work will have implications for panel maintenance measures in online probability panels.
Year of publication2014
Bibliographic typeConferences, workshops, tutorials, presentations