The effects of biological data collection in longitudinal surveys on subsequent wave nonresponse
Jul 26, 15:20
There is an increasing trend for social researchers to add components to existing longitudinal surveys that were unanticipated at the outset (e.g. biological data collection, data linkage requests) in order to complement the existing data. However, the effects of the unexpected increase in respondent burden can potentially increase subsequent wave nonresponse and attrition not only in the wave containing the new component, but also over the life of the survey.
One such additional component with a particularly large respondent burden, but also of potentially high utility, is the collection of biological data. This can include anthropomorphic measures, saliva and blood samples. Such additional data collection can increase the length of the interview and/or the number of visits as well as the number of consent procedures required and may involve some physical discomfort and/or anxiety for the respondent. For participants and non-participants alike, these factors may be off-putting to a segment of respondents and adversely impact their participation in future waves of the study.
The purpose of this chapter is to quantify the relationship between biological data collection and nonresponse/attrition in a longitudinal survey context. This chapter will first set out the current literature on respondent burden and wave nonresponse/attrition with a special focus on how this relates to biological data collection in longitudinal surveys. Following this, the effects of adding biological data collection to a large-scale longitudinal survey will be tested by analysing differences in subsequent wave nonresponse and attrition between respondents who were and were not invited to take part.
The data for the analysis will be from wave 2 of Understanding Society (UKHLS) and will use a quasi-experimental design. In this survey, biological data were collected over a period of two years using a trained nurse during a separate visit. In year 1, all eligible respondents were included but in year 2, a subsample was selected. The response propensities for UKHLS wave 3 for those respondents eligible but not included in the nurse visit will be compared to those who were included. The time to attrition will also be compared over the subsequent waves of the survey.
The outcomes of this analysis will help to inform survey researchers of potential issues with subsequent wave nonresponse and attrition to be considered when adding biological data collection to ongoing longitudinal surveys and also when using the resulting data in statistical analyses.