Reducing Seminar

Identifying and reducing survey nonresponse bias relies on supplementary data which are available for all sampled units. In the absence of useful frame data, paradata have been identified as a potential source of such information. A particularly rich subset of paradata are call records; the contact histories describing the recruitment process. These are appealing because they are cheap and abundant, often being collected as part of standard fieldwork protocol. However their longitudinal structure complicates analyses, and there is little consensus on how these data might be used in practise. This short talk will outline some of the issues involved in using call record data to adjust for survey nonresponse bias. Applications to a nationally-representative household survey will be presented.