Conferences

Nonresponse is a major problem facing research in the social sciences. Response rates in many surveys have been falling in many countries. To improve response rates and to decrease non-response bias, it is necessary to better understand the nonresponse process and why and how sample members reply to a survey. In face-to-face surveys the interviewer plays a vital role in contacting sample members and achieving their cooperation.
This paper investigates the influence of interviewers on unit-nonresponse across a number of UK government surveys. The analysis focuses on the effects of socio-demographic characteristics of the interviewer, interviewer experience, interviewing strategies and interviewer behaviours and attitudes. Of particular interest is the interaction between household and interviewer characteristics. Since several surveys are combined in one study, survey specific and survey independent interviewer effects will also be investigated.
The study makes use of a unique dataset, which links the survey outcome of several surveys to a rich set of auxiliary variables for both responding and nonresponding households, including socio-demographic information from the 2001 UK census, detailed information about the interviewer and information about the geographical area. The data available have a hierarchical structure with households nested within interviewers and interviewers being cross-classified with geographical areas. The paper explores the use of multilevel cross-classified models to analyse the effects of interviewers taking account of household level and area level characteristics. Implications for survey practice for the improvement of interviewer training and allocation will be discussed.