The impact of the interviewer : non-response and response variance in social surveys

• Main Author: Campanelli, Pamela Comber
• Published: London School of Economics and Political Science (University of London) 1999
• Subjects: 301.072
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.645532

Interviewer-based data collection is the norm for social and market research surveys in the United Kingdom and is likely to remain so for the foreseeable future. But what impact do interviewers have on survey resuhs. Interviewers are often seen as valuable allies in the data collection process for their role in minimising many potential sources of survey error (e.g., through motivating the respondent and controlling the response process). Yet, at the same time, interviewers can also be one of the principal causes of nonresponse error and response variance in quantitative surveys. In terms of nonresponse, it is widely known that different interviewers have different response rates, but comparatively little is known about the extent to which this is due to differences among interviewers and their characteristics as opposed to differences among the respondents in those areas allocated to the interviewers. What is also unclear is the extent to which such interviewer differences persist over time in longitudinal surveys, the extent to which interviewers differ in their effectiveness at reducing the refusal as opposed to the non-contact component of nonresponse, and the extent to which the fieldwork strategy of matching the same interviewers to the same respondents, 'interviewer continuity', is useful for raising response rates in longitudinal surveys. In terms of response variance, it is rare to find studies in which both the complex sampling variance and the complex interviewer variance are both computed or in which the effects of interviewer continuity on response quality is examined. This thesis investigates these issues by using the interpenetrated sample design experiment designed by C. O'Muircheartaigh (my supervisor) and myself for implementation in Wave 2 of the British Household Panel Study. The analysis is facilitated by the use of cross-classified multilevel modelling in addition to other techniques. The thesis also looks at the issue of 'interviewer continuity' qualitatively, through the impressions of the interviewers themselves.