Modelling survey participation in surveys involving multiple phases of data collection

Abstract

This thesis aims to link the theory of response effects (Sudman and Bradburn, 1984), the conceptual theory of survey participation (Groves and Couper, 1996) and the route map of social exclusion developed (Atkinson, 1998), extending the survey non-response framework for studying the associations between social exclusion and non-response.  In the empirical part, we examine the Finnish Health 2000 survey data with direct linkage to auxiliary information at individual level.  In addition, we have conducted an interviewer perception survey amongst the interviewers who participated in the fieldwork of Health 2000.  We model survey participation behaviour of individuals in the presence of high response burden, analysing survey attrition across multiple data collection phases. Using multilevel sequential logit modelling, we incorporate the interviewer level information into the survey participation analysis.

We have found that the survey participation behaviour of individuals is greatly affected by their socio-economic circumstances, social capital, and social connectedness.  People with affluent circumstances are more co-operative than people with any of the social exclusion risk factors.  However, rehabilitation programs supporting people to actively participate to the society and help their entry or return to the labour market are found to increase survey co-operation.  We demonstrate that a single model oversimplifies the survey participation in a survey with multiple data collection phases. We show that the interviewer effect in face-to-face interviewing survey may impact participation at further data collection components, which by survey design are independent from the presence of the interviewers.  Finally, we illustrate that the survey estimates can be improved, if the survey non-response propensity weighting is accounted for depending also on the characteristics and perceptions of the interviewers. This finding shows that the interviewer effect can contaminate the obtained survey information not only at individual level, but also at the level population distributions for the survey estimates.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata