Mixing modes in longitudinal household panels: recent developments and new findings
Jul 26, 11:00
This paper addresses the complexities of mixed mode designs in household panel surveys with a focus on how they affect initial and longitudinal response rates, sample composition and costs. In comparison with cross-sectional surveys, mixed mode designs in longitudinal surveys add the complexity that a respondent may complete a survey in a different mode at different waves which may compromise the comparability of repeated measures over time. For household panels a further complexity is that members of the same household may complete the questionnaire in different modes.
Our contribution provides first an overview of the combinations of modes that are already used for the purpose of mitigating nonresponse and attrition. For example, the German Socio-Economic Panel (SOEP) allows respondents to complete a paper-and-pencil questionnaire if a face-to-face interview (CAPI) cannot be scheduled. In the Household, Income and Labour Dynamics Australia (HILDA) telephone interviews are conducted as a last resort and to cover remote locations. The Swiss Household Panel offers CAPI if no telephone number is available, and CAPI and web interviewing (CAWI) to reluctant respondents. The extent to which such mixed mode designs improve participation and affect the sample composition has not been widely studied, however.
Recent experiments in the Innovation Panel of the UK Household Longitudinal Study (UKHLS), where CAWI (in combination with other modes) was introduced as an alternative mode in Wave 5 and continued in Wave 6 and 7, have demonstrated the potential benefits to be gained from incorporating web-based data collection. In preparation for its third refreshment sample, the SHP is similarly investigating the costs and benefits of a mixed mode design involving CAWI. A distinguishing feature of the SHP is that interviews are done by telephone, leaving it particularly at risk of bias from noncoverage and nonresponse. A further implication is that a household visit is not necessary, so offering CAWI may be especially cost-efficient in combination with CATI. In this paper we present the design and findings of the first wave of a two-wave pilot study testing the feasibility of using CAWI both as the primary mode, as well as sequentially, in combination with CATI/CAPI, aimed at reducing selection errors at recruitment and over time, as well as fieldwork costs. The use of a register-based sample offers the additional advantage of auxiliary data for assessing the extent of selection error, in relation to the costs associated with alternative survey designs.