Simulating the Consequences of Adaptive Survey Design in Two Household Panel Studies

Abstract

Abstract Adaptive survey design has been proposed as a solution to the decreasing response rates and higher costs associated with surveys. Much of the adaptive survey design research to date focuses on cross-sectional surveys, yet it seems there is greater potential for this approach within longitudinal surveys as data on the respondent and their survey experience builds each wave. We consider the short- to medium-term impacts of modifications to fieldwork processes in the context of two household panels. We use waves 11–16 of the Household, Income and Labour Dynamics in Australia (HILDA) Survey and waves 1–6 of the UK Household Longitudinal Study (also known as Understanding Society) to simulate eight alternative follow-up strategies. Cases are targeted based on their likelihood to improve sample balance (representativeness), as measured by the R-indicator, or to provide a response, or a combination of these components. Both individual- and household-level targeting are considered. We assess the extent to which these adjustments to fieldwork efforts impact the response rates, sample representativity, and cost. We find that if the follow-up fieldwork effort were reduced by 25 percent the least detrimental strategy is where the best households in terms of improving the R-indicator or the response rates are issued to field for follow-up. This approach resulted in the same sample balance as obtained with full follow-up practice, saved between 17 and 25 percent of follow-up calls but dropped the full balanced panel response rate over four years by 8–12 percentage points.