Reducing Data Collection Costs for A Household Survey of Children: Leveraging Appended Variables on Address Sampling Frame

Household surveys of children often require large address-based samples because eligibility can only be determined through household screening, substantially increasing data collection costs. Improving the efficiency of identifying households with children is therefore a central design challenge. Prior research showed that the presence-of-children flag appended to address-based sampling (ABS) frames lacks sufficient accuracy to support efficient oversampling of households with children in a general population survey.

In this paper, we focus on screening children aged 5–17 as the predictive performance of ABS-appended variables differs between young children and older age groups. Using the household screening result from a national mail push-to-Web household survey, we demonstrate that a stratified design based on a combination of ABS-appended variables can more efficiently identify households with children and reduce household screening costs. We also illustrate the conceptual framework for evaluating the impact on total cost for both screening and topical data collection. Effective sample size, which accounts for differential sampling rates across strata, is shown to provide a more appropriate measure of statistical efficiency than nominal sample size. Finally, we emphasize the importance of monitoring the quality of appended variables, as changes in accuracy can affect the efficiency gains achievable under a stratified design.