Abstract:
|
Life in Australia™ is an online probability panel built to represent contemporary Australian resident adult population. The initial recruitment of 3,322 panelists was completed in 2016 via dual frame random digit dialing. Since then, the panel retention and completion rates have declined resulting in an increased over-representation of university graduates, 55 to 74 year-olds and females and increased under-representation of people with less than a bachelor’s degree, 18 to 44 year-olds and males. To maintain the size of the panel while ensuring ongoing representativeness and viability, the panel was refreshed twice in a way that over-samples under-represented cohorts. This presents a challenge in the context of the traditional design-based approach to weighting as the chances of selection of the individual panelists become increasingly difficult (if not infeasible) to calculate. Furthermore, adjustments for attrition combined with the unbalanced sample design for the top-up negatively impact on sample efficiency and panel utility. In this paper, we describe options and trade-offs for deriving a parsimonious probability-based panel weight for panel with multiple recruitment waves.
|