RDD telephone surveys: Toward a single-frame cell-phone design

Abstract

Given the tremendous societal and technological changes in recent years, survey research has moved from landline telephone survey design to dual-frame telephone survey designs that include both landline and cell-phone numbers. Consequently, researchers have developed methods to allocate samples and combine the data from the two frames. However, current statistical procedures will fail to detect the need for a single-frame cell-phone RDD survey design due to exclusion of key variance components, and optimal random-digit-dial (RDD) telephone survey design with lowest mean square error (MSE) may soon exclude landline numbers. This study provides evidence that a cell-phone design (currently deemed a radical departure from standard practice) is feasible and sometimes an optimal design. We discuss hindrances to its consideration as an alternative, propose an approach to compare different sampling designs, and present an empirical evaluation between current dual-frame designs and a cell-phone design. We find that the variance efficiencies from the use of a cell-phone design are gradually outweighing the coverage bias trade-off. We also find that the exclusion of users of both types of telephone service selected through the landline frame can lead to nontrivial bias even with near-complete population coverage by the cell-phone frame. This suggests the need for future research, such as weighting of single-frame cell-phone RDD surveys.