Address-based Area Sampling: An Efficient Hybrid between Address-based Sampling and Area Sampling

Cost savings in area sampling (AS) due to use of address-based sampling (ABS) frame for listing addresses without making field visits are appealing but may lead to coverage bias due to problems in implementing it because some addresses, such as P.O. boxes and rural routes, cannot be linked to housing units (HU). In an analogous manner, cost savings in ABS due to telephone interviewing instead of field interviewing in AS are appealing but may lead to coverage bias due to problems in linking telephone numbers to all selected HUs. We propose an alternative design termed address-based area sampling (ABAS) which encompasses best of both AS and ABS. Instead of preparing a complete cluster or segment list by field staff, an initial segment list of only selected addresses from ABS is first prepared at the central office, and then a field staff or prompter is dispatched to verify validity of selected addresses, substitute new ones for invalid HUs, update addresses of selected list, and finally drop off short questionnaires to set up a telephone interview by the call center for administering the main or long questionnaire in the second phase. Substitution of invalid addresses is based on the methodology of rejection sampling recently developed by Singh and Wolter (2010) in the context of coverage of missing HUs under AS. It is observed from general practical considerations that the proposed ABAS design may provide an efficient compromise between a relatively inexpensive ABS design and expensive AS design while overcoming several limitations of both.