Geographic Screening

Most surveys target a specific geopolitical area, so that estimates produced from their data can be representative of that area. For some surveys, the area consists of an entire nation, but other surveys aim to produce regional estimates (such as those for states, counties, or zip codes). Thus, such surveys require some sort of geographic screening, or determination that a sampled case falls within the target geography, to establish study eligibility. If the screening is inherent in the sampling design itself, no further information is required. Other studies require additional screening steps, either prior to sample release or during the field period. Decisions about the level of geographic screening for a study arise from the sampling frame to be used.

When the sampling frame for a desired geographic area can be tied clearly to that area, no screening is needed beyond the design of the sample itself. For example, the sampling frame for a mail-based survey is composed of addresses that are known to be within a specific geographic area. Thus, geographic screening is part of the sampling design itself. Similarly, the sampling frame for an area probability sample is, by definition, geopolitically based, and therefore, no additional geographic screening is needed.

Telephone surveys typically use sampling frames that are denned by areas such as the nation as a whole, states, counties, cities, Census tracts, or zip codes. Samples of telephone numbers are generated by linking telephone exchanges to the desired target geography. In random-digit dialing (RDD) surveys of relatively small areas, it is impossible to match exactly telephone numbers with the boundaries of the target area. Researchers must determine whether the level of agreement between sampled telephone exchanges and the geography of interest is sufficient for their purposes or whether further questioning of the respondents to establish their location is warranted. This questioning can be complex and difficult to operationalize, thus leading to errors of omission and commission in which some eligible people are incorrectly screened out and some ineligible people are incorrectly screened in.

Implementation of additional screening steps increases the likelihood that all sample units are within the target geography, though it is likely to increase study costs as well—both of these are factors that influence a researcher's decision on the need for further screening. Decisions on additional screening are also influenced by considerations of incidence and coverage. Incidence is the degree to which the geography is represented in the sampled telephone exchanges, whereas coverage is the proportion of sampled telephone exchanges that are represented within the target geography. RDD sampling frames are often used to maximize coverage—ideally, all telephone numbers in an area are included in the frame for the RDD sample, including unlisted numbers, which would be missed if the sample were selected from telephone listings. However, RDD samples usually do not enjoy a perfect match between telephone exchanges and the target geography, so some sampled cases may lie outside the study boundaries. In such situations, the screening interview must ask respondents whether they reside in the target area. For example, respondents may be asked whether they live in a given county, zip code, or an area bounded by selected roads or other geographic markers.

...