Address-Based Sampling

Address-based sampling (ABS) involves the selection of a random sample of addresses from a frame listing of residential addresses. The technique was developed in response to concerns about random-digit dialed (RDD) telephone surveys conducted in the United States because of declining landline frame coverage brought on by an increase in cell phone only households and diminishing geographic specificity as a result of telephone number portability. The development and maintenance of large, computerized address databases can provide researchers with a relatively inexpensive alternative to RDD for drawing household samples. In the United States, address files made available by the U.S. Postal Service (USPS) contain all delivery addresses serviced by the USPS, with the exception of general delivery. Each delivery point is a separate record that conforms to all USPS addressing standards, making the files easy to work with for sampling purposes.

Initial evaluations of the USPS address frame focused on using the information to reduce the costs associated with enumeration of primarily urban households in area probability surveys or in replacing traditional counting and listing methods altogether. These studies showed that for a survey of the general population, the USPS address frame offers coverage of approximately 97% of U.S. households. The frame's standardized format also facilitates geocoding of addresses and linkage to other external data sources, such as the U.S. Census Zip Code Tabulation Areas data. These data can be used to stratify the frame for sampling target populations.

Use of ABS in conjunction with the USPS address frame does have some drawbacks. Researchers cannot obtain the address frame directly from the USPS but must purchase the information through private list vendors. The quality and completeness of the address information obtained from these vendors can vary significantly based on (a) how frequently the company updates the listings, (b) the degree to which the listings are augmented with information from other available databases, and (c) if the company purges records based on requests from householders not to release their information. Moreover, vendors differ in their experience with and ability to draw probability samples from the USPS list. This can be problematic for researchers who do not wish to draw their own samples and tend to rely upon vendor expertise for this task.

Another drawback is that coverage in rural areas tends to be somewhat lower than in urban areas. Additionally, in some rural areas, the USPS files contain simplified (i.e. city, state, and zip code only) listings rather than full street addresses. The percentage of these types of addresses in the database is declining, however, as local governments adopt emergency 911 protocols, which require that all households be identified with a street address. Therefore, over time, simplified address designations are expected to be replaced by full street address information. Another potential issue is that the USPS address frame contains post office (P.O.) boxes and multi-drop addresses (i.e. multiple persons associated with the same address), which may be problematic for both in-person and telephone surveys in which a street address is required to locate the household or to identify a telephone number associated with the household. Such addresses may be less problematic for mail surveys, where the initial goal is to ensure that the mailed questionnaire is delivered to the sampled household.

...