Compromise Solutions: Probability Sampling Adaptations

Cognizant of the advantages of probability sampling, yet confronted with limited resources, many researchers have used compromise solutions to the sampling problems posed. These are intended to achieve some of the benefits of probability sampling from the precise universe of interest, without incurring the typically high associated costs.

Area sampling Modifications:

One such compromise involves the selection interview areas, using probability sampling, but identifying respondents for interview in the selected areas by non-probability methods. For example, a several stage probability sample might be used to identify areas (e.g. blocks) as places of interview. Then, sample respondents could be identified for interview using a systematic pattern (e.g. every fifth house) from an arbitrarily designated point within each sample area (e.g. the southeast corner) Methods akin to this are widely used.

Probability Sampling with Quotas:

One application assigns quotas, within probability selected areas, based on estimated numbers of four groups within each area: mean under and over 30, and unemployed versus employed women. Such a system retains some of the features of probability sampling, in that objective selection of where to interview is maintained rather than leaving it to interviewer choice. The imposition of quota controls ensures that certain hard to find people are represented, with appropriate frequency, in the sample. The procedure is substantially faster and cheaper than standard probability sampling for individual, in-home field interviews. These virtues are bought at the expenses of the assurance provided by pure probability sampling.

Approximately the Relevant Universe with a Listed One:

Another compromise solution is to use probability sampling from a universe that approximates the precise universe of interest. In effect, the precise universe of interest is assumed to be little different from the approximated universe for which a satisfactory list of universe elements exists.

Probability sampling of Telephone Homes:

A very common example of the approximating approach is to assume that telephone accessible households are equivalent to all households for a particular study. Given that some 95 percent of households have a phone, this assumption will be satisfactory for many consumer research purposes. The principal advantages of this approach (versus, for example an area sample of personal interviews) are savings in cost and time, as well as relative simplicity The major disadvantages are those linked to telephone interviewing.

Such telephone probability samples are often used for “tracking studies” where in the analyst tracks trends in such consumer response variables as brand and advertising awareness, brand usage, attitudes, and so on. They are especially useful in evaluating the early progress of a new product introduction. They are also used to evaluate the effects of changes in marketing mix (e.g. an increase in media expenditure). Telephone probability samples are also sometimes used for screening purposes, to locate hard to identify product users. For example, one study identified a sample of male contact lens wearers for follow up personal interview.

This discussion has implicitly assumed that the researcher has a complete list of telephone households. For limited purposes (e.g. when sampling in a local area, and for information which must be obtained quickly), ordinary telephone directories may be adequate. At the other extreme, when high quality data are required on a comprehensive and continuing basis from a universe of all telephone households, then ordinary directories will not be adequate.

Because unlisted telephones may represent a significant part of the total universe, it may be necessary to use the special sampling technique called random digit dialing, which provides for proportional representation in the sample of unlisted numbers. Use of this more expensive, but more accurate, technique is particularly important if households with unlisted phones significantly differ from listed households in the characteristic of survey interest.
In practice, telephone probability samples are typically selected by some type of systematic sampling. In its simplest form, this means taking every kth listing from a local directory. In large scale, national studies the sample design will usually be stratified on such bases as geography and degree of urbanization, and will employ random-digit dialing. Most commercial research firms have access to telephone probability samples based on random digit dialing.