Stephan, Frederick F.
Professor of Social Statistics, Princeton University, Princeton, N. J.
McCarthy, Philip J.
Professor of Statistics, Cornell University, Ithaca, N.Y.
Pages: 6 Published: Jan 1952
In the great mass markets served by modern industry, the consumers of a product are so numerous and so widely dispersed geographically that the manufacturer cannot ask each potential user what qualities he wants the product to have or find out from each actual consumer how well he is satisfied by the product he gets. Consequently, unless a producer has unusually accurate information flowing back to him from dealers and salesmen through the channels of distribution, he must find out about the public response to his product in some other way. This often leads to a survey of the wants and reactions of a sample of people chosen in the hope that their responses will give him a fair indication of what all consumers desire. The designer of a new product has no group of regular consumers and must likewise turn to a sample of people to test the acceptability of the product in the entire market. Sampling people differs from sampling bulk materials or sampling a product on the production line because the people who are chosen must accept their selection and cooperate actively. Many other human factors are involved in the sampling of dispersed populations. However, some of the techniques of sampling that have been developed for use in engineering, agriculture, and testing can be applied to the problem of obtaining samples of people to represent accurately a large group of consumers or some other population group. During the past fifteen years or more, very impressive use has been made of these techniques in censuses and surveys conducted in America and elsewhere for a great variety of purposes. Many improvements have been made in the methods, and a substantial body of sampling theory is now available to guide the application of the techniques to human populations. With these modern sampling procedures it is possible to make quite accurate determinations from relatively small samples when the other phases of the process of measurement are also up to the same high standard of accuracy. However, these procedures are not always simple, inexpensive, or free of important technical and operating problems. Our analyses of a number of surveys and comparison of their performance have convinced us that there is no one sampling procedure that is uniformly most effective. The more successful instances of sampling are those in which the component techniques and operations were selected and combined in a design that is fitted to the characteristics of the population and adjusted to the specific purposes of the survey.
Paper ID: STP47685S