Probability Sampling in the Field: A Case History

Analysis of the results of the 1948 Elmira voting study in terms of calls sheds light on some of the problems of administering a probability sample. In general, for both individual characteristics and opinion and action items, the distribution of results did not change appreciably after the third call, although five or more calls were made. Mortality amounted to 19 per cent of the original sample, thus giving rise to the question how great mortality can be without destroying reliability. This experience suggests that the results of even a carefully designed and supervised probability sample survey cannot be evaluated without some knowledge as to how it was administered in the field. The author is a research executive on the staff of Elmo Roper.