Key points are not available for this paper at this time.
Assuming a sound design, a polling sample is always later subjected to losses of respondents because of not-homes, refusals, and so on. From a survey of the 1950 election in Madison, Wisconsin, a report is given of therelation of such losses to number of calls made, some factors associated with the losses are cited, and statistical tests are made to determine whether the observed mortality resulted in sampling bias. Comparison is made throughout with the Elmira, New York, survey, as a norm. Francis E. Lowe during the survey was a graduate student of sociology at the University of Wisconsin and is now a member of the staff of Market Research Corporation of America in Chicago. The late Thomas C. McCormick was Professor of Sociology at the University of Wisconsin.
Lowe et al. (Sat,) studied this question.