Stouffer, S. A.
Harvard University, Cambridge, Mass.
Pages: 5 Published: Jan 1952
The methods to be discussed in this paper, dealing with psychological measurement with special reference to the problem of measuring consumer wants are old in principle, but their development in present form is new, and in many respects revolutionary. Asking people what they want or what they believe is as old as elections. But techniques for finding out quickly and accurately the attitudes of a population by querying only a small representative sample of the population are relatively new, and their potentialities are still imperfectly understood, not only by laymen, but even by those who are supposed to be, in various respects, experts. Some people still think of research in this area in terms of public opinion polls of state and national elections. Actually, the forecasting of elections is only a minute fraction of the work in this field. Despite all the perils of estimating not only what the public's preferences are at a given moment, but also of calculating the probability of a change between the time of interview and time of election, and of estimating who won't even bother to vote, election forecasting has an impressive batting average. One spectacular failure to name the winner, as in 1948, may close some people's eyes to the important fact that errors in scores of elections have always been within a range of 3 to 4 per cent. An error of only 3 to 4 per cent on most matters as variable as public attitudes is within almost any conceivable practical limit of tolerance. I do not want to leave the impression that all attitude or opinion research is as accurate as this. Some of, it may be more accurate; some of it may be much less. Research is now being undertaken to improve the techniques so that they will have greater reliability and validity.
Paper ID: STP47682S