POLLS AND PUBLIC OPINION

New Zealand Listener, Volume 19, Issue 492, 26 November 1948, Page 19

POLLS AND PUBLIC

OPINION

Written for "The Listener" by

R H. THURLOW

THOMPSON

M.A. Department of Philosophy, |

Canterbury University College

HE public opinion experts knew exactly who was going to win the recent American presidential election — but they were wrong, and all the polls, national, State and local, were caught out completely. The results have given a certain amount of amusement to those unaccustomed fo the idea of polling, and many people are relaxing happily in this apparent confirmation of their conviction that public opinion cannot be measured in this way. However, though Gallup and his fellow workers may have been wrong, they have never yet been so wrong as those who fondly imagine that polls Can now be safely ignored or regarded with tolerant amusement. Public opinion polls are here to stay. There is nothing new in the attempt to see which Way" the political wind is blowing. The first "straw vote" was reported in the "Harrisburg Pennsylvanian in 1824 in an attempt to predict who would succeed Monroe in the presidency "without Discrimination of Parties." With the realisation that what people were thinking had news value, polling gradually increased until by 1928 there were 85 straw polls (75 local) concerned’. with the prediction of the presidential election results. Apart from the polls run by a few large newspapers, the early polls tended to be theatrical, unreliable, badly organised and frequently dishonest. One poll stood out as a model. Since 1916 the Literary Digest had achieved a remarkable degree of accuracy in predicting election results by sending out questionnaires to people listed as owners of telephones or motor-cars. It was able to say, with some justification before the 1936 presidential election: ""The Digest poll is still the Bible of millions." For the 1936 election, however, the Literary Digest was to have some new competitors using rather different methods involving careful statistical calculations which they had developed in the field of market research. The most important of these competitors were Crossley, of the Hearst newspapers, Roper, of the Fortune magazine,

and Gallup with his newly founded American Institute of Public Opinion (A.LP.O.). Six weeks before the Literary Digest and national oracle was even fo begin its polling operations for the election, Gallup warned his subscribers that the oldfashioned methods of the Digest would point to the wrong man, and predicted the inaccurate division of votes that the _ poll would achieve. The editor of the Literary Digest was annoyed. The election was a worrying ‘one for Gallup, who had sold his service on a

money-back guarantee and had to watch the fate of his new venture hang in the balance. The result of the election was a complete vindication for Gallup and those who had adopted the new scientific methods of polling. Complex and Intricate The principle underlying | public opinion polling is simply this: "by sounding the opinions of: a relatively small number of persons, proportionate to each major population group in every section of the country, the opinions of thé whole population can be determined with a high degree of accuracy." In other words, a cross-section of the population is worked out, and on the basis of this cross-section public opinion can be sampled with the surety that the opinion expressed by the sample is representative of the opinion of the population as a whole. The working out of the cross-section and sample is an extremely complex matter. However, the framing of the questionnaire and the interviewing which are equally essential parts ‘of the polling technique, are fraught to an even greater extent with delicate and intricate problems which are at the moment only partially understood. Public opinion experts using this technique now conduct polls in many countries and manage to measure, to within some 3 per cent. of accuracy, the views of the electorate on current social and political questions. The high degree of accuracy achieved can be judged from the results of three of the main American polls in the past presidential elections. The table indicates the degree to which the polls miscalculated the division of votes, the figures showing the percentage error.

This record of accuracy has unfortunately led people to regard public opinion polls with a feeling almost of veneration, as being virtually infallible. The experts themselves, however, have been uncomfortably aware of the way in which their remarkably accurate predictions have covered a multitude of inaccuracies that only a great deal of research will eliminate. Roper’s prediction, for example, in the 1940 presidential election, was apparently more accurate than that of Gallup, but his (continued on next page)

POLLS AND PUBLIC OPINION

(continued from previous page) greater success lay in the fact that heunlike Gallup-made two serious miscalculations which cancelled themselves out, so bringing him luckily almost exactly on to the mark. Variable Factors In view of past success, why did the polls suddenly fail so miserably in the last presidential elections? Until there is more information available it is impossible to do more than speculate on the possible causes, but there are certain factors to which attention can be drawn. First, as Gallup and others have always pointed out-the very statistical procedures that underlie the polling technique and assure a certain degree of accuracy 95 times in 100 also imply the possibility that five times in 100 the polls may be wrong. This is especially important in close elections. Secondly, while elections are the only objective check on the accuracy of public opinion measurement they are unsatisfactory as indications of accuracy for several reasons. Prediction in any field implies that all other factors remain constant. A chemist mixing one chemical with another can predict the result of the interaction only providing no one else had added other chemicals unknown to him. In polling elections there are certain factors that cannot be held constant. A wet day, for example, may upset prediction by lessening the number of votes cast in rural areas in a gréater proportion than in urban areas. The election is thus not necessarily an altogether fair check on accuracy in opinion measurement.

There are various possible reasons for the failure. Perhaps some new factor entered American politics which had not been allowed for in stratifying the sample; perhaps an inadequate. knowledge of the processes of public opinion prevented appreciation of the point where the "bandwagon" effect whith was apparently carrying Mr. Dewey to victory, began to work against his interests by removing the incentive to vote from his supporters; or perhaps it was just that the polls by many small deficiencies are not yet sufficiently developed to cope with such an election. It is impossible to say. The polling organisation and the public opinion research centres are the only ones with the material and the machinery to find out the cause of the debacle. Some Value Even in Failure There will no doubt be some who will no longer place any faith in the results of the polling organisations, but stch an attitude is as ill-advised as one of uncritical acceptance. The polls still remain the most important way by which we can really find out what public opinion is on an issue, the most important way by which the areas of public ignorance can be defined, and thus remedied, the most important way by which we can gain information on the processes of public opinion, and an important way of clarifying issues and

stimulating thought about those issues. Though the technique of public opinion polling is already one of the most important techniques in the social sciences, it is young and has much to learn. The recent failure may result, on the one hand, in the loss temporarily of some public support, but is unlikely to affect the polls adversely to any serious degree. On the other hand, the election results may give the general public a more healthily critical attitude towards the ‘polls, and provide the experts with material, which when analysed may yield a great deal of valuable information regarding the processes of public opinion and the technique of its measurement,

1936 1940 1944 *% ROPER (Fortune) ee 1.0 0.5 0.2 GALLUP CROSSLEY ( = a in% Honest ) 6 (approx. ) (approx.) 3.0 1.8 1.8 " 1.3