THE STRAW VOTE

Evening Post, Volume CXXVIII, Issue 38, 14 August 1939, Page 10

THE STRAW VOTE

AMERICAN OPINION

METHODS OF TESTING IT

SUCCESSFUL RESULT

In: the United - States, testing public opinion on the issues of the day and fo.vocasting the results of elections have bs-en carried to a remarkable degree of accuracy, says a writer in the "Mel-ibcr-arne Age." , For more than thirty years "straw i fcialilots," as they are called, have been conducted by American newspapers •aaid periodicals, chiefly during election jsreunpaigns, and some of these ballots came very close to the actual results ;revealed when the votes were counted. '.But straw ballots of this kind cannot loe; conducted on a scientific statistical 'rtjasis, because it is practically impossible for any newspaper or periodical ip ascertain through its readers, through mailing lists, and from names taiken from directories, a reliable poll, representative of all grades of the comruunity. These straw ballots have, however, Xed the way to the establishment of ibusiness organisations in the United ilStates which specialise in the collecL'tion of statistical information on public feeling on questions of the day by means of personal interviews with numerous representatives of all sections of the community. The two chief organisations of the kind are the American Institute of Public Opinion, and that controlled by the monthly magazine "Fortune." NO MACHINERY EXISTING. In a recent issue of "Harper's Magazine"' Mr. Jerome H. Spingarn gave some interesting particulars about the methods of these two organisations in collecting information on which to forecast public opinion. Many of the questions handled by these organisations are issues on which there is no political machinery in existence for obtaining a direct expression of public opinion. Such questions as "Should sit-down strikes be made illegal?" and "Do you approve of the proposal that all labour unions be incorporated in order that they may be held legally liable for the contracts they make?" come up for discussion during election campaigns, but elections are not decided on such direct issues alone. Such issues are bunched with others of greater or lesser importance in the platform of candidates standing for election. The two organisations mentioned which specialise in testing public opinion deal with these questions, and with others which are of public interest, but which lie outside the range of party politics. They do not concern themselves solely with political issues and election results. One of them places its services at the disposal of advertising agencies, corporations, and trade associations for ascertaining the preferences of housewives with regard to particular goods on the market. The American Institute of Public Opinion employs about 640 part-time workers, located in cities and rural areas throughout the country. They receive blank ballot papers from the central office, and call on people in their homes, in offices, and on farms, and stop people in the street, to obtain their opinions. They are paid at the rate of about 2s 6d an hour, plus travel in rural districts. "These field workers," states Mr. Spingarn, "put in an average of five or six hours a week on the job. HONEST AND UNBIASED. "Most of them are college graduates," all have had references from prominent local citizens attesting to their honesty and freedom from bias. A large number are young married women. Their work is closely watched by the central office, ani i.r an interviewer consistently sends in results which contrast with the results sent in by others, his territory may be checked by mail, or by another field worker, to determine whether he is cheating. "Interviewers work in their own time —some in the morning, some in the afternoon, and others at night. In industrial towns like Detroit it is practically impossible to find men at home during the day, and all interviewing must be done at night, or over the weekend. People are very willing to express their views, and as the surveys become better known, they feel a certain pride in being represented in the nation's cross-section. The names of persons interviewed are not taken, but the name of the street in which they live is noted. Unless the subject is unusually garrulous, an interview takes about ten minutes. "Generally when a ballot is mailed to interviewers on Monday, it reaches the most distant ones by Wednesday morning; field work is completed by Saturday night, and on Monday morning the bulk of the returns come in to the central office, with the last returns following on Tuesday and Wednesday. The tabulation is done by machine, and can be completed within a few days after the last ballot comes in. In emergencies a 'rush' poli on a question can be made in seventytwo hours. THE FACTOR OF SIZE. "This scientific random sampling simply boils down to taking the opinion of a given number of residents of Park Avenue (let us say) selected at random as a composite of what Park Avenue people are thinking; then doing the same thing in a factory district, with a much larger number of people, because there are many more factory workers; and having another interviewer visit farm houses in Wisconsin. If all groups of people are sounded out, and each is represented in proportion to its voting strength at elections, a representative cross-section of the electorate is obtained. The size of the sample used is important. If we question only one New England farmer we may get the village atheist, whose ideas are unique in the community. If we question twenty we counteract such errors. The statisticians who have made exhaustive study of the laws of chance say that 'a sample is adequate in size when an increase in the number of cases fails to produce significant differences in results.' "As a general rule about 5000 persons, properly distributed among all grades of the community, are questioned. A much larger number are questioned when public opinion on a matter of great public importance is being tested; but it has been found that a test of 50,000 voters will be only about 2 per cent, more accurate than one of 3000. Minors are not usually questioned, nor are negroes in the south, where they may not vote. Results from States are weighed in accordance with their proportionate voting population. Thus while lowa has roughly the same population as Georgia, her opinions are given five times as much weight as Georgia's,-because approximately 50 per cent, of lowans vote, while only 10 per cent, of Georgians do. j ABSENCE OF PARTIALITY. "More than seventy, newspapers in the United States subscribe to the surveys (as these polls are called) con- 1 ducted by the American Institute of '. Public Opinion. They pay a price < | lor the right to publish the results, <

together with interpretative comment by the institute. The papers are noi allowed to alter any of the copy in a way which might reflect on the impartiality of the poll.

"Even before the surveys were begun their organisers anticipated the difficulty they would ' encounter in steering a neutral course and in avoiding charges of bias, especially from people who are disappointed in their findings," writes Mr. Spingarn. "No one has ever proved any charge against the American Institute of Public Opinion or the 'Fortune' surveys. Both organisations are anxious to gain acceptance as accurate indicators of what the country is thinking."

Although election forecasts are not the chief concern of the surveys, election results provide the best means of testing the accuracy of the methods employed for obtaining information on which to interpret public opinion on questions on which there is no constitutional machinery for obtaining such results. Both organisations, therefore, devote considerable attention to election forecasts. In recent State elections their forecasts have been within 3 or 4 per cent, of the actual results. In the Presidential election of 1936, in which President Roosevelt received 60.7 per cent, of the total of 45,646,817 votes recorded, "Fortune" forecast a poll of 61.7 per cent, for Roosevelt, and the American Institute of Public Opinion forecast 53.8 per cent \