(3 of 6)
No Rugged Individualism. He has determined the complexion of the U.S. people with near-mathematical exactitude (see chart) from census statistics, Government reports, etc., has made his own calculation of the U.S. voting population by such adjustments as canceling out non-voting Southern Negroes & poor whites. He knows, for instance, that 28% of U.S. voters live in the Middle Atlantic states, that 34% of them live in cities of over 100,000 population, that only 23% of them are of average means (i.e., skilled workers, white-collar employees, small shopkeepers), that 43% of them are between the ages of 30 and 49. If necessary, Gallup statisticians can dig deeper: 16% of them are farmers; 42% of them have gone to high school; 38% now call themselves Democrats, 36% Republicans, 26% independents, third party or no party.
As long as his polltakers keep precisely to those percentages in choosing their subjects to be interviewed, the whole U.S. electorate can safely be reduced to a miniature sample of only 3,000.* And despite the cherished American illusion of rugged individualism, the factors which determine voters opinions are the factors neatly worked out on Dr. Gallup's chart.
Like all theories, Gallup's has its obvious limits. No poll is any better than its interviewers. Though the polltakers' instructions carefully specify the cross-section to be taken, some Gallup pollsters are reluctant to venture into poorer districts; others fill out their ballots by punching doorbells in the daytime, thus missing jobholders.
Then there is the exacting job of framing the questions and weighing the answers. In pollsters' jargon, that breaks down into such specific problems as how to get around the "prestige answer" (what the "respondent" thinks he ought to think) and how to evaluate the "intensity factor" (how strongly the "respondent" feels about it). To solve such difficulties Pollster Roper has developed the "cafeteria" question, which gives a choice of answers from soup to nuts. Pollster Gallup has developed a system called the "quintamensional plan of question design," which measures not only the yes & no, but also the "respondent's" knowledge and his reasons for & against any issue.
When taking election polls, Gallup has two extra headaches. Ever since the early days of the New Deal, it has been proved that the heavier the vote, the larger the Democratic percentage. Thus, to arrive at his final winning percentage, Gallup must estimate in advance how many voters will go to the polls. His other headache is doping out the electoral vote; in 1944 Ohio went Republican by only 2/10 of 1% of the ballots. His fondest dream is that Congress will some day abolish the Electoral College.
In 1936, he underestimated Franklin Roosevelt's popular vote by 7% (Roper was off only 1%, Crossley was off 7%). But in 196 elections since then his average error in estimating the popular vote has never been greater than 4%; since 1940, never greater than 3%. In 1940 he called the turn within 3% (Roper was within 1%, Crossley within 4%); in 1944 within 1% of the civilian vote (Roper within less than 1% and Crossley within 2% of the combined civilian and soldier vote).