Election-Year Polls Saturate the Public
POLITICAL opinion polls, like water, are usually welcomed by the public when they come as occasional nourishing rains. But cast your eye skyward; in this presidential election year political polls are falling steadily everywhere and have reached the flood stage.
What so many political polls appear to do in the age of instant communication is abridge the legitimacy of considered, independent thought; polls can crowd voters into being less individual and circumspect, and in some cases they probably do the thinking for people harried by the pace of a galloping world and looking for help. If enough polls indicate a trend or direction, then perhaps they are true and right.
But pause for a moment to consider the 1936 presidential election and the infamous poll of Literary Digest magazine, the poll that changed professional polling techniques forever.
Polling was in its infancy in 1936. Using a sample drawn mainly from automobile registrations and owners of telephones around the country, Literary Digest, one of the most popular magazines of the day, mailed out more than 10 million straw vote ballots (as part of a promotion campaign).
The final tally of responses just before election day gave Republican candidate Alf Landon 1,293,699 votes (55 percent), Democratic candidate Franklin D. Roosevelt 972,897 votes (41 percent), and the rest for a third candidate.
Of course on election day Roosevelt won by a landslide with 61 percent of the vote to Landon's 37 percent. The electoral vote was a staggering 523 to 8. Literary Digest had previously conducted polls on each presidential election since l920 and had predicted the winners. The magazine's poll was popular, credible, and carefully watched. There is even evidence that Landon's campaign was considerably buoyed by the results.
Three other pollsters of the time, Gallup, Roper, and Crossley, quietly predicted that Roosevelt would win, but the Literary Digest poll was the Big Mac among the lesser hamburgers of the day and got all the attention.
From a polling perspective, what went wrong?
Peverill Square, an assistant professor of political science at the University of Iowa, has studied the infamous poll along with a subsequent Gallup poll of 1937, which analyzed the 1936 poll. "The response bias of the 1936 poll was that the people who actually responded were different from the people who were sampled," he says.
In other words, it appears that the yuppie-like crowd with cars and telephones in those Depression days even had an upscale subgroup of Republicans, and these people formed the largest response group. Too many of them responded and thereby biased the poll.
Many of the others in the sample, those who presumably voted for Roosevelt, choose not to respond in great numbers (possibly because at the time they couldn't afford a five-cent stamp). "If you have a lot of people who refuse to respond," says Squire, "they may have a lot in common with each other and therefore you're missing their opinion. This is still a problem today [in polling], although less of a problem if one is using the right techniques. But the Literary Digest poll taught us the reasons for ra ndom samples from a population."
What this 1936 poll did was put an end to the assumption among professional pollsters that the quantity of respondents in polling was the road to accuracy. More proved not to be better. From then on the art of polling became more sophisticated. It turned to cautious random sampling as a better indicator of attitudes and preferences, and therefore enhanced the ability to forecast results with some degree of accuracy.
"Literary Digest had a great reputation going into the election," says Squire, "but the poll contributed to their demise." In 1938 the magazine ceased publication.
Polls live on in controversy. They are criticized less often for the kind of amateurism of the 1936 poll, but more for the way the polling industry seems to saturate campaign coverage with dozens of polls. In a presidential campaign the question can be raised: Who do polls really benefit, the candidates or the electorate?