Polls: the measuring stick of modern politics

In 1936 Literary Digest magazine, citing its own poll, boldly predicted that Republican Alf Landon would defeat President Franklin D. Roosevelt. FDR, of course, was re-elected by a landslide.

Oops.

Literary Digest is long dead. Polls, still in their infancy in 1936, are thriving. It's hard to imagine modern American politics without them.

The concept behind polling is simple: ask a question of a representative sample of the population, and you should get a pretty accurate sense of what the country as a whole is thinking.

Candidates use polls to monitor how their campaigns are faring, and to find out which buttons they might push to win your vote. The news media conduct polls to find out what's on voters' minds, and to track the "horse race."

Polls are controversial. Some people say they discourage good potential candidates with low name familiarity from running for office. Others say polls help obscure differences between candidates: they all read the polls and tailor their pitches accordingly.

Others say polls may measure public opinion, but often don't measure the intensity of that opinion.

And of course nobody knows how much those responding have thought about the candidate or issue.

Reading polls intelligently

Public-opinion researchers know much more about how to conduct a poll now than they did when the Literary Digest staged its spectacular bellyflop. Still, polls are imperfect tools, subject to manipulation and misinterpretation. Here are some questions you should ask whenever you're looking at poll results. Red flags should go up if the people behind the polls don't provide the answers.

1. Who sponsored the poll?
   Special-interest groups love to commission and release polls that support their points of view. If the National Rifle Association sponsors a poll that shows most Americans want to keep assault weapons under their beds, or if the Sierra Club sponsors a poll that shows people want the spotted owl to replace the bald eagle as the nation's symbol, you might well question the survey's objectivity.

2. How big was the sample?
   Obviously, it's impossible to poll every voter. But it doesn't take a rocket scientist to figure out that the more people you poll, the more likely you are to obtain results that reflect the attitudes of the population as a whole.
   How far from reality the poll results can be expected to be is called the margin of error. The larger the sample polled, the smaller the margin.
   For example, 570 Washingtonians responded to a statewide Seattle Times/Front Porch Forum in April 1996, and 18 percent said they were very concerned about keeping their jobs. Statisticians calculate that the margin of error for a sample that size is 4 percentage points. That means that if every adult in the state were questioned, between 14 and 22 percent would say they are very concerned about job security.
   A smaller sample would have a larger margin of error.

3. Who was sampled?
   Good pollsters strive for randomness. Theoretically, every person in the target population should have an equal chance of being surveyed.
   That's where the Literary Digest poll fell short. The magazine surveyed only owners of automobiles and telephones; in 1936, many poorer voters owned neither. So wealthier voters were over-represented in the sample, and the poll results were skewed.
   Today many pollsters dial phone numbers at random, so people with unlisted numbers will have as much chance of being sampled as people listed in phone books.

4. What methodology was used?
   Polls can be conducted in person, by telephone or by mail. Mail surveys, often employed by politicians to gauge voter opinion on issues, are the least accurate; the people who take the time to return the survey forms usually aren't a representative sample of all the people who receive them.

5. When was the poll conducted?
   Voters change their minds; an old poll may not reflect new developments or current attitudes. Pollsters expect a "bump" in a candidate's standing, for instance, at the end of a national convention with all its positive coverage.
   Pollsters incorrectly predicted Thomas Dewey would defeat President Harry Truman in 1948 partly because they didn't poll close enough to election day.

6. What was the exact wording of the question?
   Leading questions can produce biased results: An example: "For president, are you leaning toward voting for war hero Bob Dole, or that dope-smoking draft-dodger Bill Clinton?"
   Bias rarely is that obvious. Still, a look at the precise wording of a question allows you to make up your own mind. Some poll questions ask respondents to choose from a limited number of responses. Other questions are open-ended.

Finally, a brief word about three kinds of polls you may hear mentioned this election year.

Exit polls are often conducted by news organizations, who question voters as they leave polling places on election day. The results are then used to "project" the winners of races long before the votes are counted. Candidates have complained when TV networks broadcast results of these polls before all voters have cast their ballots, and these days most exit polls aren't released until after the end of the voting day in the state where the polling was done.

Tracking polls are small-sample surveys used by campaigns and others to track the standing of candidates on a regular basis, sometimes daily.

Push polls aren't really polls at all, but campaign propaganda masquerading as research. They aim not to measure public opinion, but to manipulate it by planting negative impressions of an opposing candidate in the form of a question. An extreme example: "Would you be more or less likely to vote for Candidate X if you learned he frequents topless bars?"

Makes no difference if that allegation is true; the purpose of a push poll is to plant that seed.

Here are several sites to visit for more information:

• "Background briefing": This is a paper by a former graduate student at the Woodrow Wilson School at Princeton on what makes polls accurate, biased, whatever.
• Web site for the Gallup Poll, one of the oldest and most respected polling organizations in the country. Site includes reports on many recent Gallup surveys.
• Web site for the Roper Center at the University of Connecticut, research center on public-opinion research. Includes online editon of their bi-monthly journal.

Copyright © 2000 The Seattle Times Company