Understanding Public Opinion Polls October 16, 2000 CNN
In the modern era, public opinion polls are everywhere, measuring the beliefs, attitudes, or expectations of Americans not only regarding preferences for presidential and/or congressional candidates, but on countless political, social, and economic issues. The Gallup Poll is universally recognized, but polls are also devised and implemented by magazines, television networks, and newspapers. Among the more prominent polls are those sponsored by Time and Newsweek, NBC News/Wall Street Journal, ABC News/Washington Post, CBS News/New York Times, and USA Today/CNN. So what is a public opinion poll? It is a method of systematically interrogating a smaller, representatively-selected sample of the public; a sample that is fundamentally an accurate "mirror" of the opinions held by the entire population as a whole. In other words, if the Gallup polling organization asks a national sample of 1500 Americans for whom they will vote in the presidential election of 2000, George W. Bush or Al Gore, then the percentage preferences found for each candidate in the sample should reflect, within an acceptable sampling error, how the much larger voting population of 100 million or more Americans will vote. How is this possible? This will be discussed a bit later, but suffice it to say that Gallup has accurately predicted the presidential winner in most election years. But even though scientific polls are far more accurate today, they still may run into problems. Such difficulties were clearly in evidence during earlier polling eras.
The Literary Digest Mistake of 1936 and The Gallup Poll's Error of 1948
Prior to the era of modern scientific polling, magazines and newspapers would solicit opinions from their readers through face-to-face straw polls or mail surveys. However, despite a large number of responses, these techniques were unreliable. The straw vote stressed quantity of responses over the quality of the sample. In other words, an accurate cross-section of the voting population was unlikely to be achieved. In 1936, The Literary Digest, a popular magazine of its day, predicted that the Democrat, incumbent President Franklin D. Roosevelt, would lose to his Republican challenger, Alfred Landon. The magazine had mailed postcard ballots to more than 10 million individuals whose names had been derived from automobile registration lists and telephone books. After more than 2 million ballots were returned and counted, The Literary Digest stated that Landon was the choice of a clear electoral majority. But the opposite happened as Roosevelt won in a landslide, capturing 60% of the vote and winning every state except Vermont and Maine. Where had The Literary Digest gone wrong? Unwittingly, the magazine had received ballots from predominantly wealthier Americans who could afford cars and telephones during a depression year when millions of other Americans were barely able to make ends meet. So, the sizable number of poor, or even average, income Americans, unable to afford the luxuries of an auto or phone, were missing from The Literary Digest's sample. It was precisely these Americans who overwhelmingly supported Roosevelt in the 1936 election. So, The Literary Digest's sample was "biased," i.e., it did not mirror the entire population, since so many poor people never had a chance to send in ballots. In short, the poll was inherently flawed and led to a completely wrong prediction. The Literary Digest went out of business not too long after this polling debacle. In subsequent years, more-sophisticated scientific pollsters developed far better techniques, using personal interviews with small samples of selected voters.
But in 1948, George Gallup's polling organization became famous for also creating a biased sample. Roughly two weeks before election day in November, Gallup confidently proclaimed Republican challenger Thomas Dewey to be the winner over the Democratic incumbent, President Harry S. Truman. Thinking Dewey to be far ahead and unbeatable, Gallup stopped interviewing voters. However, many voters changed their minds, switching from Dewey to Truman during that two week period. Gallup interviewers were not there to record those voting changes. It must be remembered that a poll only represents a "snapshot of opinion" at a particular point in time. The result was a Truman upset victory over Dewey. So the clear lessons of 1936 and 1948 are that the sample of individuals polled must be an accurate microcosm of the much larger population and that interviewing must continue up to election day since public opinion can be very fluid and unpredictable. Indeed, today, pollsters use exit polls, surveys of actual voters taken as they leave their polling stations on election day, in order to determine likely winners, well before all voting precincts have closed in the nation.
Modern Polling Procedures
For a poll to be accurate, it must be based upon a representative, randomized sample, employ valid or reliable questions, and have polling personnel carefully communicate with those interviewed. First, a representative sample is chosen through the process of random selection, which is basically a lottery system whereby every individual in the population (by age, religion, race, living area, etc.) in the nation has an equal mathematical chance of being included in the sample, just as in a lottery every number has the same probability of being selected. In other words, a properly constructed sample will include all segments of the population. If the national sample is large enough (typically 1500 to 2000 people interviewed) and is truly randomly formed, then those laws of probability state that final opinion results will be satisfactory, usually within a small margin of error-typically no more than plus or minus 3 to 5 percent in national polls. So, if in a 3 percent "margin of error" example, a poll projects that Candidate Jones will receive 52% of the vote on election day, then his actual percentage could vary within a range of 49-55%. How are individuals in the sample chosen? One of the most simple and inexpensive methods is to use computerized random digit dialing that automatically chooses telephone numbers. In this way, both listed and unlisted phone numbers can be tapped, hence avoiding bias in the sample. Appropriate and "willing" household members are then subsequently interviewed. Second, the manner in which questions are phrased is quite important to a poll's final results. Questions that are emotionally loaded can predetermine responses from those being interviewed. For example, the question "Do you agree with a woman's freedom of choice to have an abortion?" will elicit different responses than will the question "Do you agree with a woman's right to murder her unborn child by means of an abortion?" Third, questioners conducting polls through face-to-face interviews, or over the phone, must avoid contaminating truthful responses through their voice inflections, pre-existing attitudes toward the questions asked, or unintentional coaching of interviewees. To avoid these problems, polling organizations exercise great care in the hiring and training of their field personnel.
Polls and the Critics
Public opinion polls are obviously important tools of democracy. Leaders, particularly presidents, rely upon them as a way of finding out what Americans' attitudes are toward the key issues of the day, as well as the status of their own, personal popularity with the voters. Presidential and congressional candidates constantly take polls to find out how they can maximize their appeal with the electorate during campaigns. The media employs opinion surveys to judge how a campaign is progressing for hopeful office-seekers or to publicize an important societal issue. Also, when polls are conducted properly, they can be very accurate. Gallup polls have predicted the presidential winner for ten of the last twelve elections in September. But polls are not without their critics. One criticism is the "band-wagon effect," where published poll results may influence voters to go for the candidate who appears far ahead of his opponent. People want to select the obvious winner. A related contention is that citizens may even decide not to vote, figuring the election is over (this seems even more relevant for state and local races). This has happened with exit polls, when voters in the far West hear the early poll returns on the TV networks which may favor a particular presidential candidate. They end up not voting, thereby reducing national turnout. Second, poll questions over-simplify issues, frequently asking those interviewed for only "yes" or "no" answers to very complex problems. Third, not all polls are of equal scientific validity (the telephone call-in type is notoriously unreliable). Too many journalists ignore this fact when they use poll data in their newspapers, thus misinforming the public. Fourth, there is the growing number and influence of undecided or independent voters in the last four or five presidential elections. Many of these voters do not reveal their preferences to pollsters, hence generating greater volatility in survey research. Thus, in 1984, several polls seriously underestimated the final margin between Reagan and Mondale (eighteen percentage points); in 1988 the polls were in error by as much as seventeen percentage points in the various presidential primaries.
The Future of Public Opinion Polls
Polls will remain an important and necessary component of American political life. Indeed, as Internet technology progresses, it is distinctly possible that Americans could register their collective opinions through home computers! For example, suppose a future American President wanted an instantaneous, national, public opinion profile. After delivering an important policy speech on national TV, he could then ask viewers to respond negatively or positively to this proposal by clicking the appropriate box on their computer screens and e-mailing their responses to a centralized information processing network that, in turn, would deliver the president tabulated poll percentages. Of course, the old questions of sample bias and scientific validity would remain. Despite technology, the traditional ways of formulating poll questions and asking them to samples of Americans will not be disappearing anytime soon. The accurate measurement of public opinion remains of extreme importance to the vitality and regeneration of American democracy.
1. Historically, why and how was The Literary Digest poll of 1936 and the Gallup Poll of 1948 flawed? What lessons about polling were learned from these two polling mistakes?
2. In a poll, why must there be a clear and accurate link between sample and population?
3. What problems can be created by public opinion polls?
4. What are exit polls? What is meant by a poll's "sampling error"?
5. Could instantaneous polls via the Internet help or hinder American politics? Why?