Peter Yacobucci, associate professor of political science, relies on political polls for his work. “But you can never rely on just one poll,” he said. “And a major, unexpected event could completely change the opinion of individuals being polled.”
A poll’s reliability depends on three main factors, according to Yacobucci: who conducts it, how the poll’s questions were designed, and how the sample of individuals questioned was developed.
Who Does It, How, and Who Answers?
“The academic research centers like Sienna or Quinnipiac conduct good polls,” he said. “So do the major news organizations like Fox and ABC/Washington Post. Polling organizations like Gallup and the Field Poll are reliable, too.” However, he cautioned that the media may not report the poll’s result in its entirety. “Editors and broadcast producers choose the items that are most likely to draw readers and viewers,” he said.
A poll’s design is critical. “There is no perfect questionnaire,” said Yacobucci, “but it’s important to craft questions carefully and to avoid possibly biased wording.” For example, a leading question like “Last weekend we observed 9/11. Do you have a favorable opinion of first responders?” may change a poll’s results by as much as 10 percent.
A well-designed poll is important because people like to give consistent answers. Yacobucci said, “If a person says they support Trump, for example, they are likely to say that they favor a wall on the Mexican border in a subsequent question. You might get a different answer regarding the wall if you had asked that first.”
It’s also important to put demographic questions—age, education, income—last. A sample of 400 people—but no fewer— is the general standard for a representative poll, assuming the sample is random. Mobile phones are not as great a problem as they once were. As they have become more widespread, pollsters have devised ways to ensure that mobile and landline users both have an equal chance of being included. Polls that survey registered voters and people who say they are likely to vote are usually the most reliable.
Understanding Poll Findings
Even if an organization takes a near-perfect poll, the results the average person sees may misrepresent the poll’s findings. “High-quality polls of a random sample of 400 respondents have a three to four percent error rate,” said Yacobucci, “so a poll that shows one candidate will get 45 percent of the vote is really giving you a range of as much as 41 to 49 percent of the vote.”
Avoid relying on any one poll’s results. “It’s just a snapshot,” said Yacobucci. “A good researcher will look at five to 10 polls done by good firms, average them, and maintain a moving average of the most recent polls.”
Finally, remember that pollsters are talking to people: Minds can change on a moment’s notice. And a small but important percentage of respondents simply lie. “One person I know responds to pollsters by pretending she’s a character from a mystery novel,” said Yacobucci, “and she answers as if she were that character. So if you want to make sense out of a poll, you have to consider the human element as well as the methodology and the pollster.”
Back to Top
Some content on this page is saved in PDF format. To view these files, download Adobe Acrobat Reader free. If you are having trouble reading a document, request an accessible copy of the PDF or Word Document.