Hot Topic: Do the Polls Really Matter?

With less than two weeks until the presidential elections, one thing is certain: Americans are obsessed with the polls. Daily tracking polls and websites that aggregate results are feeding the fixation. The polls show a tight race but President Obama maintains an edge in crucial swing states. How accurate are these surveys at predicting the election outcome and how much attention do they really deserve?  Presidential historian David Greenberg, an associate professor of journalism and media studies and of history, and Cliff Zukin, a national expert on opinion polling, mass media and American politics, explain why the polls get so much attention and shed light on how accurate the polls have been in the past.

Rutgers Today: Is there more of a focus on the polls in this election, or are there more polls than in the past? Why are we hearing so much about them?

Greenberg: We’re hearing so much about the polls for two reasons. The first is that there are more of them. The second is that they’re accessible to everyone on the web. It used to be that news outlets reported only their own polls (conducted by professional firms). The New York Times reported on the Times poll, The Wall Street Journal told you about their polls, The Washington Post told you about theirs. Starting a few election cycles ago, websites like Real Clear Politics and Pollster.com began compiling all the polls for laypeople to read. Now it’s often possible to get PDFs that show not just the results but the exact questions, the responses, and even the methods used to take and weigh the polls. So political junkies now click every day on RCP or FiveThirtyEight (Nate Silver’s polling site, now hosted by The New York Times).

Zukin: I don’t think there is more of a focus than in 2004 or 2000. I think it is the same over-coverage of polls as we’ve seen for some time. The reason for this is journalism’s preoccupation with the horse race – who’s ahead and who’s behind, and how has the public reacted to the events of the campaign, such as Romney’s 47 percent comment and the debates. No question that it is an unhealthy preoccupation as campaign dynamics drives out other news.

Rutgers Today: On the same day different polls have shown the presidential race tied, Obama maintaining a small lead and or Romeny pulling ahead. How can polls produce different results and what are some of the factors that influence the outcome of the polls?

Greenberg: Polls are not perfect, and there’s always going to be a fair amount of variation among them. A poll that’s wildly out of line with others is probably wrong. But small leads by one candidate or the other are usually within a margin of error. Also, different polls employ different methods. Some may use “robocalls” – automated questioners, not real people. Those are considered less reliable. Others may not call cell-phone users. They have different ways of weighting their samples, too, in order to make sure the people polled reflect a true cross-section of likely voters.

Zukin: They are not really wildly different.  Polls can vary by a few percentage points because they are samples drawn to estimate the public; they are not meant to be accurate down to 1 or 2 percentage points.  Also, different pollsters have different ways to determine likely voters, use different sampling strategies, field procedures or weight the data slightly differently.

Rutgers Today: Do you think that polls accurately reflect how people plan to vote, or do they determine it? Do the polls become a self-fulfilling prophesy?

Greenberg: The polls are fairly accurate as to how people plan to vote when they’re asked. But, as the cliché goes, the only poll that counts is the one on Election Day. People change their minds, or they decide not to vote. There is usually some variation, even if it’s slight, between even the best last-minute polls and the final tally.

Zukin: There’s been no evidence in the history of presidential polling of a bandwagon effect, where people follow polls and adjust their votes accordingly.  In fact, the critical last 10 percent to decide in an election are the least interested and informed, who may not even be aware of poll findings.

Rutgers Today: Historically, how accurate are the polls at predicting the outcome of the election? Is there another example of when the polls were this tight this close to the election and what happened?

Greenberg: In 1936 the Literary Digest poll was supposed to be the best – but it predicted Alf Landon would beat FDR. He lost in a landslide. By 1948 George Gallup had dethroned Literary Digest – but he predicted that Tom Dewey would beat Harry Truman. Polls have never been as good at predicting as the pollsters would like us to believe.

Polls today are somewhat better at predicting but in close races, like 2000 (Bush vs. Gore) and 2004 (Bush vs. Kerry), some predicted one, some predicted the other – so obviously some polls had to be incorrect. The same is likely to happen this year. Because the race will be close, it’s unlikely that everyone will predict Obama or everyone will predict Romney. So some will be right and some wrong.

Zukin: Historically polls have been VERY accurate. The website Real ClearPolitics aggregated all the final polls in the 2008 election and found that the average was a margin of 7.6 percent for Obama. The actual vote margin was 7.3 percentage points.  In 2004 the average of final pre-election polls had Bush winning by 1.5 percentage points; he actually won by 2.4 points.