Media Contact:
For help contacting the expert listed below, contact Chad Austin at NC State News Services, at 919/515-3470.

Oct. 5, 2004

NC State Expert Available to Discuss Disparities in Presidential Polls

FOR IMMEDIATE RELEASE

Recent election polling appears to be extremely volatile. Some recent polls showed President Bush with a large lead (Time reported a Bush lead of 11 percent in early September) over his rival, Sen. John Kerry, while others showed the race as a dead-heat (the same week as the Time poll, The Christian Science Monitor reported a 47-47 tie). A recent Gallup Poll seems to diverge significantly from most other polls, showing a larger and growing Bush lead. What should Americans make of the different polling numbers?

Dr. Michael Cobb, assistant professor of political science at North Carolina State University who specializes in survey research and polling methods, says it’s important for the media to explain to its audience how polling works and why different polls produce different numbers. Citizens are mistakenly left with the impression that polls are not reliable snapshots of political attitudes, or that the race for the White House is already over. In fact, neither is true. Reported polls differ on several key aspects, and these can sometimes produce dramatically different results.

For example, some organizations report results for registered voters while others look at likely voters. Ascertaining who is a “likely voter,” however, is more art than science. Likewise, sometimes pollsters push respondents harder to make a voting decision – these are so-called “leaners” – while other times these people are counted as undecided. People that have not fully made up their minds, though, often hold different attitudes than those who are more certain, which can skew the results. Most importantly, some polls believe the electorate will be comprised of a certain percentage of Democrats and Republicans, and they adjust – or “weight” – their polling data to reflect that prediction. Gallup, for example, has apparently made some odd predictions about who is going to vote, so its poll had an oversample of Republicans who – not surprisingly – prefer Bush. Indeed, the infamous declaration of “Dewey Defeats Truman” in 1948 was based in part on the faulty polling that included too many Republicans in the sample.

Cobb is willing to talk with those interested in learning about the reasons different polling numbers are being reported this political season. He can be reached at 919/513-3709 or mdcobb@social.chass.ncsu.edu.

- 30 -