“The exact wrong thing, which a lot of people do, is they pick the result they like best and rationalize as to why it’s the most reliable,” said Scott Rasmussen, president of the Rasmussen Reports polling agency. “That’s a natural human instinct, but you should always look at as much polling data as you can, and rather than picking apart the differences, look for common ground.”
Last week’s common ground was that four of the five national polls were close — either inside the poll’s stated “margin of error” or within 1 percentage point of it.
“If the ‘registered voter’ polls are showing a little bit of a lead toward Obama and the ‘likely voter’ polls are showing a few points more in Romney’s direction, this race is very close,” Rasmussen said. “That is the overriding message.”
How close? As of Friday afternoon, the website PollHeadlines.com, which tracks and averages polls from dozens of sources over time, showed the race is neck-and neck with Obama ahead nationally 45.64 to 45.62.
Campaigns’ reaction
Ryan Mahoney, regional press secretary for the Republican National Committee, made a similar point, saying the party tries to look at a combination of polls over time.
The campaigns sometimes pay as much attention to specific poll topics and demographic trends as the overall results.
“You dig deep into (the polls) and the one constant is voters care about the economy,” Mahoney said. “That’s a crucial indicator, and the fact that the president continues to struggle with those economic numbers is our indication of where things are.”
Obama supporters can point to this week’s Ohio-specific polls as good signs, as Quinnipiac Polling put Obama ahead 47-38 in the state, one day after Public Policy Polling showed Obama with a 47-44 edge here.
“While everyone expects this to be a close race due to the historical challenges the president faced when taking office, polls have made clear that the people of Ohio just aren’t buying Romney’s out of touch ‘doom and gloom’ rhetoric about the Ohio economy,” said Obama Ohio spokeswoman Jessica Kershaw.
Poll variables, bias
While the campaigns try to use the poll data to their advantage, the pollsters themselves debate how to make their polls as accurate as possible.
Peter Brown, assistant director of the Quinnipiac University Polling Institute, said there are many factors that determine reliability of polls — the number of people questioned, the use of registered voters or likely voters, the effectiveness of human callers vs. computers, and more.
“Everybody is accused of being biased at one time or another,” Brown said. “But we believe our methodology is solid. We believe our questions are written in as impartial a manner as possible. We use human beings to make the calls, which is considered the gold standard. We call cellphones, which is considered the gold standard.”
Rasmussen has been accused by some Democrats of skewing his polls in favor of Republicans. Rasmussen Reports and Gallup Poll are the two organizations that release daily tracking polls on the presidential race. In the past two weeks, Rasmussen’s results have been more favorable to Romney than Gallup’s have, on 13 of 14 days.
Rasmussen said it’s a sampling issue, as Gallup Poll surveys only registered voters, while Rasmussen uses the less certain “likely voters.”
“Right now, Republicans are far more engaged in the election, paying more attention to it, and that’s one of the very best indicators (of a likely voter),” he said. If the Democratic base is less energetic, it may affect the number of each political party that make it into the polling sample.
Accuracy, timing
Brown stressed that polls give an accurate snapshot of people’s feelings at one moment in time, adding that the same question asked a week later will likely draw different results. He pointed to 1980 polls showing that Jimmy Carter led eventual winner Ronald Reagan for much of the summer, and 1988 polls where Michael Dukakis was well ahead of George H.W. Bush, as proof of that fluidity.
“Polls are very good at telling you what people think today, and they can be OK at what they’re going to think two days from now,” Brown said. “They’re not very good about a week from Tuesday, and they’re really bad about 4 ½ months from now — that’s an eternity. And that’s the only poll that matters.”
There are times when polls are simply off-base. Polls’ margins of error are usually stated “at the 95 percent confidence level.” That means you can expect the unexpected about 5 percent of the time. The June 20 Bloomberg presidential poll, which showed a 13-point margin for Obama when all other polls were close, could be an example.
“If you see one that’s way out of sync, just assume it’s an outlier,” Rasmussen said. “Statistical theory says 1 of 20 polls will be outside the margin of error.”
Rasmussen added that occasionally, the entire polling industry just misses one. He said virtually every poll had Obama comfortably winning the 2008 New Hampshire primary, but he finished second to Hillary Clinton.
But pollsters usually do well just before Election Day, when more of the uncertainty has disappeared. A Fordham University survey on the 2008 presidential election showed, on average, 23 major polling agencies missed the Obama-McCain margin by just 1.4 percentage points on their final polls. The final Rasmussen and Pew Research polls were almost exact matches of the national results.
About the Author
