Why Presidential Polls Are Wrong

Blame bad timing and bad media practices for the surprise in the New Hampshire primary on the Democratic side, two political watchers say.

Hillary Clinton's popularity resurged quickly after her defeat in the Jan. 3 Iowa caucuses, so it was difficult to measure in the four days running up to Tuesday, said Dick Bennett, president of American Research Group, based in Manchester, New Hampshire.

ARG's final results prior to the primary put Barack Obama ahead by 9 percentage points. Similar results were predicted by pollsters Reuters/C-Span/Zogby, Rasmussen, CNN/WMUR/UNH, Marist and CBS News.

Bennett says Clinton, who won the primary with 39 percent of the vote compared to 37 percent for Obama (with 96 percent of precincts counted), got a surge of support as a result of her witty response to a question about her "likability" compared to Obama from co-moderator Scott Spradling of WMUR-TV during a televised ABC News/Facebook Democratic candidates debate on Saturday. Her "teary" response to a campaign stress question during a Monday stop in Portsmouth, NH, also didn't hurt.

Timing problem


Bennett had 30 staffers making phone calls the day before the primary. ARG was one of the few polling groups to see Clinton trending up among voters in the days between Iowa and New Hampshire, but "we didn't know the extent of it."

"The 'emotional moment' thing was played extensively, and it was played negatively. And they played it a lot. And everybody saw it and it rallied women, and she went back to where she was in the middle of December. She wasn't breaking any new ground. They left her because of Iowa. It was a timing thing," Bennett said.

The "emotional moment" story had an impact and was reported the night before the primary, he said. "We were in the field for [only] three hours after the story broke. That's not a polling problem, that's a timing problem," he said.

Bigger polling problems

The larger problem with many of today's political pollsters is that surveys are conducted in affiliation with media organizations, said Shawn Parry-Giles, a political communications professor at the University of Maryland who camped out in New Hampshire prior to the primary to make observations.

"Media aren't going to be self-reflexive about their poll," Parry-Giles said. "The journalists themselves just bought into the fact that [Obama] was so far ahead and it was inevitable. I was stunned by the coverage."

The media should stop treating polls as if they are factual information, she said.

"This is about what the voters say and do, and media has to be very careful about how they frame the polls," Parry-Giles told LiveScience.

One poll by CNN/WMUR/UNH on the anticipated results in New Hampshire had a relatively small sample size (which cripples a survey's accuracy) and a fairly large margin of error, but it was reported as accurate and went unquestioned, she said.

Other factors

An additional factor: New Hampshire voters pride themselves on being contrary.

"As you go from event to event, voters talked about 'how we're going to set our own trend. The country is going to follow New Hampshire, not Iowa,'" Parry-Giles said.

In general, voters trust polls too much, Bennett said.

"We've fallen into the trap that a poll, which relies on linear math, can explain a very complex system," he said. "And it can't, but we're lucky that we can get close enough."