Pollsters struggle to explain Clinton win

— -- A day after Hillary Rodham Clinton's surprise victory in the New Hampshire primary, the country's leading political pollsters found themselves trying to explain how and why they got it so spectacularly wrong.

Although most pre-election voter surveys accurately predicted Republican John McCain's win, most had Barack Obama crushing Clinton in the Democratic race. The average spread was 8.3 percentage points, according to the RealClearPolitics website. A USA TODAY/Gallup Poll published Monday had Obama leading by 13 points.

Even early surveys of people after they voted had Obama winning by 5 points. But when the votes were counted Tuesday night, Clinton won with 39% to Obama's 36%.

Pollsters came forth Wednesday with explanations that were part mea culpa, part admonishment that certain elections are simply impossible to predict.

Some, including Gary Langer, polling director for ABC News, called for an independent panel of experts to investigate what happened.

Some pollsters said the record-shattering turnout may have produced a different electorate than the one envisioned in their models designed to predict who will vote. Others pointed to the surveys as voters left the polls showing that 17% made up their minds on primary day, which they said may have confounded the pre-election surveys.

"It's absolutely a cautionary tale to both the people who do polls and the people who read polls," said Richard Morin, a public-opinion expert at the Pew Research Center. "Pre-primary polling is fraught with dangers. That is particularly the case in New Hampshire, which I have called the graveyard of political pollsters."

Andy Smith, a pollster for the University of New Hampshire, noted that previous pre-election polls in New Hampshire have gotten the margins wrong, underestimating, for example, McCain's 18-point defeat of George W. Bush in the 2000 GOP primary. Smith's final poll, for CNN, had Obama up by 9 points.

Usually, the vast majority of pre-election polls get the result right, said Langer, who did not conduct a New Hampshire poll after the Iowa caucuses, when most surveys began showing large Obama leads.

"That's what makes this failure, I believe, unprecedented," he said.

Morin believes several factors drove the error, chief among them the massive turnout of 500,000 voters. Typically, primaries are low-turnout affairs, and polls use questions designed to discard people they think are not likely to vote.

"This was really an unprecedented turnout, and I have a feeling that it simply overwhelmed many of the assumptions on which likely voters were identified," Morin said.

But Frank Newport, editor in chief of the Gallup Poll, said the demographics of his sample, including the percentage of older women, were very close to those of the actual voters, according to the voter surveys.

He and other pollsters — including John Zogby, who had Obama up by 13 — hypothesize that large numbers of people changed their minds after most poll interviews were conducted over the weekend.

"I have polled many races, especially close ones, where 4% to 8% have said they finally decided on their vote the day of the election, and that can wreak havoc on those of us who are in the business of capturing pre-election movements and trends," Zogby wrote on his website.

Nearly one-fifth deciding on election day, he said, was "unprecedented."

Actually, 15% said they decided on election day in the 2000 Democratic New Hampshire primary, according to the USA TODAY/Gallup Poll. But no one disputes that there were dramatic moments in the last two days of this year's race.

"There are many theories that there was a lot of last-minute movement in this hothouse environment," Newport said, such as "the intriguing potential impact of the 'verge of tears' video"— when Clinton's eyes grew moist as she answered a question about her personal motivations for campaigning.

And Clinton herself pointed to her debate performance Saturday night as a factor in the turnaround, he said.

Two-thirds of the interviews for the USA TODAY/Gallup Poll were conducted before the debate, Newport said. Gallup intends to interview those surveyed in an effort to determine who they voted for and why, he said.

"I think, just based on anecdotal evidence, that Clinton was able to connect with women voters who had been with her all along except for the last couple of weeks," said Mark Blumenthal, the publisher of Pollster.com, a website devoted to opinion research.

While Obama narrowly prevailed among women in Iowa, Clinton beat him 46%-34% among women in New Hampshire, the surveys of voters showed.

Langer disputes the idea that large numbers of people changed their minds at the last minute. He notes that Clinton won by 2 percentage points among those who reported making their decisions before election day, according to the voter surveys.

"I hope this is a big wake-up call for journalists, for candidates, for pundits," said Susan Herbst, public policy professor at Georgia Tech.

How various New Hampshire polls taken after the Iowa caucuses compare with the final results:

Sources: The Associated Press and RealClearPolitics

Contributing: Martha T. Moore, Kathy Kiely and Jill Lawrence in Manchester, N.H.; Mark Memmott in McLean, Va.