Pollsters flummoxed by New Hampshire primary

WASHINGTON (Reuters) - Hillary Clinton’s victory in New Hampshire’s Democratic presidential nominating contest confounded pollsters, who found themselves trying to explain how opinion polls got it so wrong.

School children and voters are reflected in a puddle while Democratic presidential candidate U.S. Senator Hillary Clinton visits a polling place at Broken Ground School in Concord, New Hampshire, January 8, 2008, the day of the New Hampshire Primary. REUTERS/Brian Snyder

Chastened experts said on Wednesday they would have to closely analyze their forecasts against the results of the New Hampshire primaries to learn why they were so right about resurgent John McCain’s win on the Republican side but so wrong about Clinton’s win among the Democrats.

Ahead of Tuesday’s vote in New Hampshire, an early battleground in the state-by-state process to choose candidates for November’s election, pollsters had widely predicted Illinois Sen. Barack Obama would beat Clinton in the Democrats’ contest, with many foreseeing a double-digit margin.

In the event, the New York senator and former first lady beat Obama, edging him out by under 3 percentage points.

In an era of instant analysis, everybody had a theory on why the polls got it wrong -- from the humanizing effect of Clinton’s teary eyes while campaigning on Monday to the suspicion that New Hampshire voters get a perverse satisfaction from doing the opposite of Iowa voters.

“It’s really a case study in the limits of momentum,” said University of New Hampshire political analyst Dante Scala, commenting on pollsters’ predictions that Obama would keep surging after a decisive win in the Iowa vote last week.

In New Hampshire, large numbers of voters decide late, a trend even more pronounced this year because the Democrats liked all their choices, said University of New Hampshire political scientist Andrew Smith.

And since many surveys either stopped 24 to 36 hours before New Hampshire started voting, or rolled three days of data into one final average, they couldn’t pick up or reflect the late pro-Clinton trends.


Women came out in large numbers for the former first lady, and she did particularly well among those 65 and older, winning 48 percent of their vote in the Democratic contest.

“Clinton did very well among women,” Smith said. “And the late deciders broke for Hillary.”

John Zogby, who does the Reuters/C-SPAN/Zogby poll, said the 18 percent of New Hampshire voters who reported making up their minds on Tuesday “is just an unprecedented number.”

Like most polls, the last Reuters/C-SPAN/Zogby survey ahead of the primary was quite near the mark for the Republican race, predicting McCain would get 36 percent to former Massachusetts Gov. Mitt Romney’s 27. The final result was 37 to 31. But on the Democratic side, the survey predicted Obama would have 42 percent of the vote to Clinton’s 29, when in fact she won narrowly.

The flood of information on the Internet can shape how voters’ loyalties can shift quickly, an effect that may have been amplified among New Hampshire’s undecided voters, said Susan Herbst, a polling expert at Georgia Institute of Technology.

“There’s more movement and more fluidity of people and their opinion,” she said. “That’s a good thing. It means that people are thinking.”

Before last week’s Iowa caucuses -- notoriously hard to predict because of their unusual rules and the role of second-choice candidates -- the pollsters collectively shrugged their shoulders and concluded “who knows?” Obama won solidly.

But in New Hampshire the pollsters thought they were on more solid ground when they detected an Obama “bounce.”

One issue making it hard for pollsters is that this is a history-making contest with no past parallels -- Obama would become the first black U.S. president, and Clinton the first female U.S. president.

Herbst said the “public-private gap,” the difference between what voters say and do, has long been tough for pollsters to gauge.

But ABC's polling director Gary Langer wrote in his blog at that he doubted the wrong results would trace back to race.

He said pollsters need to look not just at factors like race but at “their own failings in sampling and likely voter modeling.”

“It is simply unprecedented for so many polls to have been so wrong. We need to know why,” he said.

(Editing by David Alexander and Frances Kerry)

For more about the U.S. political campaign, visit Reuters "Tales from the Trail: 2008" online at