2010 Campaign: Why Political Polls Sometimes Get It Wrong

WASHINGTON, Aug. 26, 2010— -- Everybody loves a good horse race. And with primary season in full swing, political enthusiasts have no trouble getting their fill.

Pollsters across the country track the ins and outs of the most hotly contested races, putting out daily predictions for who will come out ahead. And poll aggregating websites, like Pollster.com and RealClearPolitics.com, have become an election buff's dream come true.

But a number of primary election polls this season have been notably off the mark in their predictions, raising questions about their reliability.

The results of Tuesday's primary offer the latest example. In Florida's Republican gubernatorial race, multi-millionaire healthcare executive Rick Scott stunned pollsters who had widely predicted state attorney general Bill McCollum would come out on top.

Polls in Alaska gave incumbent Republican Sen. Lisa Murkowski as much as a 30 point lead over her rival, Tea Party favorite Joe Miller in the weeks leading up to the election. Now the two are locked neck and neck in a race awaiting the tally of absentee ballots.

And in Arizona, young Republican congressional candidate Ben Quayle edged out a win ahead of a large field, despite polls to the contrary and allegations of his involvement with a raunchy website.

Experts say this year's primaries are highlighting some of the pitfalls of political polling.

"As a general rule, primaries are much harder to predict than general elections," said Brad Coker, managing director of Mason-Dixon Polling and Research. "The hard part is figuring out who's going to show up."

Pollsters say developing a fundamental sense of who is going to vote is harder to do in primary elections when turnout is historically lower and more variable. There's also the early and absentee voting factor.

Coker said Tuesday's Florida Republican gubernatorial primary was the first time in more than 20 years of polling that he didn't have the winning candidate leading in his final poll. And he was not alone.

"It was a late August primary, snowbirds were mailing in their ballots, it was the first day of school, the weather was bad and primaries typically have low turnout," Coker said, explaining the outcome vis-a-vis what most polls predicted.

Pollsters Defend Practice, Cite Primary Difficulties

Also complicating political polling during primaries is the absence of a clear partisan divide, which helps create distinct choices for voters and pollsters who track them.

"You're a Democrat, you're a Republican – it drives vote. That helps stabilize estimates in general elections," said ABC News polling director Gary Langer. "It's an absent factor in primary polling which makes it more difficult to arrive at a good estimate in these races."

Perhaps the biggest surprise on Tuesday was the Alaska GOP senate primary between Murkowski and Miller. The race is still too close to call.

One poll from Hellenthal and Associates in Anchorage, completed just ten days before the vote, had Murkowski leading Miller 61 percent to 36 percent respectively.

Political consultant and pollster Marc Hellenthal said that despite the spread, an unusual late surge in financial support for Miller made the outcome unsurprising. "We expected the race to be close. We've been saying it since Friday," he said.

Still, it's also clear a relatively small pool of polling data from Alaska also clouded the picture.

"Alaska is a very tough place to poll," said ABC News political director Amy Walter. "It's difficult to get a handle on Alaska because you don't have many pollsters there."

So with all the nuances and pitfalls of primary season polling data, is the attention given to the numbers ahead of election day misplaced?

"We all could do a little bit better to lay off the horse race in these surveys," said Langer. "We tend to disregard a really valuable use of pre-election polls which is to try to understand what the voters in these elections care about.... This single-minded focus on who's ahead and who's going to win is less enlightening and often less accurate."

But pollsters themselves don't necessarily agree.

"It's kind of like saying, 'Do we eat too much high fat ice cream?' or 'Do we watch too much TV?'" said Peter Brown with the Quinnipiac University Polling Institute. "Yeah, we probably do, and we keep doing it. They certainly provide information about a political race and they're a staple in American politics."

ABC News' Huma Khan contributed to this report.