The pollsters all agreed.
Anchorage Mayor Mark Begich would whip Republican Sen. Ted Stevens in Tuesday's election. Maybe by a little. Maybe, as some polls predicted, by a landslide.
For months, the same pollsters predicted a solid win for Democrat Ethan Berkowitz over incumbent Republican Rep. Don Young in the battle for Alaska's lone U.S. House seat.
Then everyone voted.
Surprise! A Young victory looks all but assured. And Begich is behind in the count -- though he still hopes to eke out a win. Even with more than 55,000 absentee and early votes still to be counted, neither Democrat could surge ahead to the kind of margin predicted by pollsters.
"I'm not happy how that came out," said pollster Del Ali, of Research 2000, a Maryland firm that tracked the House and Senate races in Alaska.
Begich was trailing Stevens by about 3,000 votes, as of Wednesday. But as recently as last week, Research 2000 reported Begich was leading Stevens by 22 percentage points. Daily Kos, a left-leaning Web site, paid for the polls.
Skeptics will assume the numbers are slanted, but consider that Anchorage pollster and political consultant David Dittman, who often works for Republicans and is a Stevens supporter, predicted a solid Begich win the day of the election.
"If I look at what we've done around the country, we're pretty much in the ballpark everywhere," Ali said.
Another national polling firm, Rasmussen Reports, accurately predicted every Senate race in the country within the margin of error in their most recent polls -- again, except for Alaska.
Local pollsters didn't fare any better, with Ivan Moore, Craciun Research Group and Hays Research Group all reporting Berkowitz and Begich leads in recent weeks.
Even the presidential race proved tough to call. Hays and Dittman said Republican John McCain and running mate Gov. Sarah Palin were only leading by 3 to 5 percentage points this week.
But the voting results on Wednesday showed the Republican ticket was winning Alaska 61 to 36 percent.
Why were the Alaska predictions wrong? Pollsters offered several theories.
The answer lies in voter turnout, Moore said.
"The pollsters didn't succeed in predicting who was going to vote very well," Moore said.
Pollsters try to find out if a person will actually vote by asking them questions, or looking at how often they've voted in the past, he said.
But this year, Palin's role in the presidential race drew Alaskans to the ballot booth who wouldn't normally vote. And they tended to vote Republican, he said.
Dittman disagreed, saying it wasn't that Palin brought in new voters, but that many Democratic voters stayed home.
Even considering the tens of thousands of ballots still to be counted, turnout is lower this year than in the last presidential election, he said. One theory is that young and first-time voters who Democrats were counting on didn't show up when it became clear Barack Obama was already headed to victory.
Ali wishes he'd polled Alaska a little bit closer to the election -- and not so soon after a jury found Stevens guilty in federal court of lying on his financial disclosure reports, when voters were most likely to think poorly of Stevens.
Alaskans are independent. Maybe they lashed out at the criticism of Stevens, he said. "As the national media and pundits were on his back, it could just be that voters said, 'You know what, I'm going to vote for the guy out of spite.' "
Young's lead is more surprising, given how unfavorably he polled, Ali said.
"I'm just trying to figure out if there's something Berkowitz did, or didn't do, in that final week that changed minds."
As for the Senate race, even polls released just before the election and after Stevens had returned to Alaska for a round of last-minute campaigning showed a Begich lead.
Dittman projected a 53-to-45 percent Begich win Tuesday. A Hays Research Group poll conducted Sunday reported Begich ahead by 8 percentage points.
"We were as surprised as anyone about the Alaskan election results," wrote Adam Hays, in an e-mail. "Of course, hindsight always has a zero percent margin of error!"
So why should Alaskans trust poll numbers next time there's an election? Pollsters insist this kind of broad misfire only happens once every few decades.
"If we were kind of permanently wrong, people wouldn't hire us anymore," Moore said.
Find Kyle Hopkins' political blog online at adn.com/alaskapolitics or call him at 257-4334.
Poll numbers: What happened?
U.S. House: Berkowitz (D) versus Rep. Young (R)
Nov. 2: Berkowitz leads 48-41 (Hays Research)
Oct. 29-Nov.2: Berkowitz leads 53-46 (Dittman Research & Communications)
Oct. 28-30: Berkowitz leads 53-44 (DailyKos.com, Research 2000)
Oct. 24-26: Berkowitz leads 47-38 (Craciun Research Group)
Oct. 17-19: Berkowitz leads 51-43 (Ivan Moore)
Stevens and Begich poll numbers, Back Page Polls: Begich versus Stevens
Nov. 2: Begich leads 48-40. (Hays Research)
Oct. 29-Nov.2: Begich leads 53-45. (Dittman Research & Communications)
Oct. 28-30: Begich leads 58-36. (DailyKos.com, Research 2000)
Oct. 24-26: Begich leads 48-36. (Craciun Research Group)
Oct. 17-19: Begich leads 46-45. (Ivan Moore)
Source: Pollster.com, Craciun Research Group, Dittman Research Group.