Pollsters don’t make predictions. We just show how things are at a given moment in time. It’s such a simple concept but one that is routinely misunderstood, mostly deliberately and for partisan purposes. Whether it’s to score political points, or get a few clicks on your blog, criticizing pollsters has become the new blood sport.
OK, bring it on. But here’s a story to illustrate how challenged and baseless these attacks are.
Back in 2002, I was lucky enough to work on Fran Ulmer’s campaign for governor. She had served as Alaska’s Lt. Governor for the previous eight years under Tony Knowles and was the Democrat in the governor’s race. She was up against U.S. Sen. Frank Murkowski, who had returned from D.C. to run as the Republican candidate.
In late August, right after the primary election was complete and the contests were determined, our poll results showed Ulmer in a 3-point lead, 46-43. Ulmer’s positive negative rating was 63-27, Murkowski’s 56-36. My candidate was very well liked and respected. In early October, a month out of the November 5 election date, we did another survey that showed Ulmer with a four point lead, 47-43. I remember another Alaska pollster did a survey for KTUU at the exact same time and showed an essentially identical result, Ulmer in the lead by 3-4 points.
As we went through October, the race tightened up. Ulmer’s positive rating was dragged back into the 50s, by mid-month the race was tied, and by the third week of October, Ulmer had fallen two points behind. In the final 10 days before the election, we started nightly tracking. On the Monday eight days out we were four points down, by Wednesday it was 7, and by the time the weekend arrived, it was 10.
Murkowski won the following Tuesday by 15 points.
Of course, nothing’s worse on a campaign than having cratering internal numbers like those taking the wind out of your sails. But our obligation was to put on happy faces and go out there and finish it up. If memory serves me, I don’t think there was even full disclosure with our client, such was the need to maintain morale and get to Tuesday.
In the 16 since, if I had gotten a dollar for every time some armchair quarterback had criticized me for “getting Ulmer-Murkowski wrong,” for crying out loud, I’d have a very large wad of cash in my pocket right now. The fact was, my polling wasn’t wrong. It matched up with other poll results at critical points during the campaign, and then I watched it shift, gradually at first, but then suddenly and precipitously at the end, in front of my eyes, because Alaskans decided (to their cost as it turned out) that they were more comfortable with Frank.
Nate Silver at fivethirtyeight.com claims that Alaska “is difficult to poll,” like he thinks we live in igloos or something and don’t have phones. The reality is, there’s nothing difficult to poll about Alaska itself; it’s Alaskans who are difficult to poll, because we change our minds.
What I was coming to understand prior to that campaign in 2002, and have learned since, is that this movement is an Alaska Shift, a systematic surge in Republican vote that has happened countless times before. It happened the first year I worked in politics, in 1990, when John Devens had Don Young’s campaign on the ropes and convinced they were going to lose, but come election day Young pulled out a miraculous recovery to win by 4 points. It happened in 1994, when Tony Knowles managed to lose a double-digit lead in the final days of the campaign and only scrape home by 500 votes. More recently, it happened in 2008 and 2014 when Don Young was challenged by Ethan Berkowitz and Forrest Dunbar, both of whom led in polls but lost hard on election day, by 5 and 10 points respectively. And it has happened again this year, with Begich-Dunleavy and Galvin-Young.
All the Democrats I’ve mentioned, perhaps none more so than Fran Ulmer, were likeable, accomplished and qualified candidates, and all of them made good initial impressions in their races. But there’s a sizeable subset of the Alaska electorate who, while they’re predisposed to personally liking the Democrat in a race at least at first, get their amygdalas lit up when campaign messaging turns, as it inevitably does in Alaska, to fear-based, nonsense talk about the things Democrats will do if we’re stupid enough to elect them. Raise taxes, grow government, spend uncontrollably and steal your PFD being principal among them. At the same time this is going on, the pleasure centers of those same voter brains are rendered awash with serotonin with promises of the continuation of the free ride. That was the case with Murkowski in 2002, and this year’s promise of a bumper dividend is a repeat performance.
The final poll I did in the governor’s race this year was done the penultimate weekend, about 10 days out of the election. It showed Begich and Dunleavy well matched and in a close race. That poll was perfectly representative of likely Alaska voters, by party affiliation, by age, by ethnicity and by geographical region. In the 10 days that followed, we saw the ritual demonization of the Democrats on the ballot reach a full crescendo, we saw Francis Dunleavy drop a chunk of his Alaska investment into the final week in support of his brother, we saw the oil industry spend untold millions promoting and getting out the “No on 1” vote and, like the cherry on top, we had Mike Dunleavy’s promise of a $6,700 payday and a restoration of the old PFD formula.
Who wouldn’t expect the race to fundamentally shift after all that?
Listen, is every poll that I do bang on target? Of course not. Like any survey measurement, we are approximating the characteristics of the population we’re studying, and it is by definition inexact. Have I had outlier polls before that I would take back if I could? Sure, every pollster has. But outliers are a different thing. They occur because of random variation, and don’t demonstrate the kind of systematic consistency I’ve described here.
Don’t worry, be happy. No taxes. Free money. Alaskans simply like that message and instinctively respond to it with their votes. Over and over again.
Ivan Moore is an Alaska pollster and owner of Alaska Survey Research.