DES MOINES – Iowa pollster Ann Selzer has made a name for herself in the US after calling presidential election contests correctly for 15 years.
In 2008, during the pivotal Democratic nomination showdown in the state between Barack Obama, Hillary Clinton and John Edwards, among the pollsters she alone predicted the one-term Illinois senator would win. (Obama bagged 38 per cent of the vote, Edwards secured 30 and Clinton 29. He went on to win the election.)
The pollster then in 2016 correctly predicted Donald Trump would beat Ms Clinton in the state, while others had the race much closer. Polls can be off by a country mile, and a case in point were the general election polls of 2016, which almost without exception predicted Ms Clinton would handily defeat Mr Trump.
Ms Selzer’s most recent poll provided a huge smile to supporters of former president Donald Trump, when it showed him on 51 points, ahead of Rick DeSantis on 19 and Nikki Haley on 16. It was the first time a candidate had enjoyed a simple majority.
With numbers like that, is there any way Mr Trump can be stopped? “Here’s the thing: having been doing this for as many caucus cycles as I have, I’ve seen everything happen,” she tells i, recounting how she was in the minority again in 2012, when she predicted Pennsylvania senator Rick Santorum would come in narrowly ahead of Mitt Romney, Ron Paul and Newt Gingrich in the Republican race.
While Mr Santorum had been in the single digits in their first sampling, night by night, his numbers grew. “And so the story was his trajectory, and that he could win,” she says. “And by a handful of votes he did.
“So anything can happen. Anything.”
While political polling has assumed increasing importance, it has become much harder to do. At least to do accurately.
The main reason is that the way pollsters contact people has changed. In 2000, 95 per cent of US households had a landline; today, that number is just 27 per cent, while 97 per cent own mobile phones.
And while history has shown polls conducted with people who have landlines and a known, fixed address, are more likely to be more accurate, trying to continue such methods is more expensive. When pollsters cut corners, they lose accuracy.
As voters in Iowa head to the caucuses on Monday to select a presidential candidate for the Republican Party, Ms Selzer and her celebrated polls stand out for getting things right, when so many get it wrong.
“Some people have turned to other methods, and I really have not,” Selzer tells i. “The reason is these other methods start taking steps away from random selection, which is the heart [of the process].”
Some people may use online polls, or turn to social media. But she says she does not trust them.
To start with, if people are volunteering for a poll, does that display some sort of inherent bias itself? Then there’s the challenge of trying to decide how representative an anonymous online poll is with the actual electorate.
“It could be 1,000 people, it could be 5,000 people. What are your methods for making your respondent pool look like a future electorate?” she says.
“I call what I do polling forward. I want my data to reveal what the future electorate looks like.”
She adds: “Many pollsters do very sophisticated modelling, but their models are based on the past. So they’re very sophisticated at predicting past elections. I want to be in a place where the future will reveal itself to me in my data.”
Selzer, 68, who in 1996 established Selzer & Company and whose polls are routinely published by The Des Moines Register, says she found her way into polling after becoming fascinated by data as a college student.
“I was always interested in gathering data about what people think … the idea I could learn something from data,” she says. “It wasn’t that I was a maths whiz or anything. And while I was in graduate school, I got a fellowship to Washington for a year and worked on Capitol Hill. So politics, and data equals polling!”
Experts say polling is notoriously difficult for the caucuses, even more so than for primaries, or other elections.
On caucus night, a person has to physically attend a show of hands for a candidate in a church hall or school gym, at an event that lasts just a couple of hours.
There is no absentee voting, and in Iowa they are invariably held on a freezing winter night, meaning anyone who has to work that evening cannot take part, and neither can anyone who is ill. For the elderly, it is a particular challenge. (The forecast for caucus night is -18°C/0.4F.)
Ms Selzer was described as “the best pollster in politics” by respected polling website FiveThirtyEight. Professor Tim Hagle of the University of Iowa, who says Ms Selzer is “very good” says the nature of the caucuses means candidates cannot simply show up and speak to voters. It is essential they have teams on the ground to organise supporters.
“It’s very cold and dark and especially some older folks don’t like to go out at night… Those are the reasons why it’s hard for a pollster to figure out who’s actually going to show up,” he says.
“It also why it’s important to have a good ground game, meaning you have volunteers that are making sure the people in their precinct know where to go, but also that they show up because otherwise… the excuses pile up.”
He added: “You’ve got to make sure as a candidate your supporters are sufficiently energised and so that’s where Selzer comes in, because she understands that and understands Iowa politics, and was able to usually properly weight it.”
A key factor in Ms Selzer’s correct 2008 prediction was was that her team was discovered that as many as 60 per cent of those planning to take part were first time voters, suggesting many younger voters had been attracted by Obama’s optimism.
“It was controversial,” she says, recalling that other pollsters assumed previous caucuses would be like the new caucus. “They were blindsided when the first exit polls showed 57 or 58 per cent were first time voters. It changed my career.”
Ms Selzer says she will publish one more poll before caucus day on Monday. The nation will be watching.