Polls get it wrong

Print More

(Host) Polls conducted for Vermont news organizations in the weeks leading up to the election all put Doug Racine ahead in the race for governor. The last poll, released days before the election, showed Racine with a ten-point lead, the largest of the campaign.

VPR’s Steve Zind looks at why the voting results were so different from the poll results.

(Zind) Like many other people, Frank Bryan’s sense of how the race for governor would turn out was shaped partly by the results of polls. Bryan is a political science professor at the University of Vermont. Bryan says as the early returns came in on election night, he realized the results would be different than those indicated by the polls.

Voters can change their minds from one poll to the next. But Bryan says the poll released last weekend clearly got it wrong. Nothing happened in the days just before the election that would have changed the thinking of voters. If anything, Bryan says, the poll results should have helped Racine.

(Bryan) “That should have created a bandwagon effect that would have helped Racine. That’s why the person that appears to be sliding is always upset when those kinds of polls come out. They don’t say, ‘Oh boy, now my people will work harder.” They’re more worried that their people will say, ‘Well, why bother.'”

(Zind) According to Del Ali, who’s done polling for a number of Vermont media outlets including VPR, reputable firms gather information in pretty much the same way. But Ali says accurate polling has become increasingly difficult. Unlisted cell phone numbers and wary households armed with caller ID to ward off unsolicited calls have made it harder to get a truly representative sampling of likely voters.

Ali says Vermont presents an additional headache for pollsters. Vermont voters aren’t required to register with a party. It’s hard for pollsters to know, for example, how many Democrats and Republicans there are in the state, to make sure their survey reflects those numbers. Ali thinks that’s what might have gone wrong in last weekend’s poll.

(Ali) “It’s possible in that particular poll that it could have been skewed in terms of there were more likely Democrats interviewed. That is really the only conclusion I can come up with for a poll showing the obvious loser not only ten points ahead, but clearly that poll indicated that he was picking up momentum.”

(Zind) Ali says even if the poll results are accurate, mistakes are made by oversimplifying them. He says that’s often the result of news outlets looking for a scoop.

(Ali) “Everybody wants to scoop their competitors. Everybody wants the story. Demographics aren’t looked at. Gender breakout isn’t looked at enough. They’ll look at a number, they’ll see ‘Candidate A – 45%’, ‘Candidate B – 40%’ and the assumption is, ahh, Candidate A is going to win the election.”

(Zind) Ali says even the best polls don’t predict the future, they’re just snapshots in time.

For Vermont Public Radio, I’m Steve Zind.

Comments are closed.