National Post (National Edition)
Blue wave falls to wash up for Democrats
Trump backers `shy' to admit support: Pollster
OTTAWA• Pollsters had projected a big blue wave would sweep Republicans out of office across the United States, but it appears to have hit the shore with nowhere near the expected force.
As of Wednesday afternoon, former vice-president Joe Biden still looked like he could garner a win, but rather than the landslide victory many pollsters had predicted he was set to pull out a close victory over President Donald Trump.
Pollsters also predicted the Democrats would take control of the U.S. Senate and keep control of the U.S. House to give them complete control over the legislative process, but most of those
Senate wins were failing to materialize.
W. Joseph Campbell, a professor in the School of Communication Studies at American University in Washington D.C., said the polls raised Democratic hopes and there are going to be a lot of questions for pollsters.
“The broad expectation was, and this is a poll driven expectation, that the Democrats would have something approaching a blue wave,” he said. “It's at best a very mixed performance for polls.”
Among the polling misses were Florida, where pollsters gave Biden as much as a five-point lead in recent weeks. North Carolina also consistently showed a Biden lead, but could end up in the Trump column when the counting is complete.
In Wisconsin, where Biden is forecast to win by about half a percentage point, polls had indicated he was as many as 17 points ahead of Trump. Biden was also predicted to win easily in Pennsylvania, but that race was proving to be a nailbiter with Trump ahead for most of the day.
Republican Senator Susan Collins, from Maine, had no major polls showing she had a chance to hold onto her seat, but she still managed to beat her challenger on election day.
Campbell said the polling misses are from all across the country.
“There are all kinds of examples you can find in polls that really miscalled the race and in particular states, especially,” he said.
In 2016, pollsters were generally thought to have missed pockets of Trump supporters and under-represented the view of non-college educated white voters in particular. Campbell said he believes it is too early to determine what went wrong this time, but he suspects it is something new.
“When polls go bad, it's for reasons that are not shared from polling failure to polling failure,” he said.
Quito Maggi, president of Mainstreet Research, a Canadian polling firm, said he believes some Trump supporters were shy about admitting their support.
“They're not the ones going to the rallies. They're not waving the flags. They're not wearing the MAGA hats, but in their heads they know they're going to vote for Trump.”
Maggi said American polling firms are expected to deliver deep dives into the results to give results broken down across a wide group of demographics.
“Every poll that is published, people expect to see not just age and gender, like what you might see here in Canada, but by income bracket, by education, by race,” he said.
Polling companies reach out to hundreds or even thousands of people, but will often get a disproportionate responses from certain demographics and have to weight their sample accordingly. For example, if a pollster reached out to 100 people and got responses from 25 women and 75 men, it would have to give more weight to the women's response to reflect the population, because in the broader population men and women are in equal numbers.
Maggi said adding that kind of weighting to account for race, income, education as polling companies do in the U.S., increases the chance of a major error.
“When you start applying, race, income, education, all of those other factors. It's the possibility for errors that grows exponentially.”
Maggi said pollsters in the U.S. also have to consider whether voters will actually get to the polls, because of issues around voting registration and long lines.
He said he is hopeful they will be able to figure out what they have been doing wrong and address it for future elections.
“They've had two elections in a row now. They're going to have good data to model it. So hopefully, by the next election, maybe the midterms, they'll make the necessary adjustments.”