Polling industry again faces criticism, but most got it right in presidential race - TribLIVE

kembaliui.blogspot.com

Days before President-elect Joe Biden declared victory, with millions of votes still being counted, pundits as well as regular voters were quick to condemn the polling industry for failing to gauge just how tight the 2020 presidential race initially seemed to be.

Democrats hoping to watch a blue wave roll through much of the nation found themselves fretting on Election Night over early and incomplete election results in places such as Michigan and Wisconsin. An especially far-off Washington Post-ABC News poll released in the last week of October suggested Biden had a 17% edge over Trump in Wisconsin, but Biden edged out the president there by less than 1%, or 20,500 votes, based on unofficial returns.

Trump supporters delighted in the president’s strong, 8-point leads in Iowa and Ohio. Polling averages suggested Trump would win the states by barely 1%.

“There were some big misses in individual states in terms of the pre-election polling,” said Tim Vercellotti, political science professor and director of the Polling Institute at Western New England University.

But many of the nation’s tightest races — including the presidential votes in Arizona, Georgia, North Carolina and Pennsylvania — were expected to be tight. And as more mail-in votes got tabulated, the margins narrowed in battleground states to give Biden the edge in all but North Carolina, which CNN called Friday for Trump.

“Pennsylvania is not a surprise,” Vercellotti said. “It was expected because the Pennsylvania state legislature had mandated that mail-in votes could not be counted until Election Day, that Pennsylvania conceivably could be the tipping-point state.”

Polls ‘did pretty well’

Nearly two weeks after Election Day, with the vast majority of votes counted but results not yet finalized, the majority of polls appear to be proving right in their presidential race projections, or at least within their surveys’ margins of error. Most suggested Biden would pull off the win, and Biden had a projected 306 electoral votes to Trump’s 232 as of Saturday.

North Carolina, where Trump overtook Biden by fewer than 74,000 votes by Friday, was “on the knife’s edge in pre-election polling,” Vercelotti said. In Arizona, which went for Trump by 3 points in 2016, Biden led by about 10,000 votes as of Sunday — pre-election polls had Biden winning the state by just nine-tenths of a percentage point.

“The polls were off by a lot in specific places, but they did pretty well in most states,” Vercellotti said. “The big takeaway is that polls were never designed to predict the future. That’s really not what they’re good at. They’re helpful at explaining what might be going on.”

Though local polls suggested a safer bet on Biden, G. Terry Madonna, director of the Center for Politics and Public Affairs at Franklin & Marshall College, said that as Pennsylvania continues to tabulate mail-in and absentee votes, the results look to be within the margin of error for their findings. F&M polling suggested that Biden would have a 4- to 6-point lead over Trump. He pointed out the F&M survey still would be within the margin of error even if Trump were to win Pennsylvania by 0.7%.

Biden’s statewide lead grew to 0.9% as of Saturday, with about 64,000 more votes than Trump.

”We’re still at the point where whoever wins, we would be within the margin of error of the polls,” Madonna said. “Remember these polls have margins of error and you can’t dismiss that.”

Rush to indict polling industry

Some political observers, however, are critical.

“Ohio went 8 points for Trump. That is just way off,” said Philip J. Harold, political science professor at Robert Morris University, noting that pre-election polls showed Biden winning Ohio by 1 percentage point.

Harold said what he sees as the failure of polls this year calls into question both the accuracy of polling today and the fashion in which pollsters interpret their results. He said “it’s a massive repudiation of these polls when we were assured” that changes in weighting errors related to education and response rates would enable this year’s polls to be far more accurate compared to Hillary Clinton’s unanticipated loss in 2016.

The Trump campaign has accused the misleading polls of undercounting his support as an attempt at voter suppression, or at least discouraging voters by making them think their candidate did not have a chance of winning. On the eve of Election Day, at a rally in North Carolina, Trump told supporters, “When you’re down by 17, people say, ‘I just can’t waste my time. I’m not going to stand in line.’ “

Predicting who will vote

The biggest challenge for pollsters — one that dates back to the Gallup poll that birthed the public opinion industry in the 1930s — is predicting who will show up to vote.

Although early polls typically rely on a random sample of registered voters adjusted for age, geographic area and party registration, final polls as Election Day draws near always focus on likely voters. Pollsters use a variety of questions, including interest level in the upcoming election and voting in the last three elections, to determine who should be considered likely voters.

“One innovation or change that the polls made after 2016 was to add mathematical weights by education because we do tend to get fewer non-college-educated responders than there are in the population. So you can weight that data up. But the question becomes, do you have the right ones in your sample?” Vercellotti said. “It could be that again in this cycle as in 2016 the surveys didn’t capture enough voters without college degrees, men, in particular.

The senior voting bloc favoring Trump also appears to have been underestimated in 2020.

“The perception was, or the expectation was, that because of the pandemic and because seniors are disproportionately affected by covid-19, in terms of illnesses, complications and death, that that particular group, which had been very strong for Donald Trump in 2016, might shift more to Joe Biden,” Vercellotti said. “Some of the early exit polling suggests that Trump won back some of that senior citizen support in places like Florida.”

Polls also appear to have underestimated split-ticket voters who went for Biden while choosing Republican congressional candidates.

The ‘shy Trump voter’ theory

Then there’s the narrative of the “shy Trump voter.”

Vercellotti said a better term could be the “defiant Trump voter” — or “someone who’s not going to plug into conventional political institutions” and “may simply be less likely to pick up the phone” when a pollster’s unknown number shows up.

“In order to investigate this, we have to talk to the very people who don’t want to talk to us,” Vercellotti said. “That’s the challenge.”

Polls bettered 2016, 2012

Despite early criticism, the overall performance of 2020 pre-election polling — at least in terms of the presidential race — appears to not only exceed the accuracy of projections in 2016 but also in 2012, when pollsters underestimated support for Barack Obama. And in recent years, the polling industry has been more consistent and accurate than it was for most of the 20th century.

Still, the trust-embattled polling industry has work to do to convince the public of its value and communicate to the public what pre-election polls do and do not mean.

“As a society, we have become addicted to political polls, and it’s to our detriment — it shapes our expectations. And then when reality doesn’t match our expectations, it undermines our faith in survey research,” Vercellotti said. “It’s very important for the polling industry to reflect on this, come up with some answers and reassure people that surveys are still a valuable to our everyday life.”

Natasha Lindstrom is a Tribune-Review staff writer. You can contact Natasha at 412-380-8514, nlindstrom@triblive.com or via Twitter .

Categories: Election | News | Politics Election | Top Stories