The polling industry and election forecasters suffered another embarrassing election night that will call into question the usefulness of public opinion surveys and horse race political coverage that appeared to once again underestimate President Trump’s support.
Votes are still being tallied in critical battleground states, meaning the final verdict on polling and forecasting may not be known for several days, but the results so far are much closer than most analysts predicted in the lead-up to Election Day.
The polls in Florida, Iowa, Ohio and Texas — all states won by President Trump — were mostly wrong, though Democratic nominee Joe Biden appears to have the upper hand in the race for 270 electoral votes and is the favorite to be the next president.
Pollsters failed to register the possibility that Trump would not only hold on to his support among white working-class voters in the Midwest, but that he would also add scores of new Latino voters to his coalition in battlegrounds that were flooded with polling, such as Florida and Texas.
The polling was even worse down-ballot, where Democrats were heavy favorites to win the Senate and expected to pick up five to 15 seats in the House to expand their majority. Republicans are now in a better position to keep control of the Senate and might pick up several House seats.
Democratic all-stars such as Jaime Harrison in South Carolina, who was running even in the polls with Sen. Lindsey Graham (R), were routed despite outspending their GOP rivals by tens of millions of dollars.
Democrat Sara Gideon has conceded to Sen. Susan Collins (R) in Maine, where not one major public poll released this year found Collins in the lead.
“To all the pollsters out there — you have no idea what you’re doing,” Graham said in his election night victory speech.
Not all of the polls were wrong. In battlegrounds such as Georgia, North Carolina and Arizona, it’s possible that the final spread will be in line with public polling.
Vote tallies indicated Biden is continuing to build on his national popular vote lead, although it is unlikely that he will end up winning by the double digit margins that many polls indicated.
The punditry and polling leading up to Tuesday suggested a broad victory for Biden was likely, both in the popular vote and the Electoral College.
In the FiveThirtyEight model, Biden started Election Day with a 90 percent chance of winning the White House, and analysts talked up Biden’s chances of winning right-leaning states like Iowa, Florida, Ohio and even Texas based on public polls.
Trump won Florida by more than 3 percentage points — a large margin for the perennial swing state — despite being down in both the FiveThirtyEight and RealClearPolitics (RCP) averages on Election Day.
Both the FiveThirtyEight average and RCP average showed Trump ahead by only 1 point in Texas. The final Dallas Morning News poll found Biden ahead by 3 points. As of early Wednesday afternoon, Trump leads by more than 6 points with 96 percent of the votes in, and he improved on his 2016 totals by 20 points in a half dozen of the counties with the heaviest Latino populations in the state.
In Ohio, both FiveThirtyEight and RCP also showed Trump leading by 1 point. Quinnipiac University, which routinely found massive leads for Biden all over the map, showed him ahead by 4 points in Ohio in their final poll. Trump leads by 8 points — his 2016 margin — with 90 percent of the vote counted.
Wisconsin is headed for a photo finish, although Biden led by 7 points in the RCP average and by 8.4 points in the FiveThirtyEight average of state polls. One late survey from ABC News and The Washington Post found Biden ahead by an astonishing 16 points in Wisconsin, a survey that has already become a talking point for the Trump campaign.
In Iowa, the averages showed Trump ahead by 1 or 2 points. He leads by 7 points.
The Des Moines Register, viewed as the gold standard in the Hawkeye State, nailed the Iowa race. However, when that poll was released, many experts dismissed it as an outlier.
One of the major issues for election analysts was the abject dismissal of polling data that was favorable for Trump throughout the cycle.
The Trafalgar Group was a prime target, as pollster Robert Cahaly told anyone who would listen that there was a submerged Trump vote that was not being picked up in states like Florida, Ohio and Wisconsin.
The Trafalgar polls — which were mocked by some of the biggest names in election analytics — are poised to be among the closest to the outcome in key states, and possibly in a few others.
“Polling is an inexact science, but we were more exact than anyone else,” Cahaly said.
Republican pollster Chris Wilson said the 2020 race is different from 2016, when the polls were close but many analysts ignored the data because they didn’t think Trump had a legitimate chance to win.
Rather, he said it’s more like the 2012 polling miss, when former President Obama’s support among Black voters was underestimated compared to the eventual turnout. This time around, it looks like pollsters failed to account for higher turnout among rural and working class whites, who were viewed as a tapped out demographic from 2016.
A lot of pollsters also missed the seismic shift among Hispanic voters, who appear poised to end the cycle as a key voting bloc in the Trump coalition.
However, Wilson saved his most withering critique for the data analysts who determined that Trump and the Republicans had a far worse chance of winning than they did in 2016.
FiveThirtyEight, after giving Trump a 30 percent chance to win in 2016, gave him only a 10 percent chance in 2020. Trump is on track to win all five of the states that the nonpartisan Cook Political Report rated as toss-ups.
The major congressional handicapping outlets also forecast significant gains for Democrats in the House and Senate, but it appears that most of the toss-ups and many of the lean Democratic seats are breaking for Republicans. Cook had estimated Democrats would pick up five to 15 House seats.
“The forecasters have a lot more to answer for,” Wilson said. “They’re just throwing math at data and what they’re producing has a lot more to do with their preferences than it does a real attempt to probe where the input data might be wrong, what it might be missing or what the counterfactual might be.”
Raghavan Mayur, the head pollster for the IBD-TIPP poll, which routinely found Trump doing better nationally than other polls, likened the 2020 experience to landing an airplane.
The final IBD-TIPP poll found Biden leading by 4 points nationally, which may end up being spot on or at least very close, while many others found Biden with a double-digit lead.
“Pilots trying to land their planes shouldn’t look outside the plane or you’ll get scared, just look at your own instruments,” Mayur said. “If you look outside it can wrack your nerves to see other polls where the difference is 10 point or more. But you shouldn’t look elsewhere, you just go by what you get and if it’s a kamikaze mission in the end, so be it.”