What Polling Got Wrong With the 2020 Election

Pollsters Robert Cahaly and Patrick Murray weighed in on what went wrong — and what they can do better next time. 

Though Joe Biden emerged as the winner of the U.S. election as many had predicted, the polling industry has drawn fire for missing the mark once again after the race turned out to be tighter than expected.

It’s impossible to quantify the extent to which the polls were off until the vote count is complete — but it’s clear President Trump overperformed in many battleground states like Michigan and won comfortably in states, such as Iowa, that were considered to be toss-ups. 

Public polls have also been the subject of much ire for the president. He has blamed the polls for his defeat, falsely claiming that they were “election interference, in the truest sense of that word.” 

Pollsters Robert Cahaly and Patrick Murray offered some explanations as to why polling saw another disparity — including one popular theory that argues that polls might be wrong because of “shy Trump voters.”

Cahaly, who is a chief pollster at Trafalgar Group, told Wake-Up Call that the polling industry overall “absolutely” made the same mistake as it did in the 2016 presidential election when it came to this hidden Trump support. Trafalgar was the lone polling group that predicted Trump’s victory in 2016 and saw this year’s election as a repeat. 

“They’re absolutely making the same mistake and I would say not for the second time, I would say for the third time,” Cahaly said of his fellow pollsters. “In 2016, we suggested there was a hidden Trump vote. We didn’t invent the concept of the shy Trump voter — we kind of discovered it as a lot of people are discovering it in the primaries.”

Cahaly said hidden Trump voters aren’t as forthcoming about their support due to a social desirability bias — something that he thinks had a detrimental effect on Trump’s poll numbers. 

“It always comes down to either a polarizing figure that people are trying to hide their support for or someone who would make them look more enlightened that they are trying to act like they support who they really don’t,” he said. 

Perhaps this was nowhere as clear as it was in Wisconsin, where Trafalgar’s final poll showed Biden leading by one point, which is sure to be far more accurate than other polls like the final ABC News/Washington Post poll, which had the Democrat up by 17 points in the state.

Murray, who runs the Monmouth University Polling Institute, believes social desirability bias was even more of an issue in the 2020 election than it was in 2016 because of increasing political polarization. He said a voter’s views of Trump’s job performance was more of a reflection of them rather than of the president himself.

“People are seen in many ways through their views of the president,” Murray said. 

Both pollsters said Trump voters were also generally less willing to participate in polls. Cahaly explained that pollsters had to “dig extra hard” to get responses.

“They just refused to participate in the polls,” Cahaly said. “So you had to work very hard to get them even a fair sample.”

Cahaly argued that this social desirability bias also could have potentially thrown off polling for down-ballot races, where polling performance appeared even worse than at the presidential level. Democrats thought their party was favored to win back the Senate and gain House seats. Instead, their path to the Senate majority is much narrower than before, and they lost seats in the House.

Cahaly pointed to Sen. Lindsey Graham’s battle with Democratic challenger, Jaime Harrison, who is Black, as a prime example. Cahaly said some South Carolina voters were more inclined to say they were supporting Harrison when they were actually supporting Graham because they didn’t want to seem as if their vote had to do with race. 

“It’s not partisan — it’s just reality,” he said. “People want to look good with somebody they don’t know. That’s not strange. So what we saw this year is the Republicans didn’t want to say anything.”

But, ultimately, this hidden Trump support wasn’t enough to help the president win in states he needed in order to clinch the White House. Cahaly pointed to the president’s decline in support among seniors. In October, Biden was leading the president among those 65 and older by the double-digits according to two separate polls in October. Many observers have attributed this to the fact that seniors have been disproportionately affected than the coronavirus pandemic. 

“Donald Trump had a way of alienating seniors like no Republicans ever had,” Cahaly said. 

Cahaly said even though the president lost support among seniors, he was able to add scores of Latino voters in key states like Florida and Texas that swung in his favor. Exit polls showed that Trump captured 32 percent of the vote among Latinos, up four points from his performance in 2016.

Murray bemoaned pollsters’ failure to predict this uptick, saying they made the mistake of over-overgeneralizing Latino voters and not taking into account their diverse political preferences. This was something even Monmouth University underestimated.

In September, Monmouth found that  58% of Florida Latinos supported Biden, compared to 32% who favored Trump, suggesting a healthy lead that didn’t actually materialize. 

“There’s no such thing as a monolithic Latino vote, there are many different aspects of that vote,” he said. “Florida’s actually the best example because even within Florida, there are so many different types of Latino vote and different communities there.”

Cahaly said if there’s one lesson that ought to be learned from the 2020 election, it’s that Black and Latino voters are never taken for granted again.

“When a community gets respect for the voting bloc they have, then America’s better,” he said. 

As to what the polling industry can learn for next time, Murray said that pollster can do a better job of conveying levels of uncertainty as polling isn’t an exact science. But he remains skeptical as to whether that will be accepted by the public at large.

“People want certainty and they will not accept the fact that the best tool that we have to do this still has a lot of error involved in it,” he said.

This story was written and reported by Tess Bonn.