Democrats, buoyed by Vice President Harris' debate performance and heartened by her rising popularity in the polls, have one doubt that bothers them more than anything else.
It's the knowledge that former President Trump has consistently outperformed in the polls in the past, and the fear that that could happen again this time.
The idea of history repeating itself is a nightmare for Democrats, given that many polls already show the votes in battleground states tied.
“I think we still need to worry about a Trump surge,” Celinda Lake, one of two lead pollsters for President Biden's 2020 campaign, told this column in an email.
One of the biggest challenges for pollsters is how to model voter turnout, which mostly involves guessing how many people from which demographics will actually vote.
That's one reason Lake is concerned.
“Trump is winning over men who haven't voted before,” she said. “Most pollsters are adjusting. Ours is now looking at two estimates of voter turnout: one is the average, the other is the [a] “Trump surge.”
If such a surge were to occur, Harris would have no leeway.
An average of polls maintained by The Hill and Decision Desk HQ (DDHQ) gives Harris a 3.4-point lead nationally, but the race is close in most key states.
In three “blue wall” states crucial to Harris' hopes – Pennsylvania, Michigan and Wisconsin – she leads by 0.4, 0.8 and 3.0 percentage points, respectively.
DDHQ currently estimates Harris has a 54 percent chance of winning in November.
Scott Tranter, the organization's director of data science, warned Democrats not to place any trust in the numbers.
In probability terms, a 54 percent chance of winning essentially means that in a hypothetical scenario of 20 elections, Harris would win 11 times and Trump would win nine times.
“If a 54 percent chance makes sense to you, then you need to understand a little bit more about probability,” Tranter said. “It's like a coin flip. Just like you wouldn't be surprised if you flipped a coin and it came up tails, nobody should be surprised if Kamala Harris wins or Donald Trump wins.”
With public attention focused on the intensity of this year's election campaign, it's worth highlighting exactly how Trump has missed the polls in the past two presidential elections.
In 2016, national polls measuring Trump's approval rating against Democratic candidate Hillary Clinton did not have a wide margin of error.
The final RealClearPolitics (RCP) average before Election Day had Clinton beating Trump by 3.2 percentage points. The actual results showed that she beat Trump in the popular vote by 2.1 percentage points, but of course lost the Electoral College vote.
The result was so shocking in part because polls were so off in crucial states: Behind the “blue wall,” pollsters measured Clinton's results remarkably accurately, but grossly underestimated Trump's share of the vote.
In the worst case, Wisconsin, the final RCP average predicted Clinton's vote share to within 0.3 percentage points of the share she actually received.
But the projections underestimated Trump's share of the vote by about seven percentage points. Trump narrowly defeated Clinton with 47.2% of the popular vote, well above the RCP's final average projection of 40.3%.
Following the shocking results in 2016, a variety of explanations have been offered for why polls may have underestimated Trump's approval rating.
These included suggestions that pollsters' turnout models were flawed and underestimated Trump's rural and suburban turnout, that Trump supporters trusted polls and the media less than Democrats and therefore were reluctant to respond to surveys, and that some respondents simply lied, supporting Trump in the privacy of their polling places and not telling pollsters they intended to do so.
While the 2016 election results prompted much soul-searching among pollsters, the industry as a whole didn't perform much better in the 2020 election.
In fact, national polls were way off in 2020 than they were in 2016. The final RCP average predicted a 7-point win for Biden, but the actual popular vote showed Biden winning by 4.5 points.
Another factor remained constant: Pollsters were good at predicting the approval ratings of the Democratic candidates but bad at predicting Trump's approval ratings. The final RCP average was within half a percentage point of Biden's actual vote share, but underestimated Trump's share by about 3 percentage points.
Of the three “blue wall” states, Wisconsin was again the biggest disappointment, with the final RCP average underestimating Trump's actual vote share by about 5 percentage points.
A misfire of this magnitude puts current polling in a tough spot in this year's race: As of Thursday evening, The Hill/DDHQ polling average of seven battleground states showed only one state in which either Trump or Harris had a lead of more than one percentage point.
The state in question, where Harris has a three-point lead, was Wisconsin.
There haven't been enough polls since Tuesday's debate to get a firm idea of whether the split has brought about any fundamental change in the race.
Harris is widely seen as the winner, and Trump insisted in a social media post on Thursday that there would be no more debates.
Also on Thursday, Trump campaign pollsters Tony Fabrizio and Travis Tounies issued a memo, released by the campaign, saying Trump's approval rating had increased by two points after the debate while Harris' approval rating had “remained flat.”
But a new Morning Consult poll, also conducted after the debate and released Thursday, showed Harris holding a 5-point lead nationally, the largest lead of any poll the organization has conducted to date.
Harris' campaign continues to insist that she is behind, but questions remain about whether her aides truly believe that or are using the claim to keep supporters highly motivated.
One thing is certain: the election is extremely close and Democratic anxiety will likely continue until the results are known.
This note is a reporting column by Niall Stanage.
