No profession is ever really “finished” after a catastrophic failure. Ask someone who opposed the Iraq war. They’ll talk your ear off about how cheerleaders for the invasion ended up failing upward in the foreign policy establishment or landing softly as “wise men” commentators in American media.
Polling won’t disappear either if it misjudges Trump’s chances again this time. Everyone will pay *some* attention to it in 2024. But the industry, and satellite industries like election modelers, will never fully regain their credibility. It’s one thing to underestimate Trump by three points in 2016, a year when you had no inkling of how strong his support might be among working-class white voters. It’s another thing to recalibrate for that, conclude that he’s trailing this time by eight or nine points instead of three — and then miss gain.
Better get the outcome right this time. All the well-meaning introspective “how we blew it” postmorterms in the world will be farts in the wind in terms or rebuilding public confidence.
GOP pollster laments how "my profession is done" if pollsters are wrong about who will win the election. pic.twitter.com/EDDIox4FhQ
— TPM Livewire (@TPMLiveWire) October 23, 2020
Modelers like Nate Silver have been reminding fans lately on social media that although Trump is currently a longshot to win (a 12 percent chance in FiveThirtyEight’s model), it’s not that long a shot. Which is true — he has about a one in eight chance if Silver’s model is right, not a one in a thousand chance. You wouldn’t write off an eight-to-one underdog in a football game. But there’s a note of anxiety in those reminders, especially since Silver’s 2016 model famously wasn’t as bullish on Hillary as some of his competitors’ models were. There were people who gave Clinton a 99 percent chance of victory; FiveThirtyEight gave her a more reasonable 71 percent chance. Silver reminds readers of that periodically too, that all the taunts about having gotten it so wrong fall a bit flat when you consider that he gave Trump a better than one in four chance of winning. Flip a coin twice and call heads or tails on each. By Election Day 2016, Trump’s odds at the presidency were better than yours were of getting both calls right, at least according to FiveThirtyEight.
As the days creep by and the polling doesn’t tighten, though, Silver’s 2020 model has slid towards greater certainty that Biden has this race by the throat. It may reach 90 percent probability of victory for him if nothing changes by November 3. If Trump ends up pulling another upset this year despite something like a 10 percent shot at winning, readers aren’t going to be soothed by after-the-fact assurances that one-in-10 odds aren’t really all that long. Missing one upset is human; missing two after you’ve had four years to retrench would suggest systemic methodological failure. In hindsight, the models will look less like complex statistical analysis than haruspicy. The most the pollsters will be able to say is that Trump turned out to be completely sui generis as a political persona and the data will turn more accurate again once American turns back to more traditional candidates.
For what it’s worth, the polls on the eve of the debate looked especially bleak for him.
NEW: Joe Biden leads in the key battleground states of Colorado, Florida, Michigan, Minnesota, Pennsylvania, and Wisconsin while Donald Trump leads in South Carolina.
— Morning Consult (@MorningConsult) October 22, 2020
Biden leading by six or more in Michigan, Wisconsin, Pennsylvania, and Florida is the makings of a bloodbath. Morning Consult also found 2016’s third-party voters breaking for Biden at a 53/21 clip, which might be sufficient to deliver him the nailbiter states from four years ago even if other demographic variables remained constant. This Gallup poll released yesterday also raised some eyebrows:
Not everyone who believes that Trump doesn’t deserve reelection will end up voting for Biden, but 56 percent in that category is a *big* number to have to overcome 11 days out from an election. Jonathan Last flagged it yesterday in a (far too confident) piece claiming that “Trump is toast,” arguing that the more it seems that turnout this year will be unusually high, the more confidence we should have that the polls will be accurate.
And right now turnout does look to be very high. Early-voting turnout is through the roof — the U.S. has already surpassed total early voting from 2016 with a week and a half still to go before Election Day — and it’s also Democrats who are mainly benefiting from it:
Democrats have opened up a yawning gap in early voting over Republicans in six of the most crucial battleground states — but that only begins to tell the story of their advantage heading into Election Day.
In a more worrisome sign for Republicans, Democrats are also turning out more low-frequency and newly registered voters than the GOP, according to internal data shared with POLITICO by Hawkfish, a new Democratic research firm, which was reviewed by Republicans and independent experts…
The GOP caught an encouraging glimpse in Florida on Tuesday, when more Republicans began casting in-person, early ballots than Democrats in Trump’s must-win state. But Democrats have dominated voting by mail and on Thursday held a historic lead in total pre-Election Day ballots cast of 463,000, or 10 percentage points, according to the state’s Division of Elections…
At a glance, the top-line Democratic margins also look huge in Arizona (16 percentage points), Michigan (24 points), North Carolina (14 points), Pennsylvania (46 points), and Wisconsin (22), according to the analysis from Hawkfish, which is funded by billionaire Michael Bloomberg, a Trump foe.
Could it be that Dems are merely cannibalizing their overall turnout by having people show up early, with Republicans set to overwhelm them on November 3? Sure. But it’s hard to grok that one chunk of their electorate would be so eager to beat Trump that they’re setting records and crushing the GOP in early voting while the remaining chunk is so ambivalent about Trump that the party will end up with mediocre turnout overall. For MAGA to beat the left this year, it increasingly appears, Trumpers will need to turn out for their guy in considerably bigger numbers than they did four years ago. It’s conceivable that they will, as there were a lot of working-class whites who sat out the last election despite Trump’s appeal who might be persuadable now. But it’s a lift. If righties do swamp the polls with a red storm, though, that’ll doubtless be the pollsters’ excuse afterward for missing so badly: No one could have foreseen turnout like this. It was a 500-year flood. You can’t hold our error against us.