Four days before Election Day, there’s a lot of fresh presidential and Senate polling of widely varying quality, with more to come over the weekend. At this point, the averages aren’t moving very much. At RealClearPolitics, Joe Biden’s national lead over Donald Trump is at 7.8 percent; it was 8.1 percent a week ago. And at FiveThirtyEight, Biden’s lead is at 8.9 percent; it was 9.9 percent a week ago. Yes, Biden’s national lead is slowly declining, but at this pace it won’t get down into the danger zone before Tuesday, which is why Democratic confidence is pretty high.
But those who need a jolt of super-confidence as the deal goes down can certainly find outlier polls telling them what they want to hear. Unsurprisingly, there are lots more of those on the Biden side of the street given his longstanding lead. Final polls giving the Democrat a double-digit lead among likely voters include Economist/YouGov (54-43), Reuters-Ipsos (52-42), and CNN (54-42). It’s not a final reading, but the USC Dornsife tracking poll — which famously predicted a Trump win in 2016 — has Biden up 54-42 as well. Trump hasn’t actually led in a major national poll since a Rasmussen survey back in September. Rasmussen now has him trailing by a manageable three points, and he’s within striking distance according to The Hill-HarrisX (four points) and Emerson (five points).
Polling in battleground states is pretty stable, too. Biden has small but relatively steady leads (using FiveThirtyEight’s averages) in Arizona (3.1 percent), Florida (2.3 percent), Georgia (1.6 percent), Iowa (0.3 percent) and North Carolina (2.3 percent). He has bigger leads in Michigan (8.8 percent), Nevada (6.1 percent) Pennsylvania (5.1 percent) and Wisconsin (8.5 percent). Trump probably needs to win all those states in the former category (definitely Florida, Georgia, and North Carolina), and at least one in the latter, to get to 270 electoral votes.
If you want some less tentative state polls, though, we’ve got ’em. Biden fans can take comfort in a Florida poll from high-quality Monmouth showing him up 51-45 among likely voters in a high turnout scenario, which is what is emerging. Monmouth also showed Biden up 50-46 among likely voters in Georgia. The mother of all late pro-Biden outliers is the ABC-Washington Post survey of Wisconsin, giving him a massive 57-40 lead among likely voters. More modestly, another high-quality poll, from NBC/Marist, gives Biden a 52-46 lead in North Carolina.
Wherever Rasmussen and Trafalgar Group conduct state polls, Trump does better than in other surveys (viz. Florida, where the former has the president up 50-47 and the latter 50-46). He got some late good news from Quinnipiac, which has been showing Democrats with extravagant leads all over the country: in Iowa, Quinnipiac has Trump up 47-46. And in crucial Pennsylvania, InsiderAdvantage shows Trump ahead 48-46 among likely voters.
You can find some rich outliers in Senate races, too. The same Quinnipiac that showed Trump regaining the lead in Iowa gave Joni Ernst (who has trailed Theresa Greenfield in most polls this fall) a 48-46 lead. Or maybe it’s a trend rather than an outlier, since New York Times/Siena also showed Ernst narrowly ahead (45-44) going down the stretch. Another Senate race where suddenly big leads are appearing that may represent a trend rather than an aberration is in Michigan, where three late polls (New York Times/Siena, Detroit News/WDIV, and MIRS/Mitchell Research) show Democrat Gary Peters’s lead over Republican John James blossoming into high single-digits.
But in a race that many view as the key to Senate control, in North Carolina, there’s sensational late-polling news for Democrats from NBC/Marist: Cal Cunningham blowing out to a 53-43 lead over Thom Tillis. It’s been a week since Tillis has been able to claim a lead in a survey by — you guessed it — the Trafalgar Group!
Soon enough votes will replace polls, and we’ll see if national surveys improve on their relatively good record (mythology aside) in 2016, and if the more abundant state polling is as far off as it was four years ago. Until then hopes springs eternal among those willing to cherry-pick the data they want.