You really have to squint/cherry-pick to see a Trump comeback in the national polls. The IBD/TIPP tracking poll has bounced around a bit in a vaguely pro-Trump direction, I guess. But the other two tracking polls (USC and SurveyMonkey) show a steady/increasing Biden lead.
There's also a lot of non-tracking poll national data, and it's just become very routine to see double-digit leads for Biden, especially in the higher-quality polls.
The other feature: the state polls *aren't* as strong for Biden as the national polls, although there haven't been that many high-quality and/or live-caller state polls lately.
Online and IVR state polls tend not to show that much change since the debate. Maybe they imply that Biden is +8 nationally or so.
Some of the live-caller state polls we do have could be consistent with a larger Biden lead, though, of +10 or thereabouts.
The non-live state polls are a weird mix. Certainly some of the higher-quality online polls (e.g. YouGov or Ipsos) are worth paying attention to, although some of them (especially YouGov) tend to do a lot of weighting in ways that make them very steady relative to other polls.
There are also a lot of quasi-partisan state polls in the form of, say, Trafalgar or InsiderAdvantage on the R-leaning side, or Civiqs and PPP on the D-leaning side. Some of their polls officially get classified as partisan by 538 and some do not, depending on who the sponsor is.
I don't want to paint all of these firms with the same brush. Civiqs is doing good work, I think. But I have to say... with *some* of these firms, it feels like they're very self-conscious about what their numbers say relative to 538/RCP averages and I don't love that.
This interview with Trafalgar Group really worried me, for instance. These are not conventional polling methods and it sounds like they provide a lot of room for subjectivity or frankly confirmation bias in how one constructs a voter universe.
Also worth revisiting the old @Nate_Cohn critique of PPP. Not because PPP is necessarily still up to this, but because it shows how a pollster can be collecting real data but applying dubious assumptions to it rather than letting the data speak for itself. newrepublic.com/article/114682…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Folks, Biden's lead didn't shrink from 7.3 points to 3.6 points in PA in a week (as per RCP) at the same time it was steady or slightly growing nationally. This is why you need poll averages that take a longer time horizon and/or adjust for house effects. projects.fivethirtyeight.com/polls/presiden…
RCP's averages are extremely subject to who happens to have polled the state recently, which is often the spammier, lower-quality pollsters, and that's been especially true recently with live-caller polls not having been terribly active in the states over the past 2 weeks.
I love many things about RCP, but if you have an average and 1/3 of it consists of Trafalgar and InsiderAdvantage and 0% of it consists of live-caller polls, it's not going to be a very reliable average.
A comparatively good morning of polls for Trump, although it says something about the state of play when you see a poll showing him 9 points behind nationally and say to yourself "hey, not bad!".
To reiterate this point, on one of Trump better days of polling recently, he only decreased Biden's odds from a 87.5% chance of winning the Electoral College as of our final model run last night to 87.2% now. (Not a statistically significant change.)
Part of that is because Trump's better polls, in the context of what we've seen recently, means state polls that look like they did before the debate (i.e. consistent with a 7-8 point Biden national lead), which is still not a great position to be in with 14 days to go.
Unless Stepien thinks Trump would be so bad that debating would lower his comeback odds, even though Trump generally wants high-variance strategies.
It's not crazy: Trump historically loses ground following debates and his messaging has been even more erratic than usual lately.
Like, if this is what's coming out of Trump these days, it's not clear that Stepien wants him debating, especially if he could also impact downballot races.
Although everyone's waiting for that last twist of fate as with the Comey letter in 2016, more often the final two weeks of a campaign can be anticlimactic, with it being too late to shift tactics or change that many minds.
But everybody's anxiety level is very, very high. So there's usually a lot of jumping the gun at minor stories, any claims about shifts in the polls, etc. Looking at polling averages can be helpful in this regard: projects.fivethirtyeight.com/polls/presiden…
Even a 10-point lead isn't safe for Biden because 1) it's closer in the swing states 2) there's still *some* time for the race to tighten 3) polls can be wrong (although they'd have to be quite wrong, not just a little wrong, but I digress...)
So far, returned mail ballots are D +31 (not a surprise given what polls show) whereas the set of mail ballots that have been requested but not yet returned are "only" D +17. Why does this matter?
If (as this data shows) Dems are returning their ballots sooner, then mail ballots cast closer to Election Day—which in many states, will also be counted later—may not be as D as mail ballots overall. They could even wind up being R-leaning, conceivably.
This means there might not be as much of a blue shift in states with late-counted mail ballots, especially ballots that arrive after Election Day in states that allow that, as people might assume.
By the slimmest possible margin, I might caution. Biden has a 51% chance of winning Georgia and Trump 49%, per our forecast. Our model's priors favor Trump in Georgia but there's been enough good polling for Biden there to counteract them. projects.fivethirtyeight.com/2020-election-…
Meanwhile, in the Georgia Senate special election, we have Loeffler (R) with a ~40% chance of winning after the runoff, Warnock (D) with a ~30% chance, and Collins (R) with a ~30% chance.
Ossoff (D) with a ~30% chance of winning the regular Georgia seat.