FOLKS! The initial results of #Election2020 are trickling in from IN and KY. Follow this thread for my analysis and live prediction of results throughout election night.
And remember you can update our election model with your own estimates here ⬇️
And remember you can update our election model with your own estimates here ⬇️
https://twitter.com/gelliottmorris/status/1322589664797229057
https://twitter.com/PpollingNumbers/status/1323768699397033986
If these are anywhere near accurate, Trump is toast.
So, these first results are mostly from absentee votes, so should skew a bit more Democratic than expected. So I am withholding judgement until we get to 100% reporting.
We cannot use these early results to predict what's going to happen in IN/KY, but we CAN use them to compare relative swings _among_ absentee voters. And so far the suburban swings are much bigger than rural ones. We'll see that holds.
Alright. Our live forecasting model is up and running (on my computer). It starts off at 93% Biden without any calls (below our pre-election forecast for now other reason except I pushed it down to be more flexible to a Trump upset).
Actually, with KY and IN calls for Trump, Biden's win prob is now down to *shudder* the exact same number
FL insta-dumps like 10% of its projected votes. Nice. This will be a quick night if Biden leads FL
The Pinellas county dump for Biden is huge, but I'm just not sure what to make of the election day results in the rest of the state. The Miami-Dade result in particular
Plugging in my the current NYT forecast in FL, GA, and NC into our election night model still gives a 92% chance of a Biden victory. The range of outcomes isn't so different from expectations yet.
also, my head hurts. just thought y'all should know.
I have to say, the needle is less anxiety-inducing without the needle. That's nice.
But the narrowing range out outcomes in FL should be a ~bit~ concerning to Democrats. Trump's strength there is pushing Trump's odds up toward the 10-12% range (it started at 7).
But the narrowing range out outcomes in FL should be a ~bit~ concerning to Democrats. Trump's strength there is pushing Trump's odds up toward the 10-12% range (it started at 7).
er, *without the jitter.
So. Different stories are emerging across FL right now. Sumter county is at a +4 point swing to Biden, which normally would be enough for a Dem, but Trump's apparent strength in Miami-Dade is upsetting a lot of the traditional patterns in the state.
Current model snapshot using the NYT uncertainty intervals in FL, NC and GA: 
Florida is a weird state. We shouldn't write off the inference from between-state correlations, but in fully-reporting counties from the other states, I'm seeing a 3 point average shift toward Biden. Seems important to note re: the Midwest.
It's still early, but what we're seeing so far is not the type of electoral map where Biden is doing 7 or 8 points better than Clinton.
With the NC and GA numbers roughly in line with expectations, our live election model is still relatively Biden-leaning at 85%. But without FL he loses a big cushion of electoral votes.
The latest update. The news in FL is devastating for Democrats, but Biden's overall win probability is not in a doomsday position yet.
The results outside of FL look pretty uniformly better for Biden, but I'm not sure we're really even in Biden +8 territory there either. It's going to be a while.
OK so what we're looking at right now is probably a 5 or 6 point Biden swing outside of Florida, with Biden performing pretty close to expectations outside of the southwest.
+ If we see an underperformance in FL + GA + NC but overperformance in TX, that's a pretty fun poll error
+ If we see an underperformance in FL + GA + NC but overperformance in TX, that's a pretty fun poll error
(southeast, sorry)
Apols for the intermittent updates folks, I'm scraping a bunch of county data by hand.
The NYT v FOX needle wars are terribly confusing when trying to update our range of outcomes
There has been a little bit of good news for the Biden campaign since the early low points in FL + NC + GA. Holding Trump a 5-point margin in OH and leading by 7pts in AZ (will go down) is much closer to expectations. So far, little evidence that the southeast errors are national
I'm just eye-balling but this looks close to 3-4 point national swing toward Biden on margin, which is certainly a Biden over performance, but probably still enough in the northern battlegrounds
https://twitter.com/kabir_here/status/1323810668928245761
Perhaps this would be a good time to say that Biden was up by just 4 points in our pure fundamentals model, and closer to +6 if you reverted to the polls a bit. That looks a lot closer to where we'll end up
So, for all that talk about fundamentals being overhyped in 2020....
Let me just say that I think people are right to be pumping the breaks a bit now. FL/GA was a bad result for Dems and even a bit surprising, but the swing toward Biden vs Clinton's margin is about 3 points on avg, and that would be enough for Biden if it holds in PA/MI/WI.
Oh, and the call just comes in in AZ. Let me re-run the numbers...
Biden winning AZ is nothing short of a massive boost to his odds. Here is our latest live forecast:
Here's where our model thinks we are now, though we're in the tails in some states so things are a bit unpredictable
(sorry forgot HI)
The histogram does
https://twitter.com/rianwatt/status/1323851919975067650
I hear that Fox is debating whether they should retract their AZ call. I wonder what we think the upper bound for Trump there is? +1? Couldn't be much higher if they called it in the first place.
NBC showing Biden up 11 in NE-02, which will be a good extra boost for Biden. But Trump is up in NE-02 which might just cancel it out in the end
Actually, there aren't two NE-02s in the country, though that would be cool! Up in NE-02, down in ME-02, as the map shows
With IA in Trump's column now the possible EV permutations are getting real thin.
Biden says “We believe we’re on track to win this election," which matches my expectations based on the results so far, with the very important caveat that we need more from PA first
Biden v Clinton vote share in MI, WI and PA for counties with >98% of expected votes cast. Looks like it's going to be a very tight finish...
these are crude estimates, but it's where we're at
Also, people need to take a deep breath with the POLLS ARE WRONG FOREVER stuff. The most likely outcome is still a Biden victory, and 290 EVs was always well within the uncertainty interval for our forecast, AKA a slightly higher than avg polling error
https://twitter.com/futurosimio/status/1323895510642483200
I am off to bed, but before I do, two tweets.
First, here is a model that predicts how fully-reported counties swung vs 2016 with a variety of other variables. The story here, as I told a colleague, is further polarization pretty much along every dimension except ethnicity.
First, here is a model that predicts how fully-reported counties swung vs 2016 with a variety of other variables. The story here, as I told a colleague, is further polarization pretty much along every dimension except ethnicity.
Second, here is where our live model stands. A Biden victory is still the likeliest outcome of the race, thanks to projected strength in WI and MI and the potential for a lucky, marginal win in GA or NC. But the tables could still turn in the morning. We'll wait for a full count.
as a bonus, here are final forecast brier scores so far. goodnight
No idea when MI + WI + PA will finish counting their ballots, but based on county-level returns and some interviews with county election chiefs I'm still relatively confident that there are enough mail ballots and votes in cities left that Biden will just barely pull off a win.
Crude chart, but, here is the correlation of polling error in 2016 v 2020, showing error in predicted two-party Democratic vote share (NOTE: not final, includes some predicted returns in partially-reporting counties)
Honestly, this feels terribly familiar
Yeah, and that's because there were always better explanations for the error last time than the edu weighting
https://twitter.com/Burghoff/status/1323987542505623558?s=20
People are really hammering us for being on the wrong side in FL, but so far every other state looks to be "called" correctly. More importantly, we gave the 290 and 306 Biden EV scenarios (the likeliest that remain) a 1-in-5 and 1-in-4 chance.
https://twitter.com/benjaminwittes/status/1323993860733915138?s=20
Now, there will be a real reckoning with polling error over the next few days. There are some big misses, But I think models broadly performed as intended — as tools to quantify what we could see happen under a 1-2 sigma polling error in Trump's direction.
I for one am certainly happier that the narrative was "Biden is up in the polls, but here's what could happen" instead of just "Biden is up in the polls.
Are we expecting a WI call soon?
With Wisconsin in Biden's column (according to the state gov), our model also sees a very likely (95%) Michigan victory which would cut off any path for Trump to 270 votes — and that's before we figure out what happens in GA, NC and PA. If all ballots get counted... it's over.
There is some confusion over whether AZ/NV are over (I'm with FOX on the AZ call), so FWIW here are the same numbers if you leave them as uncalled
Just based on the relationships to totals in other states, yea. But if I give the model some rough AZ bounds (including Trump +1) it changes to this.
But again, FOX and AP have called AZ, and that seems right to me.
But again, FOX and AP have called AZ, and that seems right to me.
https://twitter.com/TGYEM2001/status/1324043916291854338?s=20
CNN has called Michigan for Biden, which all but seals deal for him. Even if I leave NV and AZ off our call sheet, our live election model is still betting on a Biden victory with about 88% likelihood. (Overall I'm very happy with how it has performed adapting to returns.)
The model with AZ called (still controversial apparently):
The model without AZ called but with relatively pro-Biden constraints on vote share:
The model with AZ and NV called for Biden (where I think we're going to end up after NV releases more data tomorrow) is showing 290 as the most likely outcome, but with a 40% chance of also picking up GA and its 16 electoral votes.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
