2020 Presidential Election Results & Exit Polls commentary thread

[Annatar]

We are up past 149.3 million votes on the cook popular vote tracker and Biden has not even reached 3% in the popular vote, Nate Silver tweeted out recently maybe he could win by 5 or even 6 points with late ballots, I doubt it, in California Biden's margin has gone from +33 to +31.3 as more votes have come in, New York should give a big boost to Biden but since election day Biden has only been winning votes by 16.5%, if we assume turnout of 160 million, a 16.5% advantage among the remaining vote would get Biden to a 3.9% margin in the popular vote.

[Annatar]

The Edison exit poll that CNN uses is not reliable, its numbers in the states do not match up with the actual results, if you take the Florida exit poll and look at the white and non-white vote result and share, Trump should only be up 1%, he is going to win by 3.4%, in Arizona the vote by race suggests a 2.8% Biden win but he is only going to win by 0.5% or less. In Iowa the exit poll suggests a 6% win, not a 8% one.

[Annatar]

Regarding changing demographics I would make a point that for whatever reason, the share of white evangelicals is not falling in exit polls despite the falling share of the white vote overall, maybe the exit polls are oversampling this group. Below is the share of white evangelicals per year in the exit polls on the left and the overall white share on the right.

2004: 23/77
2008: 26/74
2012: 26/72
2016: 26/71
2020: 28/67

[Annatar]

Biden's margin in CA has dropped below 31%, it is at 30.7%, looks like 2020 will be the year CA trends Republican.

[Annatar]

While Trump's EC advantage was still large, 2020 was definitely an indication that it's not something Republicans can count on for much longer.

The tipping point state this year will be even more Republican compared to 2016 and the 2 big states that are gaining EV's post 2020, Florida and Texas are not trending democrat very strongly, Florida not at all, if we have a close election in 2024 there could be massive split between the electoral college and popular vote.

[Annatar]

The swing to Biden in New Jersey is down to 1.8% now with more ballots counted, quite possible New Jersey will trend Republican this year again like it did in 2016, here is an interesting statistic, Biden is doing only 2.4% worse in Colorado than New Jersey, could Colorado vote to the left of New Jersey in 2024. 

[Annatar]

The swing to Biden in New Jersey is down to 1.8% now with more ballots counted, quite possible New Jersey will trend Republican this year again like it did in 2016, here is an interesting statistic, Biden is doing only 2.4% worse in Colorado than New Jersey, could Colorado vote to the left of New Jersey in 2024. 

I'd honestly be shocked if Biden did worse than Hillary in NJ... I guess just an impact of higher turnout among rural Rs?

I'm talking about the trend, not the swing, Biden will do better than Hillary in NJ, I am sure of that but it will shift to the right compared to the national swing.

[Annatar]

I would note regarding Nevada that it is one of the more Hispanic states in the country and so any swing against Trump among whites in Nevada could have been offset by gains among Hispanics in a way that was not possible in whiter states.

[Annatar]

I have always thought Biden would win by 4% or a bit more, Clinton had a slight lead over Trump the morning after election day and she won by 2.1%, so she gained roughly 2% over the vote county post election day, Biden had a 2% lead the morning after the election, so if he gains like Clinton he should win by 4%.

[Annatar]

Comparing the pre-election polls to the exit polls as well as the results, 2 things stand out in terms of what the pre-election polls, both at the state level and national level appear to have got wrong.

1. Consistently in state and national polls, the swing towards Trump among Hispanics was picked up as was his marginal gains with blacks and losses among college whites vs 2016, what the polls failed to capture was that there would be no large swing of non-college whites towards Biden, the only explanation is the polls were reaching a disproportionately anti Trump group of non-college whites.

2. The polls accurately captured how democrats would vote as well as how independents would swing to voting for Biden this year compared for Trump in 2016 what the polls missed was how many Republican defections there would be, the NYT/Siena poll in NV for example only had Trump winning Republicans by an 82% margin. For whatever reason the polls were reaching a particularly anti-Trump group of republicans.


Basically it seems the pre-election polls got two things wrong, they were sampling a disproportionately democratic leaning group of non college whites and they were sampling a disproportionately anti Trump group of Republicans, the two things can of course be related.

I guess it was something like the NYT calling random towns in PA and the non-college whites who picked up the phone were librarians who were democrats and that threw off their whole non-college white sample. Would not surprise me if pollsters oversampled the types of non-college whites who vote more democrat like librarians, teachers etc and under sampled those who vote Republican. 

[Annatar]

Comparing the pre-election polls to the exit polls as well as the results, 2 things stand out in terms of what the pre-election polls, both at the state level and national level appear to have got wrong.

1. Consistently in state and national polls, the swing towards Trump among Hispanics was picked up as was his marginal gains with blacks and losses among college whites vs 2016, what the polls failed to capture was that there would be no large swing of non-college whites towards Biden, the only explanation is the polls were reaching a disproportionately anti Trump group of non-college whites.

2. The polls accurately captured how democrats would vote as well as how independents would swing to voting for Biden this year compared for Trump in 2016 what the polls missed was how many Republican defections there would be, the NYT/Siena poll in NV for example only had Trump winning Republicans by an 82% margin. For whatever reason the polls were reaching a particularly anti-Trump group of republicans.


Basically it seems the pre-election polls got two things wrong, they were sampling a disproportionately democratic leaning group of non college whites and they were sampling a disproportionately anti Trump group of Republicans, the two things can of course be related.

I guess it was something like the NYT calling random towns in PA and the non-college whites who picked up the phone were librarians who were democrats and that threw off their whole non-college white sample. Would not surprise me if pollsters oversampled the types of non-college whites who vote more democrat like librarians, teachers etc and under sampled those who vote Republican. 

Interesting theory! I have a couple of questions.

Does that track state by state? The polls were off more in WI, TX, and others than in GA or AZ. Is that because there are fewer of the subgroups you identified in the latter states?

I'm curious, too, about how this theory about sampling error fares when looking at down ballot races.

Regarding errors by region, if you take the final 538 polling average across the 7 northern swing states, MN, WI, MI, PA, NH, IA and OH, I assume OH will end up at a 7% margin for Trump and PA at a 1% margin for Biden, the polls overestimated Biden by 5.1% on average, these are the states with a lot of non-college whites.

In the 6 southern/western swing states, NC, GA, TX, FL, AZ and NV, polls overestimated Biden by an average of 3.3%, so the polling error was roughly 2% smaller in the states with fewer non-college whites but it was still fairly large.

One thing that was present in 2016 that disappeared this year was in some swing states Clinton outperformed her polls like Nevada, this year, in all 13 swing states Trump outperformed his polls, in that sense 2020 represented a systematic failure in polling unlike 2016 where polls erred in some swing states in the other direction.

[Annatar]

One thing I am curious about is all these big surveys like the CCES, nationscape etc showed all these obama trump voters unhappy with Trump and liking Biden, mainly non-college whites in the midwest yet there was no swing, in the driftless region there was no swing, Trump did better in many counties in north-eastern Iowa and south-western WI than the did in 2016.

So were all these surveys trash, 2 years of survey data about non-college whites who had voted Obama than Trump  but were disillusioned with Trump going back to Biden, all this data just seems fake now, none of it materialised.

[Annatar]

One thing I am curious about is all these big surveys like the CCES, nationscape etc showed all these obama trump voters unhappy with Trump and liking Biden, mainly non-college whites in the midwest yet there was no swing, in the driftless region there was no swing, Trump did better in many counties in north-eastern Iowa and south-western WI than the did in 2016.

So were all these surveys trash, 2 years of survey data about non-college whites who had voted Obama than Trump  but were disillusioned with Trump going back to Biden, all this data just seems fake now, none of it materialised.

There are some specific parts of the rural/small-town Midwest that swung R, but based on what we're seeing right now, most of Middle America did swing to Biden (and a good part of it even trended D). Of course, it's not the 6-8 points polls were showing, and that will have to be seriously examined, but let's not act like the swing didn't even exist.

The swing even in Middle America appears to have been college whites though, even small town counties in Missouri have college whites that could swing to Biden.

[Annatar]

Looks like Trump may have marginally increased his share of the black vote in LA, from 7% to 10% in 70% plus black precincts. Across multiple states the evidence appears to point to a 2 - 4% rise in Trump's black vote share vs 2016, instead of 8% in 2016, Trump probably got 11% of the black vote this year.

https://twitter.com/WinWithJMC/status/1328680296586354688

[Annatar]

The final results in Ohio will shift the state a point towards dems but right now Trump is up 8% in Ohio and the House GOP won the vote by 14% in Ohio with every seat being contested, that is a pretty big difference of 6%, seems a lot of voters in Ohio liked Biden but also voted Republican.

[Annatar]

Does anyone know why Richland county IL swung so much in 2016-2020, is there some specific thing going on in that county.

[Annatar]

I have been looking through both the Edison and AP exit polls, both have issues but I am of the opinion the AP exit poll is more accurate, the margins of different groups make more sense.

[Annatar]

2 things to note in the 2018 AP exits vs 2020. College white voters and high income voters voted like they did in 2018, Trump did better with lower income voters, non-college whites and non-college non whites vs the GOP in 2018. Below are the margins:

Non college Whites:
2018: R+20
2020: R+25

College White:
2018: -4R
2020: -6R

Non-college non-white:
2018: -57R
2020: -48R

College non-white:
2018: -49R
2020: -48R

Incomes:
2018:  <$50,000: -17R
2020: <$50,000: -8R

2018: $50-$99,000: -2R
2020: $50-$99,000: +2R

2018: >$100,000: -5R
2020: >$100,000: -4R

Basically, Trump improved by 9% among voters under $50,000 vs 2018, improved 4% among voters earning $50,000 - $99,000 and gained 1% among voters making over $100,000.

The conclusion is college whites making over $100,000 were likely the only group that became more democratic between 2018-2020.

[Annatar]

Been comparing the last NYT polls in AZ, PA, WI & FL to the AP exit poll to see where the differences lie, biggest difference appears to be among older voters, pollsters seem to have been reaching a disproportionately democratic group of 65+ voters across multiple states.

NYT WI poll: 65+: Biden +15
AP WI Exit poll: 65+: Trump +7

NYT PA Poll: 65+: Biden+12
AP PA Exit Poll: 65+: Trump+6

NYT FL Poll: 65+: +2 Biden
AP FL Exit Poll: 65+: +11 Trump

NYT AZ Poll: 65+: +1 Trump
AP AZ Exit Poll: 65+: +7 Trump



Here are the final NYT margins vs the results

WI: Biden +11 : Biden +0.6
PA: Biden +6   : Biden +1.2
FL: Biden +3   : Trump +3.4
AZ: Biden +6  : Biden +0.3

Average error: 6.8% treating each state independently. 

[Annatar]

Looks like NC will end its 4 election streak of trending democratic, 2004, 2008, 2012 and 2016 it trended democratic, this election it will trend Republican for the first time since 2000.

[Annatar]

Interesting thing to note in the Fox exit polls, there is a 10% difference in margin between non-college white and and women R+31 and R+21 but a massive 26% gap between college white men and college white women, R+6 and D+20, it seems college moves women much more towards the democrats than it does men.

[Annatar]

Interesting thing to note in the Fox exit polls, there is a 10% difference in margin between non-college white and and women R+31 and R+21 but a massive 26% gap between college white men and college white women, R+6 and D+20, it seems college moves women much more towards the democrats than it does men.

as i can read the numbers the fox polls overestimated Trump % and abit underestimated the Biden %, so, the polls numbers need some works

The age crosstabs suggest a Biden win of 3.9%, he will win by 4.4% so it is only off 0.5%.

[Annatar]

It is interesting that now that only 2 states have trended Republican in every election since 2004, Arkansas and Pennsylvania, states like West Virginia, Tennessee etc all trended democratic, but Arkansas and Pennsylvania remain the only 2 states that have trended Republican 4 times in a row.

[Annatar]

The AP and Navigator white vote share is off by a lot as previous posters have said, whites were estimated to comprise 67% of eligible voters this year, whites turnout out at slightly above the average turnout as Hispanics and Asians turn out at lower rates, whites were unlikely more than 70% of all voters, the AP at 74% white and Navigator at 75% white is far to high.