Random Experiment: 2020 polling averages adjusted by 2016 polling errors

[DabbingSanta]

Map:

278-244 Biden win with Michigan uncalled.

Data:


My Thoughts:

1) Southern states could really be at play this year.  Statewide polling in GA & TX were near perfect in 2016 and 2018. Let's see if the trend continues.

2) Midwestern states are still only a polling error away for Trump, including MI/MN/PA/WI.

3) Note: This is assuming the election is held today. Polling aggregates will change between now and November and the race could look much different heading in to election day.

4) My point is that everything is still up in the air. Trump could still win on election day. Likewise, Biden could sweep all the states on my list. Judging by some of the comments on this board I think we have already forgotten polls should not be taken for gospel.

[ProgressiveModerate]

Interesting. We'll have to see if have the same silent vote, particuarly that in rural areas that he got in 2016. I doubt there's as much closet support for Trump know that he's such a known entity.

[SevenEleven]

I'm not a fan of this style of analysis because these polling margins do take into consider undecided voters and therefore you create a systemic bias.

It would be much better to compare Trump polling average to Trump results and Clinton polling average to Clinton results.

1) Southern states could really be at play this year.  Statewide polling in GA & TX were near perfect in 2016 and 2018. Let's see if the trend continues.

These states should be easier to poll given how monolithic various voting demographics are, any good likely voter screen should do well. It probably shouldn't surprise anyone that the least accurate polls are in states with the most (brain damaged?) Obama-Trump voters.

[TrendsareUsuallyReal]

This is bullsh**t, Trump and other Republicans performed about 3% worse than their polling averages in Texas and Arizona

[DabbingSanta]

I'm not a fan of this style of analysis because these polling margins do take into consider undecided voters and therefore you create a systemic bias.

It would be much better to compare Trump polling average to Trump results and Clinton polling average to Clinton results.

Fair enough, I'll do some more research. Also worth noting that some of these poll aggregate numbers do not include polling with third party candidates, which may impact things as well

This is bullsh**t, Trump and other Republicans performed about 3% worse than their polling averages in Texas and Arizona

source:

https://www.270towin.com/2016-polls-clinton-trump/texas/
https://www.270towin.com/2016-polls-clinton-trump/arizona/

[Non Swing Voter]

Seems like the pollsters would have refocused on states where they got the margin wrong by a lot.  I think Pennsylvania and Michigan will probably be more accurate this time than e.g., Arizona.

[MT Treasurer]

The fact that even all this pro-R "unskewing" where we lazily assume that the same polling error occurs again AND to the same extent (which wasn’t the case in 2018) only yields a map where Biden is already at 278 EV is pretty devastating for Trump.

E: Is there a particular reason you didn’t "unskew" the Atlas polling averages in those battleground states, e.g. MI?

[Reaganfan]

The fact that even all this pro-R "unskewing" where we lazily assume that the same polling error occurs again AND to the same extent (which wasn’t the case in 2018) only yields a map where Biden is already at 278 EV is pretty devastating for Trump.

I don't think so. I think we might be at a crossroads for polling. In the same way the "New York Times" and major newspapers or late night TV like "The Tonight Show" are now essentially arms of the left, we may see polling join that group, especially if polls overestimate Biden like they did many Democrats since 2014.

[WD]

The fact that even all this pro-R "unskewing" where we lazily assume that the same polling error occurs again AND to the same extent (which wasn’t the case in 2018) only yields a map where Biden is already at 278 EV is pretty devastating for Trump.

I don't think so. I think we might be at a crossroads for polling. In the same way the "New York Times" and major newspapers or late night TV like "The Tonight Show" are now essentially arms of the left, we may see polling join that group, especially if polls overestimate Biden like they did many Democrats since 2014.

lol

[MT Treasurer]

The fact that even all this pro-R "unskewing" where we lazily assume that the same polling error occurs again AND to the same extent (which wasn’t the case in 2018) only yields a map where Biden is already at 278 EV is pretty devastating for Trump.

I don't think so. I think we might be at a crossroads for polling. In the same way the "New York Times" and major newspapers or late night TV like "The Tonight Show" are now essentially arms of the left, we may see polling join that group, especially if polls overestimate Biden like they did many Democrats since 2014.

I’m not saying that the race couldn’t tighten a little or that there aren’t polls which are understating R strength, but the big picture has been clear, unambiguous, and consistent since 2017. Biden's lead in national polls has been larger and more stable than Clinton's ever was (in case we forget, national polling wasn’t dramatically off in 2016), and he’s polling far better than her in several swing/R must-win states (especially MI, TX, GA, AZ; and in those last three states polling has not exactly underestimated R strength in recent elections).

It’s more than polling, though. From an electoral standpoint, his entire presidency has been a disaster for the GOP, and I’m not just talking about low-turnout special elections here. There’s no denying on my part that midterm elections have tended to result in some serious losses for the incumbent's party in recent history, but when you consider how much Republicans had going for them in 2018 (galvanized base and high turnout, booming economy, Kavanaugh aftermath, most lopsided/pro-R Senate map since 1938, no major war like in 2006, etc.), it’s astonishing how devastating the defeat was for the party. That election should have never been another 2006 or worse.

COVID-19 and the protests have made things worse for Trump, but it was obvious even before those developments that this election was going to be an uphill battle for him without a historically unpopular opponent to run against and his inability to broaden his base (all the while losing ground in vast parts of the country that were key to his victory in 2016, especially in suburban areas/among female voters/"lesser-of-two-evil" voters/etc.). Trump has always been an unusually weak position for an incumbent (it’s been this way since January 2017, honestly), but much of it makes sense if you interpret 2016 not as an embrace of Trump but as a rejection of Hillary Clinton. He’s as close to DOA as you possibly get under current conditions.

E: There’s nothing wrong with being cautious and refusing to blindly trust polling, but I’d definitely advise against trusting your "gut feeling" too much lest you set yourself up for disappointment. I know the feeling because my "gut" told me that 2014 and 2016 would be much better elections for the GOP than predicted by most election observers, but my "gut" has disappointed me just as often if not more during Trump's presidency.

[catographer]

Interesting hypothetical. I would’ve used RCP averages, but regardless this seems pretty informative. It’s basically a best-case scenario for Trump, the thing he is hoping happens again.

[tagimaucia]

Just think about what would happen if you had adjusted 2016 polling averages by 2012 polling errors and maybe you’ll realize what a silly exercise this is.

[DabbingSanta]

Just think about what would happen if you had adjusted 2016 polling averages by 2012 polling errors and maybe you’ll realize what a silly exercise this is.

Did you not read my disclaimer? Obviously the odds of this playing out exactly like 2016 are near zero. I'm just pointing out that Trump isn't necessarily screwed like everyone on here says, especially given we are five months out. Biden could sweep all the swing states in November, but he also could lose them. Worth noting Clinton had a huge lead in the summer, and at that point even I thought she would win a large victory.

[Figueira]

Interesting. We'll have to see if have the same silent vote, particuarly that in rural areas that he got in 2016. I doubt there's as much closet support for Trump know that he's such a known entity.

Is there any actual evidence for the "shy Trump supporter" concept? I thought the polling error in 2016 was mostly due to pollsters not factoring in education.

[Left Wing]

 If you use RCP, Biden flips more states

[John Dule]

It's an interesting thought experiment, but the polling will be much more accurate this year. MI, WI, and PA were all severely underpolled in 2016, and the smaller sample size was part of the reason why they missed so wildly.

[lfromnj]

It's an interesting thought experiment, but the polling will be much more accurate this year. MI, WI, and PA were all severely underpolled in 2016, and the smaller sample size was part of the reason why they missed so wildly.

Were they?
A lot of pollsters even in 2018 didnt fix the education problem.
However the good news for Wisconsin is it's likely we have gold standard Marquette back

[John Dule]

It's an interesting thought experiment, but the polling will be much more accurate this year. MI, WI, and PA were all severely underpolled in 2016, and the smaller sample size was part of the reason why they missed so wildly.

Were they?
A lot of pollsters even in 2018 didnt fix the education problem.
However the good news for Wisconsin is it's likely we have gold standard Marquette back

Leaving the methodology aside, there just weren't many polls run in those states that I saw on the RCP tracker back then. Not even enough to calculate an average with at some points. I don't know what they've done to improve the methodology, but I guarantee you that pollsters won't neglect to put their resources there this time.

[morgieb]

Be interesting to see the numbers by 2018 polling error. I think there's some evidence that they've been fixed, though I still expect a slight polling error in the Midwest which may boost the Reps.

[Flo 2.0]

Oh dude this looks so cool!! It's so interesting to see some of these results (such as Georgia voting more Democratic than Minnesota) but it brings up a very good point with the absolute uncertainty in this (or any) election!! Good job on this, my guy! I hope you can make at least one more before election day,! 😊😸

[Alben Barkley]

The fact that even all this pro-R "unskewing" where we lazily assume that the same polling error occurs again AND to the same extent (which wasn’t the case in 2018) only yields a map where Biden is already at 278 EV is pretty devastating for Trump.

I don't think so. I think we might be at a crossroads for polling. In the same way the "New York Times" and major newspapers or late night TV like "The Tonight Show" are now essentially arms of the left, we may see polling join that group, especially if polls overestimate Biden like they did many Democrats since 2014.

God I hope more of you people keep doing Donald Math and mental gymnastics like this to deny reality with your “alternative facts.”

Will just make the victory all the more sweet if you’re all shocked when it happens.

[politics_king]

Map:

278-244 Biden win with Michigan uncalled.

Data:


My Thoughts:

1) Southern states could really be at play this year.  Statewide polling in GA & TX were near perfect in 2016 and 2018. Let's see if the trend continues.

2) Midwestern states are still only a polling error away for Trump, including MI/MN/PA/WI.

3) Note: This is assuming the election is held today. Polling aggregates will change between now and November and the race could look much different heading in to election day.

4) My point is that everything is still up in the air. Trump could still win on election day. Likewise, Biden could sweep all the states on my list. Judging by some of the comments on this board I think we have already forgotten polls should not be taken for gospel.

If the 2016 map stays the same but Biden flips Arizona & Florida, it's over.