I Don't Know If Joe Can Do It

[ondma]

You like to play both sides with polling. Republicans losing are winning but democrats leading are losing. The only poll showing it close in that race is SurveyUSA which is a high quality poll but the aggregate of polling still shows Smith by +5. Better than .5 for Ernst, or nothing for Graham, or -2 for Tillis.

I didnt say Lewis was going to win. But it certainly is discouraging that someone so radical could even be in striking distance of taking a US senate seat from a, IMO, very good sitting senator. Would you not agree? Like it or not, the Democrats have certainly lost the support of the rural areas and the male white middle class worker, even in a traditionally very strong democratic state like Minnesota.

Edit: as for playing both sides of the polls--- 2016 would like to speak to you.

 

[fskimospy]

I didnt say Lewis was going to win. But it certainly is discouraging that someone so radical could even be in striking distance of taking a US senate seat from a, IMO, very good sitting senator. Would you not agree? Like it or not, the Democrats have certainly lost the support of the rural areas and the male white middle class worker, even in a traditionally very strong democratic state like Minnesota.

Edit: as for playing both sides of the polls--- 2016 would like to speak to you.

Didn’t you say the polls didn’t count?

 

[Ken g6]

Trafalgar assumes there are "shy" Trump voters - voters who will vote for Trump, but are afraid to discuss it.

On 60 Minutes tonight they talked with "shy" Biden voters in Arizona, who will vote for Biden but are afraid to discuss it among their (they assume) Trump-loving neighbors.

So the polling errors this year could be the other way around.

 

[Dave_5k]

Sorry but this is not accurate. 4/5 points is within the margin of error for any single poll, yes, but 538 and other sites use polling averages, for which the margin of error is significantly lower as they combine large number of polls.

While it’s probably true that if you take the extreme edges of the residual margin of error they might overlap what the 538 estimate mostly represents at this point is the idea that the polls are wrong in a systemic way.

I should have been more clear, the primary reason I think margin of error is understated is indeed systemic bias, as demonstrated in multiple historical results. Some potential sources: no polls are doing the work to actually take a random sample out of all registered voters in a state, and then do the legwork to contact all of those randomly selected individuals. And almost no polls report on nor include in their statistics a major uncertainty factor - the non-response rate. And additionally have major bias risk in that 30-40% of registered voters don't actually vote, so can have a systemic bias in who choses to vote, which pollsters try to model with "likely" voters sub-selected by various methods, but is another huge driver of uncertainty and potential bias that pollsters do very poor job of signaling or quantifying.

So IF the polls were actually contacting a random sample of registered voters, and everyone selected responded, the 3-5% claimed error range on single polls, and a potential reduced 2-3 point error range with a large set of similar polls would be factually accurate - for registered voters. As it is, the error band is unknown, particularly without disclosure of the non-response rates and without actual random sampling of actual voters. But we can look back at historical errors in state races to get an idea of the minimum uncertainty ranges vs. polls, assuming no new sources of systemic bias.

Uncertainty at state level has been demonstrated to be far higher than reported: e.g. 2016 Pennsylvania was forecast for Clinton by ~4% on average, some polls as high as +7%, with the Trump win outside of the reported error bands on majority of polls. Similar situation in Michigan. And Wisconsin (Clinton +5 polls vs. Trump +1 actual). And in California in the opposite direction, with Clinton actually winning by 30% vs. forecast 23% average poll margin. Contrast these results with +/- "3.5%" uncertainty claimed for individual higher quality state polls.

Not to say this uncertainty/bias is unlimited - if polls are getting 90%+ response rate, and are truly contacting random sample of voters (which most polls are doing their best at), aggregate state polls are likely to be accurate within 6-8% for heavily polled states. But this demonstrated historical uncertainty of +/- 6-8% on aggregate polling of state races is far larger than the pollster claims of uncertainty bands at the state level, which is ultimately what matters for the presidential election.

 

[fskimospy]

I should have been more clear, the primary reason I think margin of error is understated is indeed systemic bias, as demonstrated in multiple historical results. Some potential sources: no polls are doing the work to actually take a random sample out of all registered voters in a state, and then do the legwork to contact all of those randomly selected individuals. And almost no polls report on nor include in their statistics a major uncertainty factor - the non-response rate. And additionally have major bias risk in that 30-40% of registered voters don't actually vote, so can have a systemic bias in who choses to vote, which pollsters try to model with "likely" voters sub-selected by various methods, but is another huge driver of uncertainty and potential bias that pollsters do very poor job of signaling or quantifying.

So IF the polls were actually contacting a random sample of registered voters, and everyone selected responded, the 3-5% claimed error range on single polls, and a potential reduced 2-3 point error range with a large set of similar polls would be factually accurate - for registered voters. As it is, the error band is unknown, particularly without disclosure of the non-response rates and without actual random sampling of actual voters. But we can look back at historical errors in state races to get an idea of the minimum uncertainty ranges vs. polls, assuming no new sources of systemic bias.

Uncertainty at state level has been demonstrated to be far higher than reported: e.g. 2016 Pennsylvania was forecast for Clinton by ~4% on average, some polls as high as +7%, with the Trump win outside of the reported error bands on majority of polls. Similar situation in Michigan. And Wisconsin (Clinton +5 polls vs. Trump +1 actual). And in California in the opposite direction, with Clinton actually winning by 30% vs. forecast 23% average poll margin. Contrast these results with +/- "3.5%" uncertainty claimed for individual higher quality state polls.

Not to say this uncertainty/bias is unlimited - if polls are getting 90%+ response rate, and are truly contacting random sample of voters (which most polls are doing their best at), aggregate state polls are likely to be accurate within 6-8% for heavily polled states. But this demonstrated historical uncertainty of +/- 6-8% on aggregate polling of state races is far larger than the pollster claims of uncertainty bands at the state level, which is ultimately what matters for the presidential election.

What model are you basing this on and what’s your basis of expertise?

To be clear I work with statistics professionally and very little of what you’re saying here makes sense. (For example non-response rate is heavily accounted for in polling)

 

[fskimospy]

If the idea is that state polling has higher uncertainty than national polling I agree because it’s harder to correctly model the electorate but the uncertainty is not that much higher and would still need to be outside of normal errors.

 

[Dave_5k]

What model are you basing this on and what’s your basis of expertise?

To be clear I work with statistics professionally and very little of what you’re saying here makes sense.

For simplicity, was looking at 538 aggregate polls in 2016 vs. actual results for my examples on demonstrated uncertainties for a handful of selected states: https://projects.fivethirtyeight.com/2016-election-forecast/

I'll take demonstrated actual historical error bands over theoretical statistical uncertainty bands (especially when the former is larger!), although my hypothesis on why there may have been such large errors is admittedly purely speculative. 20+ years removed from classwork in statistics, although currently work professionally on risks on energy investment projects including weather uncertainty, regulatory uncertainty, and future price uncertainty among other interesting risks in investment decisions (for wind and solar).

Walking through some potential sources of selection bias on even a "perfect" poll:
Take a hypothetically perfect poll that is a completely unbiased, completely random sample, with large random sample of all registered voters. Even so, this "perfect" sample poll will end up having multiple potential systemic biases:
Selection bias 1: This was from snapshot of registered voters as of a point in time, may be missing recent voter registration
Selection bias 2: This is limited to registered voters that they were able to get a phone number for (most voter registration lists do not typically include phone number - and won't be able to be contacted at all if unlisted phone or no phone)
Selection bias 3: The survey results are limited to subset of people that choose to answer phone call from unknown caller
Selection bias 4: And then further limited to subset that agree to answer poll
Selection bias 5: What time of day do they try to reach the registered voters - e.g. only calling in evening? (potentially under-sampling those that work night shifts)

Recognizing that polls can't obtain a random sample of actual voters is a key issue increasing uncertainty, and all phone polling faces the same hurdles and potential biases. Not that any of those individual particular items may in fact be biased to one party or the other, but they could be and we don't know. Are democrats or republicans more likely to have unlisted phone number or no phone? Are democrats or republicans more likely to refuse to answer phone from unknown caller or refuse to take time to answer poll?

 

[fskimospy]

For simplicity, was looking at 538 aggregate polls in 2016 vs. actual results for my examples on demonstrated uncertainties for a handful of selected states: https://projects.fivethirtyeight.com/2016-election-forecast/

I'll take demonstrated actual historical error bands over theoretical statistical uncertainty bands (especially when the former is larger!), although my hypothesis on why there may have been such large errors is admittedly purely speculative. 20+ years removed from classwork in statistics, although currently work professionally on risks on energy investment projects including weather uncertainty, regulatory uncertainty, and future price uncertainty among other interesting risks in investment decisions (for wind and solar).

Walking through some potential sources of selection bias on even a "perfect" poll:
Take a hypothetically perfect poll that is a completely unbiased, completely random sample, with large random sample of all registered voters. Even so, this "perfect" sample poll will end up having multiple potential systemic biases:
Selection bias 1: This was from snapshot of registered voters as of a point in time, may be missing recent voter registration
Selection bias 2: This is limited to registered voters that they were able to get a phone number for (most voter registration lists do not typically include phone number - and won't be able to be contacted at all if unlisted phone or no phone)
Selection bias 3: The survey results are limited to subset of people that choose to answer phone call from unknown caller
Selection bias 4: And then further limited to subset that agree to answer poll
Selection bias 5: What time of day do they try to reach the registered voters - e.g. only calling in evening? (potentially under-sampling those that work night shifts)

Recognizing that polls can't obtain a random sample of actual voters is a key issue increasing uncertainty, and all phone polling faces the same hurdles and potential biases. Not that any of those individual particular items may in fact be biased to one party or the other, but they could be and we don't know. Are democrats or republicans more likely to have unlisted phone number or no phone? Are democrats or republicans more likely to refuse to answer phone from unknown caller or refuse to take time to answer poll?

First question - why are you taking one year as opposed to looking at average polling error over time? As for your concerns about selection bias an enormous amount of work is put into correcting for this.

My point is that while it is certainly not impossible that polling errors are larger than Biden’s lead it is also well outside the margin of error as people commonly understand it.

 

[ondma]

Didn’t you say the polls didn’t count?

You accuse me of playing both sides of the polls in one post and then claiming I said they dont matter in another. Which is it? Or are you just throwing as many insults as you can and hoping one will stick?

Edit: I dont recall ever saying that polls "dont matter." If I did it was in context of a relative margin of error or in a particular situation. If every state had Biden up by 20%, then obviously an electoral college victory would almost be certain. (Basically I dont consider anything 100% probable, there is always some uncertainty). However, especially in a lot of swing states, I still consider Biden's lead problematical and there can be enough last minute shifts to swing enough states to give Trump an EC victory. Again, 2016 is calling.

 

[fskimospy]

You accuse me of playing both sides of the polls in one post and them claiming I said they dont matter in another. Which is it? Or are you just throwing as many insults as you can and hoping one will stick?

I’m saying either the polls count or they don’t.

If they do, that means taking empirical polling averages from competent statisticians. No cherry picking, no nonsense.

If they don’t then it’s a free for all and people believe what they want.

Pick one.

 

[ondma]

Nice oversimplification. Polls obviously "matter". Would I rather see Biden up by 5% than Trump up by 5%? Certainly. I just dont give the single digit leads at best that Biden has in swing states as much confidence as you do. And yes, it may not be statistically accurate, but I give Biden's lead less confidence than if Trump had a similar lead for 2 reasons. First is the undeniable rabid enthusiasm Trump supporters have and second is that we have seen a last minute surge toward Trump already in 2016, and I dont rule out it happening again.

 

[ecogen]

we have seen a last minute surge toward Trump already in 2016, and I dont rule out it happening again.

You don't rule out Comey coming out with a letter about reopening the investigation into Biden's emails?

 

[K1052]

You don't rule out Comey coming out with a letter about reopening the investigation into Biden's emails?

While Trump has tried painstakingly to recreate the exact conditions of October 2016 reality has not been very accommodating.

 

[Roger Wilco]

538 just upgraded Biden's odds to 90/9.

 

[K1052]

Trafalgar assumes there are "shy" Trump voters - voters who will vote for Trump, but are afraid to discuss it.

On 60 Minutes tonight they talked with "shy" Biden voters in Arizona, who will vote for Biden but are afraid to discuss it among their (they assume) Trump-loving neighbors.

So the polling errors this year could be the other way around.

Trafalgar's lead pollster went on Fox and claimed PA has systematic voter fraud that will cost Trump 4 to 5 points.

https://twitter.com/x/status/1323087162938720256

I know people love to point at Trafalgar as getting it right in 2016 to justify faith in their numbers but when things went down people also pointed at the USC polling which had Trump ahead in 2016 but shows something a little different this year:

 

[fskimospy]

Trafalgar's lead pollster went on Fox and claimed PA has systematic voter fraud that will cost Trump 4 to 5 points.

https://twitter.com/x/status/1323087162938720256

I know people love to point at Trafalgar as getting it right in 2016 to justify faith in their numbers but when things went down people also pointed at the USC polling which had Trump ahead in 2016 but shows something a little different this year:

View attachment 32773

USC was wrong in 2016, the fact that they keep trying to claim otherwise is embarrassing.

 

[K1052]

USC was wrong in 2016, the fact that they keep trying to claim otherwise is embarrassing.

Yes they were wrong but in a way that made some people think they were right.

Looking out at the polling landscape perhaps the thing that interests me most is how dramatically Trump is underperforming in states he won handily in 2016. Morning Consult released their last slate of state level surveys and I had to blink a bit at Trump being +9 in MO and +11 in IN. Even an R partisan pollster in MO has been pegging him at +5 or +6 for a while. I mean all this could be wrong but if it's not it would seem to signal a major collapse with white voters in Biden's favor.

 

[s0me0nesmind1]

Probably the most truthful thing I've seen about these last elections:

 

[fskimospy]

Probably the most truthful thing I've seen about these last elections:

View attachment 32777

This is a great quote because it perfectly encapsulates libertarianism - sounds good until you put literally any thought into it.

 

[K1052]

Guess which party sued to get the Libertarians off the ballot in Texas this year.

 

[s0me0nesmind1]

Guess which party sued to get the Libertarians off the ballot in Texas this year.

what does that have to do with anything said?

Guess who SUCCESSFULLY got the green party off the ballot in multiple states?

 

[K1052]

what does that have to do with anything said?

Guess who SUCCESSFULLY got the green party off the ballot in multiple states?

AFAIK the Libertarians nowhere fucked up their paperwork as badly as the Greens have resulting in them not appearing on several state ballots.

 

[K1052]

To clarify my position on 3rd parties: I want more of them but only in conjunction with RCV and jettisoning the EC.

 

[VRAMdemon]

Now ... Trump’s messages about the vaccine. He says it would have been delivered by now except that "The Deep State" decided to delay the vaccine in order to sabotage his re-election. It’s important to keep in mind it’s Trump saying this.

Trump is now in the fourth year of his presidency. If he was going to take charge of the government, he would have done it by now. So if "The Deep State" is still around and able to attack Trump, isn’t that an acknowledgement that Trump can’t control "The Deep State"? If Trump couldn’t stop "The Deep State" in his first term, why should we expect anything to change in his second term? Re-electing Trump means we get another four years of what we’ve just had? Regardless of what Trump’s plans are, "The Deep State" will stop him. Trump ran for office claiming that he could outthink, outdeal, and outmanuever any opposition that was thrown at him. That was the center of his campaign: he was somebody who could get things done. Nobody could stop him.

Trump didn’t present himself as a person who would just fight the good fight. He promised to win. He said that he would be able to get results that other men couldn’t. He would outfox the Chinese and the Russians. He would get our allies to pay all the money they owed us. He would dictate treaties to other countries and compel them to sign them. He would solve all the problems in the Middle East. He would build a wall and make Mexico pay for it. He would score the game winning touchdown and have sex with the head cheerleader and her hot sister. Trump told his supporters that he was a winner.

I'm wondering how his cult followers reconcile Trump’s claims that he can defeat any enemy with his admission that The Deep State is still around and still thwarting him. Do they feel that Trump could destroy Deep State is he wanted to? If so, why hasn’t he? Or do they acknowledge that Trump can’t defeat Deep State and by doing so, admit there are some problems that Trump can’t fix?

 

[vi edit]

The only deep state is systemic republican corruption attempting to invalidate legal votes of tens of thousands Americans.