ISF Logo   IS Forum
Forum Index Register Members List Events Mark Forums Read Help

Go Back   International Skeptics Forum » General Topics » USA Politics
 


Welcome to the International Skeptics Forum, where we discuss skepticism, critical thinking, the paranormal and science in a friendly but lively way. You are currently viewing the forum as a guest, which means you are missing out on discussing matters that are of interest to you. Please consider registering so you can gain full use of the forum features and interact with other Members. Registration is simple, fast and free! Click here to register today.
View Poll Results: On the whole, do you believe political polls are trustworthy/reliable?
Yes. 6 27.27%
Deez Nutz. 16 72.73%
Voters: 22. You may not vote on this poll

Reply
Old 7th November 2020, 07:53 PM   #41
Segnosaur
Penultimate Amazing
 
Join Date: Jan 2002
Posts: 16,335
I took a look at the last national polls (from Nov3) on 538 and compared them to the election results (as we have now.... counting is still ongoing)...

Biden: Poll: 51.8, election: 50.6
Trump: Poll 43.3, election 47.7

So, the national polls weren't really that far off this time. Biden's results were ~1% off, and Trump's results were ~4% off. What is interesting is that there seem to be a lot more unaccounted for in the polling averages than in the final results. I wonder if this could be due to 3rd party voters that decided to switch to Trump at the end?
__________________
Trust me, I know what I'm doing. - Sledgehammer

I'm Mary Poppin's Y'all! - Yondu

We are Groot - Groot
Segnosaur is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 08:12 PM   #42
Venom
Illuminator
 
Join Date: Apr 2011
Location: United States
Posts: 4,722
Can polls ask more than 1000 people at a time?

Can we have a poll or set of polls over a month after asking 8,000-10,000 real people in a state?
Venom is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 08:45 PM   #43
angrysoba
Philosophile
 
angrysoba's Avatar
 
Join Date: Dec 2009
Location: Osaka, Japan
Posts: 28,361
Originally Posted by Venom View Post
Can polls ask more than 1000 people at a time?

Can we have a poll or set of polls over a month after asking 8,000-10,000 real people in a state?
Yeah, you can do it of course. But they cost money the more respondents you have, so why bother?
__________________
"The thief and the murderer follow nature just as much as the philanthropist. Cosmic evolution may teach us how the good and the evil tendencies of man may have come about; but, in itself, it is incompetent to furnish any better reason why what we call good is preferable to what we call evil than we had before."

"Evolution and Ethics" T.H. Huxley (1893)
angrysoba is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 09:10 PM   #44
Puppycow
Penultimate Amazing
 
Puppycow's Avatar
 
Join Date: Jan 2003
Location: Yokohama, Japan
Posts: 25,445
Originally Posted by zooterkin View Post
Basically, he said the polls have been fairly consistent over the last few decades. One problem is that the margin of error is often larger than the actual margin in key states, so whether they predict the correct result becomes a matter of chance.
If there was just one single poll, it would be to be expected that it might be off by the margin of error. But with how many polls they do these days, random errors like that should cancel out when you take an average of polls. There's a bias, not just a neutral error due to randomness. Because the vast majority of the errors went in the same direction. I don't mean an intentional bias, but a response bias of some sort. Whatever the reason was, the Trump voters were just less likely to respond to pollsters. That's my hypothesis.
__________________
A fool thinks himself to be wise, but a wise man knows himself to be a fool.
William Shakespeare
Puppycow is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 09:14 PM   #45
Puppycow
Penultimate Amazing
 
Puppycow's Avatar
 
Join Date: Jan 2003
Location: Yokohama, Japan
Posts: 25,445
Originally Posted by Segnosaur View Post
I took a look at the last national polls (from Nov3) on 538 and compared them to the election results (as we have now.... counting is still ongoing)...

Biden: Poll: 51.8, election: 50.6
Trump: Poll 43.3, election 47.7

So, the national polls weren't really that far off this time. Biden's results were ~1% off, and Trump's results were ~4% off. What is interesting is that there seem to be a lot more unaccounted for in the polling averages than in the final results. I wonder if this could be due to 3rd party voters that decided to switch to Trump at the end?
That's a cumulative 5.5% error, since the error for Biden was an overestimate and that for Trump was an underestimate.
__________________
A fool thinks himself to be wise, but a wise man knows himself to be a fool.
William Shakespeare
Puppycow is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 09:15 PM   #46
Puppycow
Penultimate Amazing
 
Puppycow's Avatar
 
Join Date: Jan 2003
Location: Yokohama, Japan
Posts: 25,445
Originally Posted by Venom View Post
Can polls ask more than 1000 people at a time?

Can we have a poll or set of polls over a month after asking 8,000-10,000 real people in a state?
Easily that many people were polled in the final month in the swing states. Not in a single poll, but when you add up multiple polls.
__________________
A fool thinks himself to be wise, but a wise man knows himself to be a fool.
William Shakespeare
Puppycow is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 10:09 PM   #47
Cain
Straussian
 
Cain's Avatar
 
Join Date: May 2002
Location: Los Angeles
Posts: 14,092
All roads lead back to QAnon. Yesterday the NYT reported an on-going study claiming a correlation between QAnon support and polls underestimating Trump support: https://www.nytimes.com/2020/11/06/t...ng-errors.html

But QAnon is ridiculous. How many people believe in that nonsense? Forbes reports, "Some 56% of Republicans believe that QAnon, a far-right conspiracy theory, is mostly or partly true, according to a new Daily Kos/Civiqs poll released Wednesday, a remarkably high number considering many of the outlandish assertions espoused by QAnon supporters."

I don't know what kind of reputation Daily Kos-commissioned polls have but any incentive to exaggerate must be counter-balanced by the all-too plausible notion that Q-ultists would be reluctant to share their demented beliefs with a stranger on the phone.

QAnon is just another part of the general kookery on the right. I'm always astonished at how some of this stuff gets mindlessly repeated. The original "fake news" (Facebook articles from four years ago), ANTIFA's gonna get your mama, Kenyan born president, Sharia Law, White Natioanlist ZOG crapola, all of the illegal votes in this election (and in 2016).

I intuited that Trump won his party's nomination in spite of his conspiracy-prone craziness, but maybe that's mistaken. The perception that they're out of power will only fuel more of this paranoid hysteria.

ETA: I should add that I feel icky believing that there's a group of wackos thwarting polling, which is itself a type of conspiracy.
__________________
April 13th, 2018:
Ranb: I can't think of anything useful you contributed to a thread in the last few years.

Last edited by Cain; 7th November 2020 at 10:10 PM.
Cain is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 7th November 2020, 10:33 PM   #48
Skeptic Ginger
Nasty Woman
 
Skeptic Ginger's Avatar
 
Join Date: Feb 2005
Posts: 84,537
I went back and looked at 538's state snake. They mostly nailed it there, even having Pennsylvania being the key swing state.

Their odds didn't give the impression of how close the race would be. They were way off on that.

But Biden did win. They got that right.

https://projects.fivethirtyeight.com...t/?cid=rrpromo
__________________
TRUMP IS A LOSER!
Skeptic Ginger is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 8th November 2020, 06:25 PM   #49
Cain
Straussian
 
Cain's Avatar
 
Join Date: May 2002
Location: Los Angeles
Posts: 14,092
I guessed more states correctly than 538. The only one I missed was Georgia, but I don't feel like an oracle. It's like if you say the Dodgers will beat the Tampa Bay Rays in a sweep, but the series has six close games. Sure, you correctly predicted Tampa Bay would lose, but it didn't go quite as planned.

I saw someone the other day say that if Silver predicts Democrats winning 67% in three separate elections, and the Democrats win all three, then it doesn't mean the model's working perfectly. If the model were working perfectly, Democrats would win two elections, so an 89% chance (or whatever) needs to have upsets in order to be right. Fair enough. But people said the polls were underestimating Trump's support, and those who took that into account were more likely to better predict the future.
__________________
April 13th, 2018:
Ranb: I can't think of anything useful you contributed to a thread in the last few years.
Cain is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 8th November 2020, 09:50 PM   #50
Roger Ramjets
Philosopher
 
Roger Ramjets's Avatar
 
Join Date: Jun 2008
Posts: 5,144
Originally Posted by Skeptic Ginger View Post
Their odds didn't give the impression of how close the race would be. They were way off on that.
That's because you don't understand odds. If the estimates are 51% Biden and 49% Trump, and the margin of error is 1%, that means Biden is virtually guaranteed to win and you would be an idiot to bet against it.

Quote:
But Biden did win. They got that right.
So in hindsight the actual odds were 100% in Biden's favor. Now I see what you mean by 'how close'.
__________________
We don't want good, sound arguments. We want arguments that sound good.
Roger Ramjets is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 8th November 2020, 10:15 PM   #51
Roger Ramjets
Philosopher
 
Roger Ramjets's Avatar
 
Join Date: Jun 2008
Posts: 5,144
Originally Posted by The Great Zaganza View Post
It's more a question of usefulness than reliability.
Indeed. Many people expect polls to do more than they are capable of, eg. predicting the outcome of an election.

Quote:
Arguably, the rosy polls for Democrats have led to massive misallocation of funds to races that where unwinnable and could have been better used elsewhere.
If so then that is a mistake made by the campaign managers. But they probably would have made more mistakes without looking at polls.

Another factor to consider is that published polls change voting behavior. Republicans see that Biden appears to be winning and it galvanizes them to get out and vote, while democrats see it and don't bother voting because they are lazy and figure they don't need to. What Biden should have done was release 'flawed' (ie. fake) polls that showed he was slightly behind, while letting Trump continue to falsely claim he was ahead. Then come election day - Boom! Trump doesn't know what hit him.

Quote:
Polling isn't the way forward, Big Data analysis a la Cambridge Analytica is.
No, the way forward is people not treating polls like psychic predictions. When a poll says x proportion of their responders chose y, that is all it says. Anything else is your interpretation.
__________________
We don't want good, sound arguments. We want arguments that sound good.
Roger Ramjets is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 9th November 2020, 01:46 AM   #52
Firestone
Proud Award Award recipient
 
Firestone's Avatar
 
Join Date: Oct 2006
Location: Belgium
Posts: 2,617
Just before the election, Nate Cohn (NYT) did the following exercise: "what if the polls are as wrong this year as they were in 2016?"

The result:
EmUIrrsXIAAjnUr.jpg

Except for Florida, that's exactly the current (and hopefully final) map.

Naive interpretation: the problems with the polls in 2016 haven't been solved.

Anyway, that the polls erred again, and in the same direction, is probably significant. Notice that the polls were quite good in 2018, except in ... Florida.

It's almost as if Trump, and only Trump, brings out a small but not negligible amount of voters that are not reached by pollsters.
__________________
The method of science is tried and true. It is not perfect, it's just the best we have. And to abandon it, with its skeptical protocols is the pathway to a dark age. -- Carl Sagan
Firestone is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 9th November 2020, 02:40 AM   #53
Ambrosia
Good of the Fods
 
Ambrosia's Avatar
 
Join Date: Jun 2009
Posts: 2,657
Every time the subject of polling comes up I see lots of references to Nate Silver and 538 and people arguing over how right or wrong he was.

I think it's worth pointing out that Nate is not a pollster. He's a poll aggregator.

Other people collect data, Nate does science with the data.
Like all data science it is subject to the laws of GIGO (Garbage In Garbage Out)

I think Nate is by far the best poll aggregator which is why his predictions are consistently more accurate than others, but he's still working with flawed datasets.

I think that the reasons the pollsters are underperforming is that they underestimate Trumps ability to turn out voters.
So far it seems to be the sole preserve of right wing populists to aggressively use facebook ads and the like to target, misinform, and get out to vote people who are usually politically apathetic. (yes everyone is spending big monies on facebook ads but I think it's the right wing that targets those ads much much better.)
We see all over the news coverage demographics like "no college education" which is mediaspeak for "lower than average intelligence, poorly informed" and it's voters like that I think that pollsters have the hardest time talking to.
So the pollsters throw some weighting on the ones they do speak to and it increases their margin of error.

When the polling margin of error is larger than the margin of victory, then surprise surprise calling the outcome correctly is basically a coinflip.

The GOP doesn't appeal to enough voters to win in a fair election and hasn't done so probably since Reagan. They keep using dirty tricks to win elections and they have got exceedingly good at it.

If we look at some of the early Florida results we see counties like Miami-Dade flipping GOP off the back of very precisely targeted attack ads equating Biden to being a radical leftwing socialist.

Most people see those ads and dismiss them as being ******** because of various reasons, but in that specific county there is a large number of Cuban immigrants who have come to America precisely to escape a real proper radical left wing communist regime and that resonate with them, most of them thought hell no I don't want more of that, and voted Trump.

That's a textbook example of team Trump conning large groups of people to vote for them using misinformation tactics (which like gerrymandering should be illegal, but I digress) and pollsters don't (can't?) account for things like that.
Ambrosia is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 9th November 2020, 02:46 AM   #54
shuttlt
Philosopher
 
Join Date: Aug 2008
Posts: 7,314
Originally Posted by Venom View Post
Can polls ask more than 1000 people at a time?

Can we have a poll or set of polls over a month after asking 8,000-10,000 real people in a state?
That would reduce random error, which isn't what we are talking about here. All you would be doing is getting a more precise and stable measure of the systematic error. If anything it would be worse because it would create the illusion of increased reliability.
shuttlt is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 9th November 2020, 12:16 PM   #55
StillSleepy
Muse
 
StillSleepy's Avatar
 
Join Date: Jan 2020
Location: Where the wild goose goes
Posts: 600
A study from the University of Southern California has some early results indicating a correlation between underestimating Trump's support in state polls and amount of QAnon activity in the respective states. Usual warnings about correlation, causation, and early results apply. [NYT]
__________________
I just want some outdoors left to enjoy. The wild is already artificial enough as it is.

"Scholars have long known that fishing eventually turns men into philosophers. Unfortunately, it is almost impossible to buy decent tackle on a philosopher's salary." ~ Patrick F. McManus, ďNever Sniff A Gift FishĒ
StillSleepy is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 9th November 2020, 04:35 PM   #56
PhantomWolf
Penultimate Amazing
 
PhantomWolf's Avatar
 
Join Date: Mar 2007
Posts: 19,924
Originally Posted by Venom View Post
Can polls ask more than 1000 people at a time?

Can we have a poll or set of polls over a month after asking 8,000-10,000 real people in a state?
The thing with statistics, you can actually represent a population with remarkable small sample sizes and get pretty good confidence levels.

Here's a basic outline for you but the upshot of all that math is that the confidence level for a population doesn't actually change a great deal between having 1,000 people and 10,000 people.

You can play with the figures here but you'll see that even if you want to survey a population of 330 million people with a 99% confidence level and a 3.5% interval, then you only need to have a sample of 1360 people. Boosting the sample to 10,000 people drops the interval to about 1.3% (assuming a roughly 50-50 split). It isn't that much of a difference to do ten times the work.
__________________

It must be fun to lead a life completely unburdened by reality. -- JayUtah
I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question. -- Charles Babbage (1791-1871)

PhantomWolf is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 9th November 2020, 04:40 PM   #57
PhantomWolf
Penultimate Amazing
 
PhantomWolf's Avatar
 
Join Date: Mar 2007
Posts: 19,924
Originally Posted by Roger Ramjets View Post
That's because you don't understand odds. If the estimates are 51% Biden and 49% Trump, and the margin of error is 1%, that means Biden is virtually guaranteed to win and you would be an idiot to bet against it.

So in hindsight the actual odds were 100% in Biden's favor. Now I see what you mean by 'how close'.
Unsure if serious.......
__________________

It must be fun to lead a life completely unburdened by reality. -- JayUtah
I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question. -- Charles Babbage (1791-1871)

PhantomWolf is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 11th November 2020, 06:59 AM   #58
zooterkin
Nitpicking dilettante
Administrator
 
zooterkin's Avatar
 
Join Date: Mar 2007
Location: Berkshire, mostly
Posts: 47,448
Originally Posted by PhantomWolf View Post
The thing with statistics, you can actually represent a population with remarkable small sample sizes and get pretty good confidence levels.

Here's a basic outline for you but the upshot of all that math is that the confidence level for a population doesn't actually change a great deal between having 1,000 people and 10,000 people.

You can play with the figures here but you'll see that even if you want to survey a population of 330 million people with a 99% confidence level and a 3.5% interval, then you only need to have a sample of 1360 people. Boosting the sample to 10,000 people drops the interval to about 1.3% (assuming a roughly 50-50 split). It isn't that much of a difference to do ten times the work.

That works if you are getting a random sampling of opinion. The problem is, with a 5% response rate, and with people of different opinions responding at different rates (e.g., it appears Biden supporters were more likely to admit to the fact than Trump supporters), the sample doesn't necessarily represent the views of the whole population.
__________________
The whole problem with the world is that fools and fanatics are always so certain of themselves, and wiser people so full of doubts.Bertrand Russell
Zooterkin is correct Darat
Nerd! Hokulele
Join the JREF Folders ! Team 13232
Ezekiel 23:20
zooterkin is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 12th November 2020, 01:00 AM   #59
Roger Ramjets
Philosopher
 
Roger Ramjets's Avatar
 
Join Date: Jun 2008
Posts: 5,144
Originally Posted by zooterkin View Post
the sample doesn't necessarily represent the views of the whole population.
Not just necessarily, it almost certainly won't. But even if it did, it wouldn't necessarily predict the outcome with 100% certainty.

Even if you polled every single voter and they all responded truthfully, it still wouldn't necessarily represent the views of the whole population or the election results. Some people think they will vote one way, but change their minds when it comes to pulling the lever. Some are too lazy to vote, and some are prevented from doing so for other reasons. There's only one 'poll' that is 100% accurate - the actual votes.

So what it comes down to is you are just quibbling about the accuracy of something that we know is not 100% accurate, and should be treated accordingly.

Nate Silver: The Polls Werenít Great. But Thatís Pretty Normal
Quote:
Iím afraid I have some bad news. If you want certainty about election outcomes, polls arenít going to give you that ó at least, not most of the time.

Itís not because the polls are bad. On the contrary, Iím amazed that polls are as good as they are. Given that response rates to polls are in the low single digits and that there are so many other things that can go wrong, from voters changing their minds after you poll them to guessing wrong about which voters will turn out ó plus the unavoidable issue of sampling error ó itís astonishing that polls get within a couple of points the large majority of the time...

The main reason that polls arenít going to provide you with the certitude you might desire is because polls have always come with a degree of uncertainty. In a highly polarized era, most elections are going to be close ó close enough as to exceed the ability of polls to provide you with a definitive answer. Say the final polling averages miss by a bit more than 3 points on average, as our forecast assumes. That means the margin of error is closer to 7 or 8 points. And every presidential election so far this century has fallen within that range.
__________________
We don't want good, sound arguments. We want arguments that sound good.
Roger Ramjets is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 12th November 2020, 01:22 AM   #60
PhantomWolf
Penultimate Amazing
 
PhantomWolf's Avatar
 
Join Date: Mar 2007
Posts: 19,924
Originally Posted by zooterkin View Post
That works if you are getting a random sampling of opinion. The problem is, with a 5% response rate, and with people of different opinions responding at different rates (e.g., it appears Biden supporters were more likely to admit to the fact than Trump supporters), the sample doesn't necessarily represent the views of the whole population.
The thing is that you don't just randomly poll though. You have an idea of the demographics of the population you are planning to poll and while the people sampled are random, you select them within that randomness to match close to your demographics.

So if you have a state with 20% of the population being low educated white males, then you make sure that you get low educated white males at about 20% of the sample. This is why when you agree to do one of these surveys they usually start out by asking things such as your age group, income level, and ethnicity so that they can determine if they still need to include you in their sample. It doesn't matter how many people don't answer the pollster, as long as they get the final sample proportions roughly the same as the population's proportions.

This is why people that do these things spend several years learning about Statistics and how to analyze and operate polling.
__________________

It must be fun to lead a life completely unburdened by reality. -- JayUtah
I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question. -- Charles Babbage (1791-1871)

PhantomWolf is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 12th November 2020, 03:19 AM   #61
Minoosh
Penultimate Amazing
 
Minoosh's Avatar
 
Join Date: Jul 2011
Posts: 11,801
We had a forum member in Argentina who seems very knowledgeable about polling - his user name was aleCcowaN (sp?) but I can't call up anything at all recent though I suspect there's something limiting the search results I get because the results are very old and mostly confined to a Global Warming thread.

I PM'd him asking if he was till active - he might have been having health issues, it's easy to lose track of members. I'd love to see his take.

I'm wondering if there needs to be a wholesale revolution when it comes to polls. My knowledge of methodology is very limited. I wonder if there should be a radical departure from randomness - building up a database of engaged and fairly trusted sources who will give honest answers to questions vetted to convey impartiality on the part of pollsters. I don't know if that's possible, or even a good idea. If it's just that if all polls are within a margin of error up or down that renders them useless except in blowout cases - then what good do they do? They wouldn't help inform a candidate's strategy, it seems.

Was Trump right about his much-vaunted 95 percent approval among Republicans? If so than virtually all Republicans will have voted for him. So if that's just assumed, then you look at registration margins and see how the results would work if both D's and R's showed up in equal force? I'm ignoring cross-party voters for this purpose, even though I am one. Do the independents then fall into predictable splits, and what is turnout amongst them?

I don't know if the polls ask the wrong questions, or if respondents are lying, or whatever else is confounding the results but obviously something is going on that is giving highly unreliable results. Maybe we should dispense with them all together. Parties could run their own internal polls to design strategy. The rest of us would just have to accept that generalized polls are so fallible they might as well be abandoned.

But then I hear that snake charts were in fact fairly accurate and I'm not really well-informed enough to have a good grasp on the whole "problem."

ETA: It's nice to see that Libertarians were a bit of a spoiler, presumably affecting the R's more than the D's. I can't imagine most of them preferring Biden over Trump, but I could be wrong.

Last edited by Minoosh; 12th November 2020 at 03:21 AM.
Minoosh is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 12th November 2020, 03:56 AM   #62
Firestone
Proud Award Award recipient
 
Firestone's Avatar
 
Join Date: Oct 2006
Location: Belgium
Posts: 2,617
Originally Posted by Minoosh View Post
Was Trump right about his much-vaunted 95 percent approval among Republicans? If so than virtually all Republicans will have voted for him.
We'll need the full results before being able to analyse what exactly went wrong.

One possibility is that the "Likely Voter" filter was wrong (underestimating turnout for non-college-educated people).
There may also be a type of Trump voters that is simply not reached by polls.

That in general (but obviously not everywhere) down-ballot Republicans outperformed Trump, though, suggests that Trump wasn't as popular with Republicans as he claimed.

Originally Posted by Minoosh View Post
But then I hear that snake charts were in fact fairly accurate and I'm not really well-informed enough to have a good grasp on the whole "problem."
538's snake missed two states: Florida and North Carolina. That's not bad given the unprecedented attributes of this election.
__________________
The method of science is tried and true. It is not perfect, it's just the best we have. And to abandon it, with its skeptical protocols is the pathway to a dark age. -- Carl Sagan

Last edited by Firestone; 12th November 2020 at 03:58 AM.
Firestone is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 12th November 2020, 04:11 AM   #63
PhantomWolf
Penultimate Amazing
 
PhantomWolf's Avatar
 
Join Date: Mar 2007
Posts: 19,924
Originally Posted by Minoosh View Post
I'm wondering if there needs to be a wholesale revolution when it comes to polls. My knowledge of methodology is very limited. I wonder if there should be a radical departure from randomness - building up a database of engaged and fairly trusted sources who will give honest answers to questions vetted to convey impartiality on the part of pollsters. I don't know if that's possible, or even a good idea.
There are already polls that do this. They have a set group as the sample and look at changes within that group between surveys.

Quote:
If it's just that if all polls are within a margin of error up or down that renders them useless except in blowout cases - then what good do they do? They wouldn't help inform a candidate's strategy, it seems.
It depends on what you expect from polls. There seems to be an expectation here that polls need to 100% accurate or they are totally useless, but that's not how statistics work. Polls are useful tools to give an idea of what is happening, but if they were 100% accurate then why even bother having an election in the first place? The fact that different polls have quite different results even when taken over the same time period should be a clue to how much a poll can vary. This is why aggregating polls tend to give a better view of what is happening rather than just looking at a single poll, but even that isn't perfect. Anyone that expects a poll to be more than indicative is really fooling themselves. Polls really work best looking at the trends between the polls rather than the numbers anyway. Seeing is a Candidate is increasing, decreasing, or holding steady in their popularity.

Quote:
I don't know if the polls ask the wrong questions, or if respondents are lying, or whatever else is confounding the results but obviously something is going on that is giving highly unreliable results. Maybe we should dispense with them all together. Parties could run their own internal polls to design strategy. The rest of us would just have to accept that generalized polls are so fallible they might as well be abandoned.
Again this comes down to expectation. I see nothing wrong with the polling in the end because I didn't expect it to be entirely accurate in all cases. I understand that there will be errors and that the final results will be off.

Quote:
But then I hear that snake charts were in fact fairly accurate and I'm not really well-informed enough to have a good grasp on the whole "problem."
These are based on aggregate polling and a bit of magic with a formula, and they were certainly on the money with most winners even if the expected results were put a bit. The only two places that missed on 538 were Florida and North Carolina, and both were statistical toss ups with a slight lean to Biden.
__________________

It must be fun to lead a life completely unburdened by reality. -- JayUtah
I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question. -- Charles Babbage (1791-1871)


Last edited by PhantomWolf; 12th November 2020 at 04:14 AM.
PhantomWolf is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 12th November 2020, 04:37 AM   #64
Darat
Lackey
Administrator
 
Darat's Avatar
 
Join Date: Aug 2001
Location: South East, UK
Posts: 94,785
Originally Posted by Minoosh View Post
We had a forum member in Argentina who seems very knowledgeable about polling - his user name was aleCcowaN (sp?) but I can't call up anything at all recent though I suspect there's something limiting the search results I get because the results are very old and mostly confined to a Global Warming thread.

I PM'd him asking if he was till active - he might have been having health issues, it's easy to lose track of members. I'd love to see his take.

I'm wondering if there needs to be a wholesale revolution when it comes to polls.

..snip...
There does but not in the way they are conducted, it is in how they are used.

The primary public use at the moment for election polls are to drive viewers/clickers/readers to certain media outlets. If the media used them more "honestly" they'd say - "Latest poll shows 52% of people will vote for Trump, 48% for Biden but the margin of error is +-3% so we can't call it".

As in the quote from Silver above the polls by the reputable pollsters are in fact very accurate, the "issue" isn't the accuracy of the polls it's how close the two sides are in the real world.
__________________
I wish I knew how to quit you
Darat is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Old 13th November 2020, 05:06 AM   #65
PhantomWolf
Penultimate Amazing
 
PhantomWolf's Avatar
 
Join Date: Mar 2007
Posts: 19,924
Nate Silver and Galen Druke of 538 discuss the polling and how bad the polls really were.
__________________

It must be fun to lead a life completely unburdened by reality. -- JayUtah
I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question. -- Charles Babbage (1791-1871)

PhantomWolf is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Reply With Quote Back to Top
Reply

International Skeptics Forum » General Topics » USA Politics

Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -7. The time now is 05:31 PM.
Powered by vBulletin. Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum began as part of the James Randi Education Foundation (JREF). However, the forum now exists as
an independent entity with no affiliation with or endorsement by the JREF, including the section in reference to "JREF" topics.

Disclaimer: Messages posted in the Forum are solely the opinion of their authors.