General Discussion
Related: Editorials & Other Articles, Issue Forums, Alliance Forums, Region ForumsWhat Went Wrong with the Polls?
What Went Wrong with the Polls?
November 14, 2020 at 7:35 am EST By Taegan Goddard 77 Comments
https://politicalwire.com/2020/11/14/what-went-wrong-with-the-polls-2/
"SNIP......
Pew Research: Looking across the 12 battleground states from the upper Midwest (where many polls missed the mark) to the Sun Belt and Southwest (where many were stronger), polls overestimated the Democratic advantage by an average of about 4 percentage points. When looking at national polls, the Democratic overstatement will end up being similar, about 4 points, depending on the final vote count. That means state polling errors are about the same as in 2016, while the national polling error is slightly larger, at least as of today.
The fact that the polling errors were not random, and that they almost uniformly involved underestimates of Republican rather than Democratic performance, points to a systematic cause or set of causes. At this early point in the post-election period, the theories about what went wrong fall roughly into four categories, each of which has different ramifications for the polling industry.
......SNIP"
Eliot Rosewater
(31,109 posts)bullimiami
(13,083 posts)Polls only can tell you the intentions of what people who are likely or intend to vote.
They cant tell you what ratfuckery is going on to suppress the vote of the other party.
Its the red-shift and it has everything to do with wide ranging deliberate suppression of Democratic votes.
uponit7771
(90,323 posts)... much as they have in the last 10 yrs.
The pollsters know something is wrong, how close are the polls in blue states with blue SoS's?
I bet we're a lot close in those states than in states lead by kGOP SoS's and have red legislatures.
+1, somethings up and its not as simple or unfalsifiable as "they lied"
Ferrets are Cool
(21,105 posts)OrangeAnus has run the country.
We don't need polls. In fact, we would be better off if they did not exist. Just Go Vote.
JI7
(89,244 posts)in the country.
NameAlreadyTaken
(977 posts)questionseverything
(9,646 posts)It always surprises me that the same folks that know repubs will cheat in a hundred different ways will insist it couldnt be the paperless counting machines,that repubs produce,sell,service and program could possibly be responsible for shaving off 2-4 points so repubs win
diva77
(7,638 posts)questionseverything
(9,646 posts)Transparency is the answer
Hand counted paper ballots with a tight chain of custody are what democracy demands so we citizens can oversee our own elections
uponit7771
(90,323 posts)... in and election and we see them doing x even to the point of openly screwing up USPS to slow down the damn mail.
You're right, people are giving the kGOP boundaries the kGOP have already blown through when it comes to democracy
Shermann
(7,409 posts)It's not clear to me what the margin of error was supposed to be. If it was 4% or more, then he nailed it.
uponit7771
(90,323 posts)... up with polling being this off for this long tilted towards one party.
The polling companies that got it wrong this time wouldn't be invited back in Spain et al.
I'm thinking they're not weighting kGOP VSM in red lead states enough because they don't want to expose the efficacy of voter suppression.
I'd like to extrapolate the blue state polling and see how accurate it was, there should NO DOUBT be a deep dive on polling after this year no doubt the polling companies owe America that.
Right now, we're running blind even in GA Senate Runoff
mr_lebowski
(33,643 posts)To make a 4 point swing.
Polls are 'likely voters'. Elections are 'people who went through with the process of voting'.
I don't see how this is complicated.
Trumpists are frothing-at-the-mouth dedicated to their Cult Leader.
Biden voters ... aren't like that.
Also, all the FUCKERY is on the side of the GOP ... but I doubt this is the primary reason personally.
kurtcagle
(1,602 posts)A little discussion about margin of error is worth understanding why I have my doubts about most interactive touch screen voting machines:
I run a site (www.datasciencecentral.com) that is focused on data science and analytics, and have written about margin of error, variance and bias in elections for a while now.
There are a few key terms here that are worth defining:
Sample Mean: This is the average value of a given sample being polled.
Population Mean: This is the average of the whole population. Think of it as the sample mean if the sample was everyone who voted.
Variance: This is a measure of the average distance between a given sample value and the sample mean. The square root of the mean (more of less) is what's known as the standard deviation. Also known as variability.
Bias: The difference between the sample mean and the population mean.
Margin of Error: this is a value that is calculated based upon variances - a high margin of error indicates uncertainty - people haven't decided who they are going to vote for yet.
In the 2016 race, the variability of the race was high, indicated by the fact that both Clinton and Trump held the leads at various times in the race. When Trump won, he was on the very outer edge of the margin of error, meaning that given the uncertainty, there was about a 1 in 3 chance that he would pull it off. He did. Down ballot races fared even worse, with the Dems losing the House and Senate altogether, despite a seeming enthusiasm lead going into the race.
In the 2020 race, the variability of the race was low - Biden retained a 53 to 42 lead almost continuously from March, a difference of 11 points. He was expected to get upwards of 340 electoral votes. He ended up with a final lead of roughly 7 points of the final population tally. So the polls were off by about 4%. What's as notable, however, is that voting by mail, even in Republican states that had a vote by mail such as Arizona, tended to be consistent with the polls and with previous voting patterns. It was only in states where you saw strong touch screen voting that you saw the extreme differences (upwards of 6%) from the polls.
Now, the thing about margin of error is that it is, almost by definition, a measure of the randomness of the data. That is to say, no poll will be totally accurate - some will be a little high, some will be a little low, but on average, with enough sampling, things should even out. That wasn't the case in 2020. Down ballot races in red states did remarkably worse than expected for Democrats, while in blue states (which tend to have VBM) you had the normal spread of the Dems take some close races and the GOP take others. What that strongly suggests is that the touch screens are a potential source for ratf*kery.
So why aren't people making this public knowledge? Several reasons, none good.
1) The Republicans do not like polls because they often reveal prevailing sentiment differs dramatically from how elections are turning out. For this reason, there is a constant effort to discredit the polls (which we even see here on DU).
2) Republicans own most touch voting machines manufacturers and have in general moved to them in most red states. This would ordinarily be a matter for extreme concern as it represents a fundamental conflict of interest that should be investigated by congress, except for the fact that:
3) The GOP factions in Congress and the states have fought took and nail to keeping such voting machine reforms from getting pushed out of committee.
4) News organizations recognize that pushing for such reforms will lose them the support of a significant percentage of their subscriber base who benefit from the situation. A few people, such as Greg Palast, have been beating this drum for a while, but they are for the most part ignored by the corporate media.
5) Trump pretty much screwed the pooch for the RNC by being so blatant about voting only on Nov 3rd in person, and trying to denigrate any other form of voting. The discrepancy between VBM and touch screen is now much more obvious, and once Biden is in office, one of the key reforms that will be needed will be to push to remove touch screens from the equation.
uponit7771
(90,323 posts)... with "not that far off" constantly titled towards one party for this long.
There's something wrong with polling I'm thinking not weight VSM correctly in kGOP controlled SoS states.
questionseverything
(9,646 posts)Lindsey grahm and McConnell both have the advantage of having very little paper to count in their states...
Georgia has paper ballots(even though theyre machine produced and counted) for the first time since the mid sixties and what you know they go blue
andym
(5,443 posts)"The suggested problem:
According to this theory, Democratic voters were more easily reachable and/or just more willing than Republican voters to respond to surveys, and routine statistical adjustments fell short in correcting for the problem. A variant of this: The overall share of Republicans in survey samples was roughly correct, but the samples underrepresented the most hard-core Trump supporters in the party. One possible corollary of this theory is that Republicans widespread lack of trust in institutions like the news media which sponsors a great deal of polling led some people to not want to participate in polls."
These voters would be invisible to pollsters, as they are uncooperative. Media mistrust fanned by Trump and mind-your-own-business-type attitudes are more prevalent among the GOP, with their emphasis on selfishness and self-interest, from their point of view: why would they "waste" their time with pollsters from whom they don't get anything back. Polls are a civic-minded activity and they are uncivic minded,.
As they state in the article, a group of systematic non-responders could potentially skew polls on everything from favorability of politicians to opinion on climate change.
The best solution is related to the one they give under the "shy Trump voter" (which probably doesn't exist): ask others who they think their friends and relatives would vote for. But actually, they should consider a second more informative question: how many of your friends and relatives would not respond to pollsters and who would they vote for? That could be used to create algorithms to weight the actual responses.
Kashkakat v.2.0
(1,752 posts)yr at least. Thats why when they trot out these theories of poor pathetic shy Trump voters, I just about lose my lunch Since when aare trump voters shy about shooting off their mouths about anything ?
"Red shift" has been often considered in past elections to be possible evidence of electronic election fraud, which is something that would happen at the tabulator level. Voter suppression tactics would have same effet.
Elsewhere Ive read it is much more pronounced in battleground states, and this article suggests as much when it states that the national polling was not as far off as the polling in the battle ground states. Why would that be, what is different about the battleground states? That's where investigators should be looking. Stop it with this shy Trump voter nonsense.
applegrove
(118,577 posts)not lie to give Trump an advantage in getting people to go to the polls for sure?
kurtcagle
(1,602 posts)Maybe less than one in a thousand. If anything, I suspect that the real Trumpsters would likely be very vocal in their beliefs.
Today was also a good indication of that. Trump's MAGA crowd was maybe 10,000 people (and that'd be pushing it). I've been in protests that were ten times that larger that had difficulty getting much mainstream news coverage.
My own take is that there is likely a red-shift going on in the touch screen machines (I've written about that elsewhere here today) that suppressed Democratic votes across the board by about 4-5%, well outside the MoE. Had that not happened, Trump would have had less than 66 million people (still way too many), Biden would have had above 80 million people as of today, and the Senate would have been in Democratic hands. Too many down-ballot races that were off from the polls by about the same 4% or so.
brooklynite
(94,461 posts)LisaL
(44,973 posts)If polls were accurate, they would have shown that bunch of key states were 50-50 (since polls can't accurately pick a difference of less than 1 %).
We would have been freaking out for months before the election.
Midwestern Democrat
(806 posts)going to have to campaign "blind" to an extent they never had to before - basically not knowing at all where they really stand in the closest states.
uponit7771
(90,323 posts)... properly or the lack of it in Red Areas properly.
No other developed country puts up with polls being this off this long tilted towards one party, they would know something is up no doubt.
That's the only variable that's not a guess and both sides agree happens especially in racial minority areas where there are long lines to vote.
https://www.democraticunderground.com/100214551130
Buckeye_Democrat
(14,853 posts)And it's indeed systemic across the country, mostly inconveniencing urban (more Democratic) voters.
It might be enlightening to compare state polls and their results for different states. Was Oregon (mail voting) polling as inaccurate, for example?
uponit7771
(90,323 posts)... than red state polling with the opposite red pols.
The polling companies know this and don't want to be the one who out the efficacy of kGOP voter suppression across the country.
VSM is the one variable both sides agree happens to some degree.
radius777
(3,635 posts)despite what some say to excuse it. Again the polls were off in the same direction in the same states - this is not a random error.
Polls only measure what people intend to do, not what the can do in reality (if faced with voter suppression, long lines, discarded ballots, etc).
The other problem is people no longer answer their phones to avoid spam and robocalls, so the only ones answering are a self-selected group. How to fix this, I don't know. It's not like the olden days when each home had a landline that was mostly answered, resulting in a better sample.
Dems also share some responsibility in this as even deep blue cities had issues with lines. The fact that the Dem base exists in densely populated areas is a 'natural voter suppression' issue before we get to the intentional suppression. One machine breaks in a big city you get long lines and people who need to go to work and won't wait on the line.
Buckeye_Democrat
(14,853 posts)I don't know if that helps explain it, but maybe?
https://amp.theguardian.com/commentisfree/2020/oct/15/ohio-us-election-voter-suppression
Thanks to an Ohio state law passed in 2006 by a Republican-controlled legislature and signed by a Republican governor, the number of in-person early voting sites is limited to just one per county. That means Vinton County, a Republican stronghold in the states southeast thats home to just 13,500 Ohioans, has approximately 97 times more polling-places-per-voter than Franklin County, the deep-blue bastion with a population of more than 1.3 million.
uponit7771
(90,323 posts)LiberalFighter
(50,825 posts)Xolodno
(6,390 posts)Many chose to vote by mail, probably isn't a good model yet for this situation.
Remember, House Minority Leader McCarthy tried to get Trump to stop his anti-mail ballot initiative, but failed. They probably had data that yes, more GOP voters would vote, but more Democrats would vote by a greater margin. So we had a record GOP turnout and a record Democratic turnout. Ironically, none of this would be possible if it wasn't for COVID.
They are also still trying to get the older models to catch up to new technology.
And there is some bias here on DU, as expected. Citing some polls that haven't quite "caught up".
Remember, in the past, these things used to be called based on exit polls, they don't do that anymore. The media networks know the current models are still too new and flawed. It may take a few cycles before they feel more confident.
I'll give credit to Fox however, they called Arizona....and I have a funny feeling they would have called the election before everyone else, but had some outside influence to stop that. They have the money to employ the best statisticians, AI modelers, etc.
LisaL
(44,973 posts)They didn't realize AZ had so many outstanding ballots left on election day. They are lucky Biden won in the end, but I don't think they could have predicted his lead was going to hold for sure.
radius777
(3,635 posts)since it was VBM(vote by mail/mail in ballots) they would lean blue, which was incorrect.
There were three buckets. (1) The early VBM was blue (counted first, which led to the blue mirage), (2) the late VBM and same-day drop offs were red, (3) the provisionals were split evenly or leaned somewhat blue.
Polybius
(15,364 posts)When polled, they chose Biden, but they never voted. Neatly every Republican voted. It's the one thing they're good at.