Pre-election polling is all lies

[Leeea]

https://www.washingtonpost.com/poli...6a9838-e7df-11eb-ba5d-55d3b5ffcaf1_story.html

Polls understated the support for Trump in nearly every state and by an average of 3.3 percentage points overall. Polls in Senate and gubernatorial races suffered from the same problem.

That is more then just a minor error.

He said that polling in 2018 was generally better than in 2016, which led some pollsters to believe the problems had been resolved. Then came 2020, and the problems reemerged.

In short, the pre-election polls cannot reliably predict anything.

“A larger polling error was found in states with more Trump supporters.”

I would love to see a randomized audit after every election. Not this partisan stupidity we have now, but pick 100 counties at random and just audit them.

 

[eelw]

I'm only referring to exit polls, but one can't get an accurate poll if many are that ashame and willing to lie who they actually voted for. And this has to spill over for some even in an anonymous poll.

 

[s0me0nesmind1]

Crazy! Who would have guessed that random asking people of a question can result in:

1) People lying about their answer
2) A non-mixed and non-diverse sample pool of people who choose to do the poll
3) A non-mixed and non-diverse sample pool of people who do not have access to the poll, even if they would take the poll.

The biggest hurdle you're going to have is just generational things: If you poll via phone calls, there is a certain generations and other demographics that are trained to simply never answer a phone call unless it's an already saved number. This results in more of the older generations being more probable - that are trained and grew up to always answer the phone when possible.

Similarly, if a poll was online - you would be less likely to have more of the older generation, etc. Goodluck getting around that.

 

[Indus]

The problems isn't polls in general but people don't answer pollsters. So you can never get a reliable sample.

 

[K1052]

The state level misses were worse than the presidential. Republicans are not cheering this on since they also incinerated absolute mountains of cash because their own internals showed the same things as the public polling.

Lindsay Graham did not spend 100M dollars because he thought he would comfortably win.

 

[fskimospy]

https://www.washingtonpost.com/poli...6a9838-e7df-11eb-ba5d-55d3b5ffcaf1_story.html

That is more then just a minor error.


In short, the pre-election polls cannot reliably predict anything.


I would love to see a randomized audit after every election. Not this partisan stupidity we have now, but pick 100 counties at random and just audit them.

1) the polls were basically as accurate in 2020 as they always are. Do you think polls have never been trustworthy?

2) It seems odd to say that polls cannot reliably predict anything when 538 was able to use the polls to correctly predict 95% of races. That’s pretty good!

The main problem with polls is the readers, not the poll. The expect the polls to be right to a level that they have never been, and to a level that’s probably impossible to achieve.

 

[woolfe9998]

Evidence of people lying to pollsters is slim. 2016 and 2020 (but not 2018) were more off than usual because in both cases the pollsters used a turnout model which underestimated republican turnout. After 2016, they tried to adjust but didn't predict that the turnout would be even higher in 2020 than it was in 2016.

The factor in common in those two elections versus all the others is Trump.

 

[Moonbeam]

Did I just vote for a worthless asshole, you want to know. Hell no.

 

[Greenman]

Within 3% seems pretty good to me. I'd be impressed if they hit within 10%.

 

[Gardener]

More accurate in 2020 than the presidential election cycle of 2016. Trump brought out a pool of new voters, and both elections had higher than average voter turnouts, both factors lead to less accurate modeling.

Inaccurate polling does not mean inaccurate election result tabulation, one thing has nothing to do with the other.

 

[pmv]

Maybe people are telling the truth in polls, they're just lying in the actual ballot?

 

[shortylickens]

Polling in America is not great. Theres assloads of articles on this but for some reason it tends to favor democrats. That might sound like a good thing for democrats but its really not. They have no idea if they are actually ahead or if the polls are skewed. And since they never know for sure it kinda makes the polls useless. Same for republicans. "Are we really behind, or are the polls just saying we're behind?" Almost a waste of money.

Exit polls are usually not great either. The most famous one was the first black man to run for mayor of Los Angeles. A lot of people on exiting claimed they had voted for him.
Dude got slaughtered when they counted the actual votes.
Later somebody did the math and realized the only way he could have lost so badly is if all of the white people and most of the black people flat out lied about voting for him on the exit polls. Had he received more support from black voters his numbers would have been much better. Turns out they just didnt like his politics. But obviously when asked about it they werent going to say so. White people too. They said what they thought would get them the least amount of grief.

 

[shortylickens]

I was wrong.
Tom Bradley won the election for mayor.
It was when he ran for governor that he lost. And he'd been polling to win.

 

[ewdotson]

Evidence of people lying to pollsters is slim. 2016 and 2020 (but not 2018) were more off than usual because in both cases the pollsters used a turnout model which underestimated republican turnout. After 2016, they tried to adjust but didn't predict that the turnout would be even higher in 2020 than it was in 2016.

The factor in common in those two elections versus all the others is Trump.

Yeah, I think that's the big question. I don't think we have enough data points to say with certainty, but it certainly *seems* plausible that it boils down to the polls being extra suspect when Trump is on the ballot and having more nominal performance when he isn't. In theory '22 should tell us more - although the way the GOP is going/has gone off the rails makes me wonder if we're going to see more of a "Trump effect" even when he's not on the ballot going forward.

 

[fskimospy]

Yeah, I think that's the big question. I don't think we have enough data points to say with certainty, but it certainly *seems* plausible that it boils down to the polls being extra suspect when Trump is on the ballot and having more nominal performance when he isn't. In theory '22 should tell us more - although the way the GOP is going/has gone off the rails makes me wonder if we're going to see more of a "Trump effect" even when he's not on the ballot going forward.

According to 538 polling accuracy was basically in line with historical norms.

The Polls Weren’t Great. But That’s Pretty Normal.

I’m not a pollster, although I’m often misidentified as one on TV. I wanted to get that out of the way because while, in practice, our lives probably get easier…

fivethirtyeight.com

 

[dank69]

Do any of these control for voter suppression? It's one thing to say "I'm voting against Trump" and a completely different thing to actually stand in line for 5 hours to do it.