CDZ A discussion about polling companies.

[Mac-7]

Started in the CDZ with the faint hope that we can just have a normal discussion. So, two "thoughts" on this:

First, as I understand it, these polling companies made significant changes in their methodology in an effort to improve their accuracy over 2016. From what I saw, they weren't really that far off in 2016 but clearly made some mistakes. Unless I'm reading something wrong, they did even WORSE in 2020.

Second, I don't think this is about honesty or partisan politics. These pollsters make a LIVING doing this. This is about their JOBS, their BUSINESSES, so watching their reputations go up in flames like this can't have been their goal.

So what gives? What role do these companies play going forward, and how do they repair their shattered reputations?

Either they are liars or they are incompetent

I think they lie

 

[blackhawk]

People need to understand polls are never going to be an accurate predictor of elections the results of any poll can by swayed by many things. Just as an example in I believe in Iowa about a week before the election you had one poll that showed Joe Biden up by 17 points another showed Trump up by 7. There is no uniform standard for conducting polls so you will always get these wild variations.

 

[Agit8r]

I suspect that both elections were affected by the Bradley Effect.

The paranoia this year could well have amplified this affect, especially with phone polling.

 

[candycorn]

Started in the CDZ with the faint hope that we can just have a normal discussion. So, two "thoughts" on this:

First, as I understand it, these polling companies made significant changes in their methodology in an effort to improve their accuracy over 2016. From what I saw, they weren't really that far off in 2016 but clearly made some mistakes. Unless I'm reading something wrong, they did even WORSE in 2020.

Second, I don't think this is about honesty or partisan politics. These pollsters make a LIVING doing this. This is about their JOBS, their BUSINESSES, so watching their reputations go up in flames like this can't have been their goal.

So what gives? What role do these companies play going forward, and how do they repair their shattered reputations?

Polling is only as good as the subject matter in many cases. If you had years to study the people you're calling on the phone, you can see their tendencies and let that guide you in your determination on how reliable their response will be to a question. Like if I were to ask you about Donald Trump and you told me that he was the greatest person ever, I would be able to hold that response up against your posting history and see that you're likely not 100% committed (to put it mildly) to that statement.

Trafalgar has shown that they are a bit more accurate than others although I question how MUCH MORE accurate they are. They gave Trump a 0.7% win in Michigan and another win in Arizona. They were wrong in both of course. That not withstanding, according to their founder (who is a Trumpian fella) was on Smerconish the other day stating that part of his "secret" is to do large sample sizes. I think that is probably the best thing to do to improve accuracy because, as stated above, you're going to have a certain percentage of respondents who are going to be flaky. If you poll 10 people and they break 50/50 but two of them are flakes, the range of results becomes 7-3, 6-4, 5-5, 4-6 or 7-3. If you ask 100 people and you have 20 flakes that is the same ratio but the likelihood that you'll get the 20 flakes is not great. It gets even less if you poll a thousand people.

As for your questions, I don't think the polling was too far off in most states if you take into account the margin of error. But its caveat emptor; relying on one poll is as dumb as relying on one news source. The go-to I have is electoral-vote.com. They average polls (when there is more than one) to show the likely outcome. I don't recall all of the results of the polling but MI/WI/PA had Biden at 52-50%. Which is about where he came out at in those states with the MOE.

One thing that they have to get better on is the General Election polls that will affect the Senate races. All of them--to the best of my knowledge---missed the effect of having Trump on the ticket with vulnerable republican Senators. I'm not going to post the entire poll but here is the polling in states that had a senate race and the difference between the final average of polls and the election results. See the blue lines dive and the red lines rise:

Arizona:

Iowa:

Maine:

North Carolina:

You get the idea. What I think happened was that pollsters ignored the voters in these states either: realizing that the White House may be gone and turning out for the republican candidate OR people splitting their tickets...not being able to vote for Trump but proudly voting for their senator.

 

[candycorn]

made significant changes in their methodology

they did?

Now, the whole issue is MARGIN OF ERROR.

well then, what new methodology would remedy that?

Seriously people like you adamantly refuse to accept the deep state, once you accept it's reality everything makes sense.

Depends how one defines 'deep state', jmho

~S~

They need to get much larger sample sizes...5 to 10 X what they are getting now.

 

[candycorn]

Mac,

You should have provided data to discuss. A generalized "the polls were wrong" provides zero context in order to have an ACTUAL analytical discussion, but instead - you'll just have opinions base on feelings.

Data, or else this discussion is meaningless cannon fodder.

I initially thought the polls were outer-space levels of wrong this election, but as the overnight happened and the mail-ins started tabulating, the polls got closer as well.

Where's the post mortem? What are the numbers? How far off were they?

The polls completely botched the Senate races

Susan Collins in Maine was given either no chance or she was tightening the race.....she won in a landslide

Lindsey Graham was supposed to be in the fight of his life, he won by a landslide.

The polls got most of the contended senate races wrong

Yes. The pollsters didn't take into account the Trump effect of an electorate having a very good feeling that either their party's president was going to lose so they had better turn out and at least salvage the Senate OR republicans who split their ticket.

 

[candycorn]

Personally, I'm done with the pollsters.

Me too*









































*Until the next election.

 

[candycorn]

I think the problem with polls is there are too many “FUK YOU” voters on the right. People who refuse to participate in polls.

This gives you a population of people who are willing to answer lengthy questions. By nature, it is short on conservatives.
That is why they consistently underestimate the Republican vote.

I don't think it's that myself. Otherwise, there would have been FU votes since the dawn of time.

I think:

if you ask the question two weeks before the election, "Who are you going to vote for" to 10 people in Maryland you will get a solid response.
if you ask the question two weeks before the election, "Who are you going to vote for" to 10 people in Florida you will get a solid response.

If you ask the same question to the same 10 people in MD the day before the election, it won't change much
If you ask the same question to the same 10 people in FL the day before the election, you may find a completely different result.

The difference? Hate to say it but the smarter the voter, the less likely they are going to change their mind based on Hunter's laptop or a photo op with a bible.

 

[Skylar]

Started in the CDZ with the faint hope that we can just have a normal discussion. So, two "thoughts" on this:

First, as I understand it, these polling companies made significant changes in their methodology in an effort to improve their accuracy over 2016. From what I saw, they weren't really that far off in 2016 but clearly made some mistakes. Unless I'm reading something wrong, they did even WORSE in 2020.

Second, I don't think this is about honesty or partisan politics. These pollsters make a LIVING doing this. This is about their JOBS, their BUSINESSES, so watching their reputations go up in flames like this can't have been their goal.

So what gives? What role do these companies play going forward, and how do they repair their shattered reputations?

My understanding is that this could be unfixable.

As the issue is the oddity of folks that answer polls.

 

[JoeB131]

My understanding is that this could be unfixable.

As the issue is the oddity of folks that answer polls.

Again- Not sure if the polls were really that off.

The national Poll in the RCP Average had Biden up by 7.2. He won nationally by 3.8. So a 3.4% difference.

And it wasn't that they overestimated the number of people voting for Biden - They predicted 51.2 vs. the 51% he actually got. They underestimated Trump who got 47.2 instead of the 44% he polled. It might be late deciders or it might be people just not wanting to admit to a pollster they are frothing racists.

As for the state polls, The only state they got completely wrong was Florida.

So let's review -

Florida - RCP - Biden up by 0.9% Trump won it by 3.3%
Georgia - RCP Trump up by 1%, Biden won it by 0.3%
North Carolina RCP Trump up by 0.2, he won it by 1.3
Arizona - Biden up by 0.9, he won by 0.3
Penn - Biden up by 1.2, he won it by 1.2.
Michigan- Biden up by 4.2, he won it by 2.8.
Wisconsin - Biden up by 6.7, he won it by .07, but this had more to do with Trump picking up undecideds.
Nevada - Biden up by 2.7% he won it by 2.4.
Minnesota- Biden up by 4.2, he won it by 7.2
Texas - Trump up by 1.3, he won by 5.8% - Again, a lot of undecideds broke for Trump.

So except for Wisconsin, none of those states were horrifically off.

Now, the pollsters did get OHIO and IOWA way off, but then you look at the polling of those states, they weren't really a lot of them. No one really considered them to be in play, so there wasn't a lot of polling.

Now, if you want to see some crazy results, let's look at NY, where the RCP Average had Biden up by 19 to 32%, but he only won it by 12.7%. Of course, the latest poll they had for NY was in September.

Final Point- Beating incumbents is actually hard. Even a bad incumbent is a known quantity.

It's why since 1900, 14 incumbants have been re-elected, but only 6 have been defeated.

To put in in perspective, in 1976, everyone assumed Jerry Ford was toast. He pardoned Nixon, he stumbled in public, he presided over runaway inflation. There was the image on his watch of evacuating the embassy in Saigon, the first time America had ever lost a war (at least in the public consciousness). Not to mention the fact that he had never been on a national ticket as President or Vice-President.

Yet he still got 48% of the vote compared to Jimmy Carters' 50.1%.

 

[Muhammed]

made significant changes in their methodology

they did?

Supposedly. I guess I could have included that word.

From what I've seen, the changes they made simply skewed the results even more in favor of the authoritarian left.

Mac, have you ever heard of the term "suppression poll"?

In a nutshell, a suppression poll is a weapon used in psychological warfare against liberal democratic republics such as the USA. If eligible voter John Doe wants candidate A to win, but they constantly see polls that say candidate B is going to win by a landslide, and John Doe believes that candidate B will win by a landslide, he may not even bother to vote because he thinks voting would just be a waste of time, an extremely valuable rare resource for all mortal beings.

 

[Muhammed]

Mac,

You should have provided data to discuss. A generalized "the polls were wrong" provides zero context in order to have an ACTUAL analytical discussion, but instead - you'll just have opinions base on feelings.

Data, or else this discussion is meaningless cannon fodder.

I initially thought the polls were outer-space levels of wrong this election, but as the overnight happened and the mail-ins started tabulating, the polls got closer as well.

Where's the post mortem? What are the numbers? How far off were they?

Here in my home state of Ohio, Quinnipiac had Biden up by 4 points just before the election with a +/- 2.6 percentage point margin of error.

Trump won Ohio by 8.2 points.

That's 12.2 points off, in favor of Biden. That's 9.6 points outside of their proclaimed margin of error.

Poll Results | Quinnipiac University Poll

Quinnipiac University Poll’s list of recent and past poll results for political races, state and national elections, and issues of public concern.

poll.qu.edu

 

[G.T.]

made significant changes in their methodology

they did?

Supposedly. I guess I could have included that word.

From what I've seen, the changes they made simply skewed the results even more in favor of the authoritarian left.

Mac, have you ever heard of the term "suppression poll"?

In a nutshell, a suppression poll is a weapon used in psychological warfare against liberal democratic republics such as the USA. If eligible voter John Doe wants candidate A to win, but they constantly see polls that say candidate B is going to win by a landslide, and John Doe believes that candidate B will win by a landslide, he may not even bother to vote because he thinks voting would just be a waste of time, an extremely valuable rare resource for all mortal beings.

That theory doesnt work just one way, so it doesn't work. If John Doe wants candidate A to win...and all indicators are that candidate A is going to win...John Doe feels comfortable enough to be lazy and skip Voting.

 

[Muhammed]

made significant changes in their methodology

they did?

Supposedly. I guess I could have included that word.

From what I've seen, the changes they made simply skewed the results even more in favor of the authoritarian left.

Mac, have you ever heard of the term "suppression poll"?

In a nutshell, a suppression poll is a weapon used in psychological warfare against liberal democratic republics such as the USA. If eligible voter John Doe wants candidate A to win, but they constantly see polls that say candidate B is going to win by a landslide, and John Doe believes that candidate B will win by a landslide, he may not even bother to vote because he thinks voting would just be a waste of time, an extremely valuable rare resource for all mortal beings.

That theory doesnt work just one way, so it doesn't work.

Wrong. It can theoretically work both ways. But in realityland it doesn't.

From a psychological standpoint, people like to be on the winning team. They don't like to be on the losing team. Winning tends to motivate people, losing doesn't. That's why winning sports teams fill up the stadium and losing ones don't.

Instead of skewing their results 12.2 +/- 2.6 points in favor of Biden in Ohio, Quinnipiac could have skewed it 12.2 points in favor Trump. Do you really think that a poll showing a 20.4 point lead for Trump would have dissuaded any of his voters?

 

[candycorn]

Started in the CDZ with the faint hope that we can just have a normal discussion. So, two "thoughts" on this:

First, as I understand it, these polling companies made significant changes in their methodology in an effort to improve their accuracy over 2016. From what I saw, they weren't really that far off in 2016 but clearly made some mistakes. Unless I'm reading something wrong, they did even WORSE in 2020.

Second, I don't think this is about honesty or partisan politics. These pollsters make a LIVING doing this. This is about their JOBS, their BUSINESSES, so watching their reputations go up in flames like this can't have been their goal.

So what gives? What role do these companies play going forward, and how do they repair their shattered reputations?

My understanding is that this could be unfixable.

As the issue is the oddity of folks that answer polls.

Trafalgar uses large sample sizes. That would seem the be the best way of mitigating the oddities.

 

[G.T.]

Started in the CDZ with the faint hope that we can just have a normal discussion. So, two "thoughts" on this:

First, as I understand it, these polling companies made significant changes in their methodology in an effort to improve their accuracy over 2016. From what I saw, they weren't really that far off in 2016 but clearly made some mistakes. Unless I'm reading something wrong, they did even WORSE in 2020.

Second, I don't think this is about honesty or partisan politics. These pollsters make a LIVING doing this. This is about their JOBS, their BUSINESSES, so watching their reputations go up in flames like this can't have been their goal.

So what gives? What role do these companies play going forward, and how do they repair their shattered reputations?

My understanding is that this could be unfixable.

As the issue is the oddity of folks that answer polls.

Trafalgar uses large sample sizes. That would seem the be the best way of mitigating the oddities.

Yeah but their name sucks. Can you do something about that? Face time the CEO, tell him the name is ugly as fuck and triggers me.

 

[candycorn]

My understanding is that this could be unfixable.

As the issue is the oddity of folks that answer polls.

Again- Not sure if the polls were really that off.

The national Poll in the RCP Average had Biden up by 7.2. He won nationally by 3.8. So a 3.4% difference.

And it wasn't that they overestimated the number of people voting for Biden - They predicted 51.2 vs. the 51% he actually got. They underestimated Trump who got 47.2 instead of the 44% he polled. It might be late deciders or it might be people just not wanting to admit to a pollster they are frothing racists.

As for the state polls, The only state they got completely wrong was Florida.

So let's review -

Florida - RCP - Biden up by 0.9% Trump won it by 3.3%
Georgia - RCP Trump up by 1%, Biden won it by 0.3%
North Carolina RCP Trump up by 0.2, he won it by 1.3
Arizona - Biden up by 0.9, he won by 0.3
Penn - Biden up by 1.2, he won it by 1.2.
Michigan- Biden up by 4.2, he won it by 2.8.
Wisconsin - Biden up by 6.7, he won it by .07, but this had more to do with Trump picking up undecideds.
Nevada - Biden up by 2.7% he won it by 2.4.
Minnesota- Biden up by 4.2, he won it by 7.2
Texas - Trump up by 1.3, he won by 5.8% - Again, a lot of undecideds broke for Trump.

So except for Wisconsin, none of those states were horrifically off.

Now, the pollsters did get OHIO and IOWA way off, but then you look at the polling of those states, they weren't really a lot of them. No one really considered them to be in play, so there wasn't a lot of polling.

Now, if you want to see some crazy results, let's look at NY, where the RCP Average had Biden up by 19 to 32%, but he only won it by 12.7%. Of course, the latest poll they had for NY was in September.

Final Point- Beating incumbents is actually hard. Even a bad incumbent is a known quantity.

It's why since 1900, 14 incumbants have been re-elected, but only 6 have been defeated.

To put in in perspective, in 1976, everyone assumed Jerry Ford was toast. He pardoned Nixon, he stumbled in public, he presided over runaway inflation. There was the image on his watch of evacuating the embassy in Saigon, the first time America had ever lost a war (at least in the public consciousness). Not to mention the fact that he had never been on a national ticket as President or Vice-President.

Yet he still got 48% of the vote compared to Jimmy Carters' 50.1%.

What about the Senate races. Collins was never leading on a public poll.

 

[G.T.]

Tr ah full garr. Eww wtf

 

[candycorn]

Started in the CDZ with the faint hope that we can just have a normal discussion. So, two "thoughts" on this:

First, as I understand it, these polling companies made significant changes in their methodology in an effort to improve their accuracy over 2016. From what I saw, they weren't really that far off in 2016 but clearly made some mistakes. Unless I'm reading something wrong, they did even WORSE in 2020.

Second, I don't think this is about honesty or partisan politics. These pollsters make a LIVING doing this. This is about their JOBS, their BUSINESSES, so watching their reputations go up in flames like this can't have been their goal.

So what gives? What role do these companies play going forward, and how do they repair their shattered reputations?

My understanding is that this could be unfixable.

As the issue is the oddity of folks that answer polls.

Trafalgar uses large sample sizes. That would seem the be the best way of mitigating the oddities.

Yeah but their name sucks. Can you do something about that? Face time the CEO, tell him the name is ugly as fuck and triggers me.

Yep. If you see the guy on the right wing kook sites, he's a blobber. But his methodologies are sound in that one aspect....sample size. "Trafalgar" is probably a name he heard somewhere that was ambiguous enough to get people to pick up the phone when their caller ID came up.

 

[JoeB131]

What about the Senate races. Collins was never leading on a public poll.

Okay, let's look at that, shall we?

RealClearPolitics - Election 2020 - Maine Senate - Collins vs. Gideon

RealClearPolitics - Election 2020 - Maine Senate - Collins vs. Gideon

www.realclearpolitics.com

Only one poll had Gideon polling above 50%. That was the Q-Poll in September.

ONly three polls were taken in October, all showing Gideon at 47-48%.

They were all small samples, about 600 people.

Again, it does make sense if you are seeing the undecided break for the incumbant.