2012 Polling Accuracy

[randomrogue]

http://fivethirtyeight.blogs.nytime...best-and-worst-in-the-2012-presidential-race/

Look at the bottom of that list. As expected Gallup and Rasmussen were way off.

I'm curious at which point some of these polling firms simply don't get used for the statistics. They already didn't use Pharos but from now on will Rasmussen and Gallup get the boot as well?

 

[PottedMeat]

lolwtf?

Merriman River Group R +15.7

 

[Ausm]

I love it must because they lean Rightwing?

 

[randomrogue]

lolwtf?

They did 1 poll and it was robo calls.

 

[buckshot24]

The final Gallup poll had Romney up 1 instead Silver assumes that Romney was always down 2.6 to figure its error of +7.2 R. Now, I have no doubt that Romney was never really up 6 or 7 at any point during the campaign but there was a consensus at one point that he was up.

 

[nageov3t]

seems like many underestimated Democratic turnout, overestimated Republican enthusiasm, and need to stop relying on only polling people with landlines.

 

[buckshot24]

seems like many underestimated Democratic turnout, overestimated Republican enthusiasm, and need to stop relying on only polling people with landlines.

I was reading Rasmussen's take on why he was off and it was turnout of the young and minority voters. He was off by a few percentage points on both which underestimated Democratic turnout. As long as you get your demographics right it shouldn't matter how you contact them.

 

[buckshot24]

I also think polling didn't take into account Obama's get out the vote machine and Romney's total failure in this department.

 

[fskimospy]

The final Gallup poll had Romney up 1 instead Silver assumes that Romney was always down 2.6 to figure its error of +7.2 R. Now, I have no doubt that Romney was never really up 6 or 7 at any point during the campaign but there was a consensus at one point that he was up.

There was never a consensus that anything other than Obama winning the electoral college was going to happen though. The outcome of the election had been obvious for months.

 

[buckshot24]

There was never a consensus that anything other than Obama winning the electoral college was going to happen though. The outcome of the election had been obvious for months.

Irrelevant. National polls had Romney up for most of October. Silver is just talking about national polls (I believe) in this article so any other factors shouldn't matter

 

[techs]

I was reading Rasmussen's take on why he was off and it was turnout of the young and minority voters. He was off by a few percentage points on both which underestimated Democratic turnout. As long as you get your demographics right it shouldn't matter how you contact them.

Did Rassmussen really make such a bullshit claim? Hell, Nate Silver has been all over the news because he was explaining why the polls had been so accurate.

The only polls that were inaccurate were polls like Rassmussen who were influenced by the Republicans claims that they were inaccurate.

Pollsters have many tools to determine probable turnout. Most of the polls were accurate this year with their margin of error. They did a great job.

The outliers, to the best of my knowledge, all used Republican claims that their supporters enthusiam was higher and the Dems lower than what the pollsters were actually measuring. They gave the pollsters no hard data to support this. But some pollsters either wanted to protect their polls against Republican retaliation or to give a result so skewed it would be 'news' and their polls would receive more publicity.

 

[Athena]

seems like many underestimated Democratic turnout, overestimated Republican enthusiasm, and need to stop relying on only polling people with landlines.

The point is, most others didn't make that mistake but the Romney campaign decided to adjust the results to conform to its own world view.

The bottom line is that Republicans start with a fixed idea of what an "American" is (or more accurately should be) then reject any objective evidence that reality may be something else.

 

[fskimospy]

Irrelevant. National polls had Romney up for most of October. Silver is just talking about national polls (I believe) in this article so any other factors shouldn't matter

No, he is definitely not just talking about national polls and so the failure in the state polling is in fact very relevant. Rasmussen again performed quite poorly and with a major Republican bias. I'm not sure why anyone took them that seriously even before the election though, considering their recent history of significant Republican bias. My guess is because people want to be told what they want to hear, they don't want to look at reality.

 

[jagec]

I was reading Rasmussen's take on why he was off and it was turnout of the young and minority voters. He was off by a few percentage points on both which underestimated Democratic turnout. As long as you get your demographics right it shouldn't matter how you contact them.

If your demographics are good enough that you can predict how people are going to vote without actually polling them, than why call ANYONE? Just make a prediction and then justify it after the fact with "demographics".

The fact of the matter is that if you only use landlines, you're going to have so few young voters in your sample that you may as well just be throwing darts. There is no justification for landline-only polls in this day and age.

 

[techs]

The point is, most others didn't make that mistake but the Romney campaign decided to adjust the results to conform to its own world view.

The bottom line is that Republicans start with a fixed idea of what an "American" is (or more accurately should be) then reject any objective evidence that reality may be something else.

Well when your idea of "science" starts with the idea the Earth is 6,000 years old the idea of magic turnout fairies helping your side to win is not so difficult to beleive.

 

[buckshot24]

Did Rassmussen really make such a bullshit claim? Hell, Nate Silver has been all over the news because he was explaining why the polls had been so accurate.

I think they were assuming 2008 levels or something close so it wasn't absurd to make those assumptions. All polling firms weight for demographics. You have to make some kind of assumptions when polling 1000 people. They had whites at 2 points higher than what actually voted.

The only polls that were inaccurate were polls like Rassmussen who were influenced by the Republicans claims that they were inaccurate.

A shit load of polls were off this time.

Pollsters have many tools to determine probable turnout. Most of the polls were accurate this year with their margin of error. They did a great job.

So was Rasmussen.

I suggest that you quit looking in the grassy knoll for conspiracies when none exist.

 

[nageov3t]

The point is, most others didn't make that mistake but the Romney campaign decided to adjust the results to conform to its own world view.

it was pretty much every internal Republican polling group, not just the Romney campaign.

The bottom line is that Republicans start with a fixed idea of what an "American" is (or more accurately should be) then reject any objective evidence that reality may be something else.

wat?

how is underestimating youth/minority turnout rejecting people as non-Americans?

 

[fskimospy]

I think they were assuming 2008 levels or something close so it wasn't absurd to make those assumptions. All polling firms weight for demographics. You have to make some kind of assumptions when polling 1000 people. They had whites at 2 points higher than what actually voted.

A shit load of polls were off this time.
So was Rasmussen.

I suggest that you quit looking in the grassy knoll for conspiracies when none exist.

Rasmussen was one of the least accurate pollsters available and they were inaccurate in the same way that they have been inaccurate for several election cycles now. This is why everyone was telling you before the election that basing your judgment on their polls was such a silly idea, and sure enough it was.

Until they refine their methods Rasmussen simply isn't a very trustworthy pollster.

 

[buckshot24]

If your demographics are good enough that you can predict how people are going to vote without actually polling them, than why call ANYONE? Just make a prediction and then justify it after the fact with "demographics".

Nobody is saying this at all, you can't do it. If for some random reason you contact 80% whites in your poll you have to weight those responses down to what you expect to show up to vote. If you get 20% under 30 you have to count their responses more heavily if you expect more than 20% turnout of that group.

The fact of the matter is that if you only use landlines, you're going to have so few young voters in your sample that you may as well just be throwing darts. There is no justification for landline-only polls in this day and age.

Maybe but not necessarily. If you get less young people (lets say half as many) because you are using landlines only then you count the opinions of the ones you get in contact with more heavily. I agree that cell phones being included are more desirable, definitely. I just don't think that you're going to get totally unreliable results when you don't included them.

 

[RabidMongoose]

I was reading Rasmussen's take on why he was off and it was turnout of the young and minority voters. He was off by a few percentage points on both which underestimated Democratic turnout. As long as you get your demographics right it shouldn't matter how you contact them.

Seems like the GOP was hoping that minorities don't exist.

 

[Athena]

how is underestimating youth/minority turnout rejecting people as non-Americans?

Read the comments from Republican operatives. Dick Morris thought that things would return to "normal" -- i.e. that the elements of the Obama coalition would melt away because they only voted for him as a symbol in 2010, not as politically active, engaged Americans. This was illustrated by Sununu's insulting statement that the President as needed to "learn to be American", followed by his dismissal of the foreign policy observations of the former Secretary of State (and former Chairman of the Joint Chiefs of Staff) as nothing more than racial solidarity. Many other Republicans said that they didn't think those voters would "make the effort" to vote (IOW these are lazy elements of society and they don't vote).

Underestimating the importance of the trends documented in the 2010 census wasn't just an unlucky oversight, it was the result of blinders built-in to the party's view of the electorate.

 

[buckshot24]

Rasmussen was one of the least accurate pollsters available and they were inaccurate in the same way that they have been inaccurate for several election cycles now. This is why everyone was telling you before the election that basing your judgment on their polls was such a silly idea, and sure enough it was.

Until they refine their methods Rasmussen simply isn't a very trustworthy pollster.

The only poll I was basing anything on was their party affiliation poll which was pretty accurate going back 10 years in predicting voter turnout. They were off by like 14 points this time in that particular poll.

 

[buckshot24]

Seems like the GOP was hoping that minorities don't exist.

I was hoping that this post didn't really exist.

 

[RabidMongoose]

Read the comments from Republican operatives. Dick Morris thought that things would return to "normal" -- i.e. that the elements of the Obama coalition would melt away because they only voted for him as a symbol in 2012, not as politically active, engaged Americans. This was illustrated by Sununu's insulting statement that the President as needed to "learn to be American", followed by his dismissal of the foreign policy observations of the former Secretary of State (and former Chairman of the Joint Chiefs of Staff) as nothing more than racial solidarity. Many other Republicans said that they didn't think those voters would "make the effort" to vote (IOW these are lazy elements of society and they don't vote).

Underestimating the importance of the trends documented in the 2010 census wasn't just an unlucky oversight, it was the result of blinders built-in to the party's view of the electorate.

Republicans have serious racial issues. It's as if their fantasy of an America with no minorities (e.g., the Romney constituency) blended with what they perceived reality to be!

 

[fskimospy]

The only poll I was basing anything on was their party affiliation poll which was pretty accurate going back 10 years in predicting voter turnout. They were off by like 14 points this time in that particular poll.

Well if you weren't taking the rest of their poll and using it that means you were attempting to graft voter turnout models from one polling firm onto the results of another firm despite the two not necessarily dealing with similar samples or similar methods. This is a big statistics no-no. That just means instead of trusting an inaccurate pollster you were just working from a fundamentally flawed understanding of how this stuff works.