CNN's Mark Preston admits tilting poll to include more R's.

[philly09]

Mark Preston just said on CNN that the post debate poll was skewered to include more republican voters.

[Alcon]

What did he say, verbatim?

[sentinel]

What did he say, verbatim?

[philly09]

i don't know, but it's all over twitter.

[Snowstalker Mk. II]

But the media still has a radical lie-beral bias, right guys?

[Averroës Nix]

Skewered? As in, "Skewer to the chest, who won the debate?"

[Wonkish1]

There is a difference between admitting that R sample is a little higher than normal and admitting that they did that intentionally.

Even if you readjust the sample to show a +2-3 D advantage the result is still a tie.

Non issue!

[5280]

This is not a deal breaker, CNN tends to skew more towards the left but not this time.

[opebo]

Haha, the right-wing media is predictable.

[Zioneer]

This is not a deal breaker, CNN tends to skew more towards the left but not this time.

It really, really doesn't skew to the left. That's Fox News projecting the idea, since they themselves skew so radically to the right (more than MSNBC skews to the left, even). CNN is aggressively centrist.

[Wonkish1]

Fox News polls don't skew to the right. If anything they are left of average in most polls. Fox News contracts to a joint operation with 1 D pollster and 1 R pollster just like Politico does with the Battleground poll.

Look all of the pollsters were getting more D respondents over a week ago and now all of the pollsters are getting more R respondents(than before) over the last week.

As I said before this is likely because of the 10% response rate in polls. If only 10% of the people you call participate than there is likely some self selection going on where the party that is in the better mode is more willing to respond to the poll.


By the way, question... These flash pollsters call people ahead of the debate to let them know that they'll be calling back after the debate, right? I mean it seems like they would have to because otherwise the amount of staff needed to turn out a flash poll in about 1 hour would insanely cost prohibitive.

[Adam Griffin]

What did he say, verbatim?

Here's what was published by CNN:

You must be logged in to read this quote.

The fact that Ryan only won a +4 plurality in a R+2 sample (and not to get into crosstabs too much here) means he is a big loser.

[Likely Voter]

its plausible that debate watchers were more Republican than all adults or RVs. Younger voters probably skipped it for more interesting TV and/or didn't bother. Preliminary numbers has viewership around 43M (vs 70M for Obama/Romney last week).

And since we all agree that each side liked its candidate, the numbers make sense.

[Wonkish1]

What did he say, verbatim?

Here's what was published by CNN:

You must be logged in to read this quote.

The fact that Ryan only won a +4 plurality in a R+2 sample (and not to get into crosstabs too much here) means he is a big loser.

You apparently can't do math. If you readjusted to D+3 it would be a tie. Tie doesn't equal loser.

[Badger]

The fact that Ryan only won a +4 plurality in a R+2 sample (and not to get into crosstabs too much here) means he is a big loser.

You apparently can't do math. If you readjusted to D+3 it would be a tie. Tie doesn't equal loser.

You don't see the irony (or the fail) here, do you?

[Adam Griffin]

What did he say, verbatim?

Here's what was published by CNN:

You must be logged in to read this quote.

The fact that Ryan only won a +4 plurality in a R+2 sample (and not to get into crosstabs too much here) means he is a big loser.

You apparently can't do math. If you readjusted to D+3 it would be a tie. Tie doesn't equal loser.

Exactly. Democrats under-perform Republicans when given two equally polarized scenarios (i.e: R+2 & D+2). It was a statistical tie with a R+2 sample, when Ryan should have been outside the MoE in a sample such as this - had he held his own or actually won the debate. Therefore, Ryan is a big loser.

[Wonkish1]

What did he say, verbatim?

Here's what was published by CNN:

You must be logged in to read this quote.

The fact that Ryan only won a +4 plurality in a R+2 sample (and not to get into crosstabs too much here) means he is a big loser.

You apparently can't do math. If you readjusted to D+3 it would be a tie. Tie doesn't equal loser.

Exactly. Democrats under-perform Republicans when given two equally polarized scenarios (i.e: R+2 & D+2). It was a statistical tie with a R+2 sample, when Ryan should have been outside the MoE in a sample such as this - had he held his own or actually won the debate. Therefore, Ryan is a big loser.

That is some of the dumbest logic I've ever heard. After you make the sample change it comes almost to a dead even tie. MOE doesn't become higher for Obama and lower Ryan in the process(which is essentially what you're saying even you don't realize it).

Take a damn statistics course.

[milhouse24]

I don't think the polls are accurate because as I have been saying all along, Obama's voters tend to be more fairweather voters and unreliable.  They are soft voters that easily change or over-state their support for Obama just to fit in with popular opinion.  These Obama voters tend to be younger, under 30, and more fickle.  I'm not being biased, this is just one of the reasons why I think the polls are inaccurate. 

[Adam Griffin]

Exactly. Democrats under-perform Republicans when given two equally polarized scenarios (i.e: R+2 & D+2). It was a statistical tie with a R+2 sample, when Ryan should have been outside the MoE in a sample such as this - had he held his own or actually won the debate. Therefore, Ryan is a big loser.

That is some of the dumbest logic I've ever heard. After you make the sample change it comes almost to a dead even tie. MOE doesn't become higher for Obama and lower Ryan in the process(which is essentially what you're saying even you don't realize it).

Take a damn statistics course.

This wasn't a "Who would you vote for?" poll, it was a "Who won the debate?" poll. It also wasn't a "What could have been if we adjust sampling after the fact" poll, it was a "What actually happened" poll. In any favorable Republican sample and assuming the two candidates perform equally well, Ryan should have won it outright - without splitting hairs or worrying about MoE. There is simply more soft support among Democrats than Republicans; a skilled Republican candidate will persuade more leaning Democrats to their side than the other way around.

Ryan held the easy ground he was guaranteed to hold given the sample. He didn't gain an inch otherwise, and in every other poll, 30-50% of Republicans conceded that he lost the debate. He didn't convince anyone else that he won. It was Ryan's debate to win or lose, much like it was Romney's debate to win or lose last week. Romney was considered the winner among 55-70% of voters. Ryan was considered the winner among 18-48% of voters. Therefore, Ryan is a big loser for failing to convince anyone outside of his realm of influence that he won the debate.

[Wonkish1]

Exactly. Democrats under-perform Republicans when given two equally polarized scenarios (i.e: R+2 & D+2). It was a statistical tie with a R+2 sample, when Ryan should have been outside the MoE in a sample such as this - had he held his own or actually won the debate. Therefore, Ryan is a big loser.

That is some of the dumbest logic I've ever heard. After you make the sample change it comes almost to a dead even tie. MOE doesn't become higher for Obama and lower Ryan in the process(which is essentially what you're saying even you don't realize it).

Take a damn statistics course.

This wasn't a "Who would you vote for?" poll, it was a "Who won the debate?" poll. It also wasn't a "What could have been if we adjust sampling after the fact" poll, it was a "What actually happened" poll. In any favorable Republican sample and assuming the two candidates perform equally well, Ryan should have won it outright - without splitting hairs or worrying about MoE. There is simply more soft support among Democrats than Republicans; a skilled Republican candidate will persuade more leaning Democrats to their side than the other way around.

Ryan held the easy ground he was guaranteed to hold given the sample. He didn't gain an inch otherwise, and in every other poll, 30-50% of Republicans conceded that he lost the debate. He didn't convince anyone else that he won. It was Ryan's debate to win or lose, much like it was Romney's debate to win or lose last week. Romney was considered the winner among 55-70% of voters. Ryan was considered the winner among 18-48% of voters. Therefore, Ryan is a big loser for failing to convince anyone outside of his realm of influence that he won the debate.

What you're saying makes no sense.

Actually it's the opposite of what you said. When asked who won the debate you'll get more crossover than who would you vote for. A small R+ sampling is worth less in that event not more(as your suggesting).

Look the math is simple add 4/5 pts to the spread and you come out Biden 0/+1. And actually that is giving you a generous assumption that every single R picked Ryan and every D picked Biden. Its actually more likely that some R's(likely male) picked Biden and that some D's(likely female) picked Ryan. When that is true the move in partisan ID spread actually moves the change in win loss by less not more.


Look this sometimes the problem with arguing with certain people on the left. They think something like this is a matter of opinion. It's not. What I'm describing is as cut and dry as it gets once we determine what shift we're going to apply to the poll.


And what other polls are you referring to? CBS? What other ones?

[Adam Griffin]

First of all, stop acting as if you have all the answers or that polling is a concrete science. If that were the case, then every poll would show the same result and there would be no error. The MoE was 5 points in this poll, which is a god-awful large one.

If you insist on discussing the what-ifs, then Biden and Ryan were statistically tied in a poll that was a net 5 points more Republican than it should have been (D+3 is, in my opinion, the most accurate reflection of current partisan breakdown). If you want to claim that your opinions are anything but opinions, then you should know that when in the MoE, there is no victor. It is tied. The result was a tie in a highly favorable Republican sample. What would you be saying if there had been a statistical tie in a D+8 sample? Maybe something to the effect of, "Biden is a big loser"?

And yes, CBS and ABC polls are the others I was referencing.

[Wonkish1]

First of all, stop acting as if you have all the answers or that polling is a concrete science. If that were the case, then every poll would show the same result and there would be no error. The MoE was 5 points in this poll, which is a god-awful large one.

If you insist on discussing the what-ifs, then Biden and Ryan were statistically tied in a poll that was a net 5 points more Republican than it should have been (D+3 is, in my opinion, the most accurate reflection of current partisan breakdown). If you want to claim that your opinions are anything but opinions, then you should know that when in the MoE, there is no victor. It is tied. The result was a tie in a highly favorable Republican sample. What would you be saying if there had been a statistical tie in a D+8 sample? Maybe something to the effect of, "Biden is a big loser"?

And yes, CBS and ABC polls are the others I was referencing.

When did I ever say that polling is a 'concrete science'. What you're trying to do(add a weight) is pretty concrete when it comes how to maintain the polls statistical use.

You don't assess the margin of error down to tie, and then make the adjustment and assume that the new result is without a margin of error or with another one. You make the weight first and then you assess the margin of error from there.

What I'm describing is not about the broad subject of polling itself, but instead encompasses 'how do you put a weight on a poll'. This action does narrow it enough that it can be definitely be answered from a statistical standpoint.

If you move the poll to tie(for MOE spread) and then assess the change you've statistically made the poll worse. If you add the weight first and then look at the MOE the poll is still as reliable as it was before except with the weight added.

Stop acting like everything that pops in your head is an opinion. It's not!

ABC poll was online. If that's the case then we might as well include CNBC.

[pbrower2a]

Fox News polls don't skew to the right. If anything they are left of average in most polls. Fox News contracts to a joint operation with 1 D pollster and 1 R pollster just like Politico does with the Battleground poll.

Look all of the pollsters were getting more D respondents over a week ago and now all of the pollsters are getting more R respondents(than before) over the last week.

As I said before this is likely because of the 10% response rate in polls. If only 10% of the people you call participate than there is likely some self selection going on where the party that is in the better mode is more willing to respond to the poll.


By the way, question... These flash pollsters call people ahead of the debate to let them know that they'll be calling back after the debate, right? I mean it seems like they would have to because otherwise the amount of staff needed to turn out a flash poll in about 1 hour would insanely cost prohibitive.

FoX News ordinarily has two sets of polls -- honest ones by Opinion Research and R-loaded polls by Rasmussen.