Pollster says election could end in landslide

[Denny Crane]

I think this theory is unsubstantiated, unless you mean to say that "some %" could mean no meaningful %. People who aren't going to vote for Obama because of race aren't going to tell pollsters they plan to vote for him. They're more likely to lie about why they aren't voting for him.

You repeated my argument. Not exactly mine, but my previous posts about it. See Bradley Effect.

Source?

3 days ago.
http://hotair.com/archives/2008/09/...rt-for-obama-unchanged-since-her-endorsement/

 

[SportsTicker]

You repeated my argument. Not exactly mine, but my previous posts about it. See Bradley Effect.

How did I repeat your argument? Your position is that the polls in general (except Zogby) are wrong because they don't account for a supposedly hidden bias. My argument is that racism is already reflected in the polls, because racists aren't going to pretend to support Obama.

3 days ago.
http://hotair.com/archives/2008/09/...rt-for-obama-unchanged-since-her-endorsement/

I don't see this substantiated in the actual poll data they claim is their source.

 

[Denny Crane]

How did I repeat your argument? Your position is that the polls in general (except Zogby) are wrong because they don't account for a supposedly hidden bias. My argument is that racism is already reflected in the polls, because racists aren't going to pretend to support Obama.

Link?

I don't see this substantiated in the actual poll data they claim is their source.

See what you want to see.

http://www.cnn.com/2008/POLITICS/08/26/clinton.backers/

Meanwhile, a CNN/Opinion Research Corp. poll taken August 23-24 found that 56 percent of registered voters have a favorable opinion of Hillary Clinton, with 40 percent having an unfavorable view. The margin of error was plus or minus 4.5 percentage points. iReport.com: Are you in Denver?


So while Clinton's die-hard supporters are in Denver in full force, the poll showed that her negative approval rating is very high among registered voters nationwide.


But there is some bad news for Obama. The poll showed that 66 percent of Clinton supporters -- registered Democrats who want Clinton as the nominee -- are now backing Obama. That's down from 75 percent in the end of June. Twenty-seven percent of them now say they'll support McCain, up from 16 percent in late June.

 

[SportsTicker]

Link?

You want more evidence, how about: "The argument of a specific 'Bradley effect,'" insisted Langer, "still looks to me to like a theory in search of data ... I don't see why this effect would be limited, before now, to a handful of elections 15 to 25 years ago. And I don't know how to understand its absence in so many other black-white races -- five [Senate and governors'] races in 2006 alone, as I note -- in which pre-election polling was dead on."

"Newton's Law of Gravity doesn't just work on Thursdays," Langer said. "You want an effect to be clearly established as an effect through analysis of empirical data, and maybe in more than one election. And to call it an effect you want it to be a consistent effect, or to explain its inconsistency."

Source: http://www.salon.com/news/feature/2008/01/24/white_voters/

Here's a whole series of analyses of the Bradley Effect:
http://www.fivethirtyeight.com/search/label/bradley effect
http://www.fivethirtyeight.com/2008/08/persistent-myth-of-bradley-effect.html

See what you want to see.

I want to see a credible and relevant (which in polling means somewhat recent) primary source. The hot air article you sourced refers to an AP article with a conclusion unsubstantiated by the published results of their own poll linked from their own article. The cnn article cites polling data which does not appear to be generally available, and which can hardly be considered relevant given that its over a month old.

 

[Denny Crane]

You want more evidence, how about: "The argument of a specific 'Bradley effect,'" insisted Langer, "still looks to me to like a theory in search of data ... I don't see why this effect would be limited, before now, to a handful of elections 15 to 25 years ago. And I don't know how to understand its absence in so many other black-white races -- five [Senate and governors'] races in 2006 alone, as I note -- in which pre-election polling was dead on."

"Newton's Law of Gravity doesn't just work on Thursdays," Langer said. "You want an effect to be clearly established as an effect through analysis of empirical data, and maybe in more than one election. And to call it an effect you want it to be a consistent effect, or to explain its inconsistency."

Source: http://www.salon.com/news/feature/2008/01/24/white_voters/

Here's a whole series of analyses of the Bradley Effect:
http://www.fivethirtyeight.com/search/label/bradley effect
http://www.fivethirtyeight.com/2008/08/persistent-myth-of-bradley-effect.html

Hillary Clinton won the New Hampshire primary.

http://townhall.com/Columnists/John...dley_effect_blew_up_the_new_hampshire_polling

There's still a lot of debate about what went wrong with the polling in New Hampshire. Personally? With the benefit of hindsight, I think that it's clearly, unequivocally the Bradley Effect at work.

Let's cover the bases on why I think that's so.

First of all, the polling on the Republican side was solid and did a good job of reflecting the actual results. So, there was obviously a factor on the Democratic side that was not in play with the Republicans.

Additionally, the polling definitely pointed towards a Barack Obama victory. There were 22 polls done in the last 3 days before the election, 20 of which had Obama winning, along with 1 tie, and 1 small Clinton victory.

Moreover, there were 4 polls that actually included data from the last day before the election and Obama won them by 5, 7, 9, and 13 points respectively. If you compare those polls to the polls that included data from the two previous days only, Barack's margin of victory actually appeared to be getting LARGER. (Average victory of 8.5% on the 4 polls that included Monday data, and a 7% average victory on the 5 polls that included only Saturday and Sunday data).

I want to see a credible and relevant (which in polling means somewhat recent) primary source. The hot air article you sourced refers to an AP article with a conclusion unsubstantiated by the published results of their own poll linked from their own article. The cnn article cites polling data which does not appear to be generally available, and which can hardly be considered relevant given that its over a month old.

Go take your own poll. I'm attributing ~ 1/3 of those from a month ago moving back to the Democratic ticket.

Or is yesterday's poll good enough?

http://www.cbsnews.com/stories/2008/09/25/opinion/polls/main4478890.shtml?source=mostpop_story

Obama leads McCain with women, moderates, Democrats, and younger voters. Sixty-one percent of those who voted for Hillary Clinton in the Democratic primary back the Illinois senator, while one in four former Clinton supporters back McCain.

 

[SportsTicker]

Hillary Clinton won the New Hampshire primary.
There's still a lot of debate about what went wrong with the polling in New Hampshire. Personally? With the benefit of hindsight, I think that it's clearly, unequivocally the Bradley Effect at work.

"Cherry Picking Results. The notion of the Bradley Effect gained a lot of currency after the New Hampshire primary, when Hillary Clinton did much better than anyone expected and won the state. However, the 8.9-point gap separating the pre-election polls and the actual results in New Hampshire represented only the seventh-largest error in the primaries. There were bigger discrepancies in Iowa, South Carolina, Alabama, Georgia, Wisconsin and Mississippi, all of which favored Barack Obama. These discrepancies did not receive as much attention as New Hampshire because they did not change the outcome of the election. But mathematically speaking, they were just as important."

http://www.fivethirtyeight.com/2008/08/persistent-myth-of-bradley-effect.html

Or is yesterday's poll good enough?

Yesterdays is great, although they don't reveal what methodology they used, so its hard to rely on their conclusion. But even stipulating that its true - it doesn't seem to be having much impact on the overall result.

 

[Denny Crane]

"Cherry Picking Results. The notion of the Bradley Effect gained a lot of currency after the New Hampshire primary, when Hillary Clinton did much better than anyone expected and won the state. However, the 8.9-point gap separating the pre-election polls and the actual results in New Hampshire represented only the seventh-largest error in the primaries. There were bigger discrepancies in Iowa, South Carolina, Alabama, Georgia, Wisconsin and Mississippi, all of which favored Barack Obama. These discrepancies did not receive as much attention as New Hampshire because they did not change the outcome of the election. But mathematically speaking, they were just as important."

http://www.fivethirtyeight.com/2008/08/persistent-myth-of-bradley-effect.html

Great. So we've gone from one wrong article that said there's ZERO evidence of a Bradley effect in 20-25 years to some other wrong article, since you're shown evidence of the Bradley effect THIS YEAR.

It wasn't one poll that was wrong by a wide margin in New Hampshire, but 20.

Yesterdays is great, although they don't reveal what methodology they used, so its hard to rely on their conclusion. But even stipulating that its true - it doesn't seem to be having much impact on the overall result.

18M ballots were cast for Clinton in the primaries. The polls favoring Obama assume that his support is going to come from people who've not participated much in the past, even though they may pass the "likely voter" part of the interview. The historical data shows that these voters do not turn out in the expected numbers when the actual election happens. The Hillary voters actually did turn out. Good luck!

Last election, roughly 120M voters turned out. 62M for Bush, 59M for Kerry. That is the only reasonable baseline to use. 3M Hillary voters going for McCain would turn those into 65M - 56M or a 9M difference in the popular vote. Popular vote doesn't win the electoral college, but you have to figure out where those 9M voters live and how it affects those states' likely outcome. Which is what Zogby is surely looking at to make his statement.

 

[SportsTicker]

Great. So we've gone from one wrong article that said there's ZERO evidence of a Bradley effect in 20-25 years to some other wrong article, since you're shown evidence of the Bradley effect THIS YEAR.
It wasn't one poll that was wrong by a wide margin in New Hampshire, but 20.

You are making two errors:
1) You conclude that the only possible explanation for the polling error is the Bradley Effect.
2) You ignore a huge amount of evidence from other elections that doesn't support your theory.

This is a pretty weak standard for such a confident conclusion that its, "clearly, unequivocally the Bradley Effect at work".

You think RCP is bunk because they average the polls, polls that include outliers. And yet, in this case you are picking the outlier, and ignoring all the other data.

You're entitled to your opinion, but I'm gonna have to respectfully disagree.

Which is what Zogby is surely looking at to make his statement.

Zogby isn't sharing his methodology with us in any detail, and considering his past bold predictions that didn't bear out, I think i'm gonna have to go with Nate Silver who is totally open and extremely detailed about his methodology when he says there is an 18% chance of a landslide for Obama, and a 2% chance for McCain. In other words a longshot for both.

I'm happy to make a friendly wager against any odds higher than what fivethirtyeight shows if you're willing to stand behind Zogby's prediction

 

[Denny Crane]

You are making two errors:
1) You conclude that the only possible explanation for the polling error is the Bradley Effect.
2) You ignore a huge amount of evidence from other elections that doesn't support your theory.

This is a pretty weak standard for such a confident conclusion that its, "clearly, unequivocally the Bradley Effect at work".

You think RCP is bunk because they average the polls, polls that include outliers. And yet, in this case you are picking the outlier, and ignoring all the other data.

You're entitled to your opinion, but I'm gonna have to respectfully disagree.

I'm looking at 22 polls that are outliers (hence they're not by definition actually outliers).

You're making the mistake of assuming that the Bradley Effect is true in every state or city. It's never been tested at the national level.

Zogby isn't sharing his methodology with us in any detail, and considering his past bold predictions that didn't bear out, I think i'm gonna have to go with Nate Silver who is totally open and extremely detailed about his methodology when he says there is an 18% chance of a landslide for Obama, and a 2% chance for McCain. In other words a longshot for both.

I'm happy to make a friendly wager against any odds higher than what fivethirtyeight shows if you're willing to stand behind Zogby's prediction

Zogby's prediction is his prediction. His boldest last prediction was that Bush would win the election (day before election day) against Kerry. His other prediction discussed here was that the election was "Kerry's to lose" as of May that year. What fault are you truly finding with his predictions?

I'm not sold on Nate Silver or sabermetrics applied to polling at this point. The jury's out. He's going to have to refigure his methodology if he's way off.

Zogby does share his methodology in detail:

http://www.zogby.com/methodology/readmeth.dbm?ID=1128
http://www.zogby.com/news/ReadNews.dbm?ID=1064
http://www.zogby.com/methodology/index.cfm
&c

 

[huevonkiller]

The bottom line is that Obama is running up the score on McCain as of late. Rasmussen has him +6 now.

 

[SportsTicker]

You're making the mistake of assuming that the Bradley Effect is true in every state or city.

I don't assume that at all. The data shows that you can have large differences between the aggregate polling data and the election outcome, and that in a statistically large number of cases, that variance CANNOT be attributed to the Bradley Effect. So this shows that what you've offered as your strongest evidence, the aggregate error, is itself not proof (its necessary but not sufficient).

The claim that the republican races in the same state didn't show the same error is irrelevant because they were different candidates and different voters, including a wildly different demographic making the two results pretty much impossible to compare and draw conclusions from. If this evidence somehow ruled in the Bradley Effect, it would also rule in the Hillary Effect, or any other "Effect" you could find between the candidates that you could claim was relevant to voters and wasn't present in the Republican primary. That is an obvious fallacy, so its clear why this claim isn't in any way evidence of the Bradley Effect.

 

[Denny Crane]

I don't assume that at all. The data shows that you can have large differences between the aggregate polling data and the election outcome, and that in a statistically large number of cases, that variance CANNOT be attributed to the Bradley Effect. So this shows that what you've offered as your strongest evidence, the aggregate error, is itself not proof (its necessary but not sufficient).

The claim that the republican races in the same state didn't show the same error is irrelevant because they were different candidates and different voters, including a wildly different demographic making the two results pretty much impossible to compare and draw conclusions from. If this evidence somehow ruled in the Bradley Effect, it would also rule in the Hillary Effect, or any other "Effect" you could find between the candidates that you could claim was relevant to voters and wasn't present in the Republican primary. That is an obvious fallacy, so its clear why this claim isn't in any way evidence of the Bradley Effect.

You win. There are no people who will vote against Obama because of his race.

It's never happened before, as it can easily be explained away by cherry picking the data.

 

[SportsTicker]

Of course there are some people who will vote against(for) Obama because of his race. But the argument is whether the polls already reflect that (are the voters lying to the pollsters about who they would support?), and by extension whether that means there is an undercurrent of racism that leaves the door open to a McCain landslide. If people change their minds at the election booth, in my opinion it will be more about identity politics and 527 fud than about anything else.

In any case I'm sure all of us following this can't wait to see what happens.