General Discussion

cthulu2016

(10,960 posts) Wed Jun 20, 2012, 06:16 PM Jun 2012

Bloomberg Poll (Complete)

Any news story about a poll that does not include a link to the actual poll results is not a news story.

And any news story about a court decision that does not include a link to the actual decision is not a news story.

IMO.

So for what it's worth, the Bloomberg poll is interesting reading:

http://media.bloomberg.com/bb/avfile/rQyA68BW5P20

Have spent some time with the full results, I think the poll is probably flawed. We have two possibilities.

1) Democratic Party affiliation and Obama support among independents have both jumped a lot, in a sudden fashion. That is possible, but would need to be confirmed by several subsequent polls. If that has happened in the real electorate that is excellent news, albeit surprising. Or,

2) Using Bloomberg's likely voter screen and a more representative sample, Obama is up 48%-50%, over Romney's 43%-45%.

Either way that is an excellent result for Obama.

But I will never understand the motive in embracing a favorable outlier with atypical internals while rejecting as absurd an unfavorable outlier with atypical internals. Goose. Gander.

7 replies

= new reply since forum marked as read

Highlight:

Bloomberg Poll (Complete) (Original Post) cthulu2016 Jun 2012 OP

they undersampled whites by 5%. hrmjustin Jun 2012 #1

Actually not by race. The problem is with party affiliation. cthulu2016 Jun 2012 #2

thanks for the correction. hrmjustin Jun 2012 #3

67% may well be too low. cthulu2016 Jun 2012 #4

I'll take it. demosincebirth Jun 2012 #7

Don't start the whole Goose Gander thing! Motown_Johnny Jun 2012 #5

Everyone else loves Ned Flanders! cthulu2016 Jun 2012 #6

hrmjustin

(71,265 posts)

1. they undersampled whites by 5%.

Reply to cthulu2016 (Original post)

Wed Jun 20, 2012, 06:20 PM

Jun 2012

2 or 3 percent more accurate. but the poll should not be taken to heart.

cthulu2016

(10,960 posts)

2. Actually not by race. The problem is with party affiliation.

Reply to hrmjustin (Reply #1)

Wed Jun 20, 2012, 06:29 PM

Jun 2012

Age and race are both normed in the poll, so the sample of race is exactly what Bloomberg thinks it should be.

Bloomberg expects the 2012 electorate to be 67% white. That is their prediction. The 2008 exit polls showed 74% white and that was a high minority-participation election, so I am not sure that Bloomberg's model is right, but that would be a problem with their model, not with the people polled.

Party preference is, however, not normed in this poll.

There are two ways to look at norming party affiliation. On the one hand, party preference (unlike age and race) can and does change back and forth and if you go into a poll saying there are X% of Republicans you may be wrong. On the other hand, party norming would eliminate some room for bad-samples.

As with all things in polling, the best is for there to be a diversity of methods among pollsters.

hrmjustin

(71,265 posts)

3. thanks for the correction.

Reply to cthulu2016 (Reply #2)

Wed Jun 20, 2012, 06:37 PM

Jun 2012

I think the president is up by 2 or 3%

cthulu2016

(10,960 posts)

4. 67% may well be too low.

Reply to hrmjustin (Reply #3)

Wed Jun 20, 2012, 06:42 PM

Jun 2012

I don't know Bloomberg's basis for norming to that level, or whether it is correct to do so.

So your statement may well be correct. 67% may well be too low.

To me, if whites were 74% of the electorate in 2008 and will be 67% of the electorate in 2012 then I don't see how Obama could lose, so if Bloomberg's model is right we're home-free.

But that is the level Bloomberg normed to. Hope they're right.