MI: Public Policy Polling: Obama up double-digits

[krazen1211]

Junk poll!

[krazen1211]

Because somebody needs to say it:

Partisan breakdown in this poll: D 32 R 28 I 40

Partisan breakdown in MI in 2008, per the exit polls: D41 R 29 I 29

So by the reasoning of many on this board, this poll is likely to underestimate Obama's lead in Michigan.

Your replication of that reasoning is quite incorrect. When partisan breakdown is examined, it is typically done by using factual registration counts from the state, not some dubious other poll.

[krazen1211]

So the people who were disputing the NBC/WSJ poll yesterday because it had a D+11 sample were wrong, because they weren't basing their concerns on factual registration counts (which, obviously, do not exist at the national level)?

Not wrong, no. They were merely looking at the registration counts of the states that do have it, and what historical partisan identification nationwide has been for decades.

[krazen1211]

Michigan is Likely Democratic, end of story. Detractors can't even blame an unfavorable sample for this one, because has a lot less Democrats than the actual registration numbers of the state.

That is amusing, given that Michigan has no partisan registration.

Facts are a tricky thing.

[krazen1211]

So the people who were disputing the NBC/WSJ poll yesterday because it had a D+11 sample were wrong, because they weren't basing their concerns on factual registration counts (which, obviously, do not exist at the national level)?

Not wrong, no. They were merely looking at the registration counts of the states that do have it, and what historical partisan identification nationwide has been for decades.

Where are they getting their numbers of historical partisan identification nationwide, if not from things like exit polls?
And how are they estimating partisan id in states like Michigan that don't have partisan registration? Because it's pretty difficult to get an accurate sum when some of the numbers are complete unknowns.

I presume they use gallup surveys and other national surveys of the public that have historically been shown to be consistent and reliable, as opposed to, say, dubious exit polling that often times is proven wrong within minutes.



Whether you choose to believe PPP or 7 other pollsters is up to you. Certainly nobody can change your mind.

[krazen1211]

Nate Silver on Michigan/PPP
(1) No the Michigan polls don't make any sense (2) Just average them, m'kay? (3) Would be nice to get a top-shelf pollster involved there.

[krazen1211]

Michigan is Likely Democratic, end of story. Detractors can't even blame an unfavorable sample for this one, because has a lot less Democrats than the actual registration numbers of the state.

That is amusing, given that Michigan has no partisan registration.

Facts are a tricky thing.

That was an error, but the FACT still stands. The data I was looking at is based off the previous exit poll and my point still stands that the sample is less Democratic than it could have realistically been. The sample has less Democrats than the 2004 exit poll.

Michigan is Likely Democratic, end of debate.

Of course, when Dr. Scholl decides, we might as well end the debate. Why have an election?

[krazen1211]

I'm just curious to know how Gallup's partisan ID numbers have been proven reliable, when there's nothing to confirm them against.

Well, aside from the fact that they tend to be confirmed by others who do the same exercise, they match up with actual electoral results. Namely, when gallup polls shifted by 5 points from the Republican party between 2004 and 2008, the GOP, well, lost roughly 5 points in Presidential and downballot races.

Certainly consistency matters, and PPP's bizarre trend of continuously releasing outlier polls (remember Colorado +14) is, well, their business I suppose.