.

[.]

.

[The Professor]

I have an extra credit assignment for you. I want you to go back through every single poll on this forum and recalculate based on your own personal opinions of what the partisan sample should be.

[pbrower2a]

At this point nobody can reliably determine who will be the likely voters. Who will have enthusiasm and who will either go apathetic or end up 'dazed and confused'? 

This is consistent with President Obama having a 2% nationwide decrease in vote share as would be reflected in Pennsylvania by PPP.

[Bacon King]

Current partisan registration breakdown, per the NC State Board of Elections:

43.4% Democratic
31.4% Republican
  0.2% Libertarian
25.0% unaffiliated

Based on that, it looks like the partisan sample of the poll isn't too far off the mark.

[Oakvale]

Oh my God, I didn't know Umengus had moved to North Carolina.

[Devils30]

a few undecided blacks...poll still consistent with PPP's

[Devils30]

anything this close in NC means Obama has already won VA and the election.

[Bacon King]

Current partisan registration breakdown, per the NC State Board of Elections:

43.4% Democratic
31.4% Republican
  0.2% Libertarian
25.0% unaffiliated

Based on that, it looks like the partisan sample of the poll isn't too far off the mark.

How can you post on a poll forum every day and not know that registerd voters are different from LIKELY voter turnout? You're right, we dont know for certain what partisan turnout will be in '12, but none of us operate in a vacuum (well, maybe some of you do). Deduce what you can from these figures, but save 1980, there's NEVER been DEM turnout as high as '08.

Well, of course registered and likely voters are different. But considering that almost 90% of RV's passed the (apparently very lenient) LV screen here, it makes sense that the partisan breakdown poll will be fairly comparable to the actual registration numbers.

[DrScholl]

Can you say Conservadems? Lots of Eastern North Carolina is registered Democratic, but votes Republican for President. If the sample was that favorable, Obama would have been up by a few points. Every poll here does not have to be re-weighted.

[krazen1211]

North Carolina is certainly a good state for Romney to campaign on traditional marriage.

[Miles]

North Carolina is certainly a good state for Romney to campaign on traditional marriage.

Yes, he should totally emphasize social issues over the state's 9.7% unemployment rate.

Good call!

[ajb]

a few undecided blacks...poll still consistent with PPP's

And if I remember correctly, PPP's sample of D+14 is 3 points MORE Democratic than the actual voters that turned out in NC in '08.

You might want to take a look at PPP's final pre-election poll in NC in 2008.  They had a sample of D+11 (49-38)... and topline results of Obama 50, McCain 49. Pretty much dead-on.
By your line of reasoning, this might have been re-weighted back to around D+4, the biggest advantage the Democrats had had in NC since Carter.

It's probably a good idea, in other words, not to get too excited about partisan ID numbers in a poll taken almost six months before the election.

http://www.publicpolicypolling.com/pdf/2008/PPP_Release_NC_1103173.pdf

[krazen1211]

a few undecided blacks...poll still consistent with PPP's

And if I remember correctly, PPP's sample of D+14 is 3 points MORE Democratic than the actual voters that turned out in NC in '08.

You might want to take a look at PPP's final pre-election poll in NC in 2008.  They had a sample of D+11 (49-38)... and topline results of Obama 50, McCain 49. Pretty much dead-on.
By your line of reasoning, this might have been re-weighted back to around D+4, the biggest advantage the Democrats had had in NC since Carter.

It's probably a good idea, in other words, not to get too excited about partisan ID numbers in a poll taken almost six months before the election.

http://www.publicpolicypolling.com/pdf/2008/PPP_Release_NC_1103173.pdf

PPP also polled amendment 1 at a mere +16.

[Bacon King]

a few undecided blacks...poll still consistent with PPP's

And if I remember correctly, PPP's sample of D+14 is 3 points MORE Democratic than the actual voters that turned out in NC in '08.

You might want to take a look at PPP's final pre-election poll in NC in 2008.  They had a sample of D+11 (49-38)... and topline results of Obama 50, McCain 49. Pretty much dead-on.
By your line of reasoning, this might have been re-weighted back to around D+4, the biggest advantage the Democrats had had in NC since Carter.

It's probably a good idea, in other words, not to get too excited about partisan ID numbers in a poll taken almost six months before the election.

http://www.publicpolicypolling.com/pdf/2008/PPP_Release_NC_1103173.pdf

PPP also polled amendment 1 at a mere +16.

Better than the second most accurate pollster, which predicted a 61% to 32% defeat of the amendment

[Dan the Roman]

There is far far too much whining about poll weightings this year. Here is a dirty little secret - the aggregate of polling, regardless of weighting, has generally been right,. These same complaints have existed in every cycle, been brought up consistently by the side convinced they were losing, and then on election day, most polls were, well pretty good with the exceptions of oddities like Nevada Senate 2010.

Whining about cross-tabs is pointless, and will be revealed in November as largely baseless. For some reason, and I have seen this since 2000, polls that are D+15 in internals tend to produce results that are pretty damn close to electorates that tell exit pollsters they are D+4.

[Dan the Roman]

There is far far too much whining about poll weightings this year. Here is a dirty little secret - the aggregate of polling, regardless of weighting, has generally been right,. These same complaints have existed in every cycle, been brought up consistently by the side convinced they were losing, and then on election day, most polls were, well pretty good with the exceptions of oddities like Nevada Senate 2010.

Whining about cross-tabs is pointless, and will be revealed in November as largely baseless. For some reason, and I have seen this since 2000, polls that are D+15 in internals tend to produce results that are pretty damn close to electorates that tell exit pollsters they are D+4.

I hope you and I both can remember to revisit this in November. I'm convinced that SurveyUSA will not be polling 15% more Democrats in their final NC poll before the election. I'm also convinced PPP won't be polling 13% more Democrats in their final NC poll before the election.

I just remember seeing this in 2008 and 2004, when there were some absurd D heavy samples(approaching D+20 or more nationally). I think part of it is racial block voting, but for some reason Democrats are less coherent, independents more liberal, and republicans more or less the same in polls compared with real life.

To be honest, I would expect NC to be about 52-48 today, which this poll would more or less support if you assume the other crosstabs are absurd, others are 4% too high, AAs are too low(they were 23% in 2008 and they are up as a % of registered voters), and whites slightly too low.

[pbrower2a]

Can you say Conservadems? Lots of Eastern North Carolina is registered Democratic, but votes Republican for President. If the sample was that favorable, Obama would have been up by a few points. Every poll here does not have to be re-weighted.

On the other side are the old Rockefeller Republicans who migrated from the Northeast to North Carolina. These people are culturally incompatible with the Religious Right -- well-heeled, but also well-educated. If the Rockefeller Republicans outnumber the (George) Wallace Democrats in a state, then the Republican nominee is in trouble in that state.

North Carolina in 2008 was so close that one could interpret the Obama victory as the consequence of the size of the vote of almost any D-leaning political minority -- Latinos, Jews, Asians, people with graduate degrees, gays...

Watch the military vote. Mitt Romney is not a war hero, and if the Obama campaign can hit Mitt Romney as inexperienced or reckless on foreign policy, then President Obama has a fresh ace up his sleeve.  

[ajb]

a few undecided blacks...poll still consistent with PPP's

And if I remember correctly, PPP's sample of D+14 is 3 points MORE Democratic than the actual voters that turned out in NC in '08.

You might want to take a look at PPP's final pre-election poll in NC in 2008.  They had a sample of D+11 (49-38)... and topline results of Obama 50, McCain 49. Pretty much dead-on.
By your line of reasoning, this might have been re-weighted back to around D+4, the biggest advantage the Democrats had had in NC since Carter.

It's probably a good idea, in other words, not to get too excited about partisan ID numbers in a poll taken almost six months before the election.

http://www.publicpolicypolling.com/pdf/2008/PPP_Release_NC_1103173.pdf

Where are you getting your facts regarding last time D turnout was that low? NC was R+1 as recently as 2004 in the Bush-Kerry race.

I think we're agreeing here. I'm certainly taking my numbers from the data you'd posted upthread. The last time the Democrats had done better than D+4 was under Jimmy Carter; as you add, 2004 had been R+1.  I'm suggesting that your reaction  to the final 2008 PPP NC poll might have been to say that the race would be no better than D+4; if you think you'd have said it would be more like R+1, then you're reinforcing my point.

You must be logged in to read this quote.

Obviously, I see that 15>11. I'm disputing your claim that emphasizing one aspect of the data (partisan ID weighting) to the exclusion of all others is a useful way of understanding poll results. Remember that PPP and SUSA don't weight for partisan ID; they do weight for age, sex and race, for example, so if you want to dispute their findings it might make more sense to look at those numbers (which, unlike party ID, derive directly from assumptions made by the polling firm). Even then, of course, the topline numbers are inherently much more reliable than the crosstabs, due to smaller sample sizes.

[ajb]

Fun example from the SUSA poll of how screwy their crosstabs can be. Here are their numbers for the 7% of their NC sample who are Hispanic:

Romney 51
Obama 27
Other 21
Undecided 0

Now 7% is likely too many Hispanics (3% four years ago in the exit polls, which are admittedly often wrong about minorities), but in this case oversampling Hispanics helped Romney!

I could rant about how this makes the poll garbage, and I'm sure there's someone on here who'd be willing to claim that Romney has a unique appeal to NC Hispanics (as do the Libertarians...), and that these numbers will turn out to be right.
But either would be silly.

[tmthforu94]

I'm not sure why so many are in denial of bawlexus' logic behind the sampling - it's very clear that the sample they used was probably more Democratic than turnout will actually be. Democrats will have an edge thanks to state trends, but it won't be that big. That'd make Romney lead by, perhaps 3 or 4 points, which isn't an outlier by any means.

[DrScholl]

Re-weighted? Look, i'll go real slow: when a poll sample is D+15 in North Carolina, that means the sample is FOUR POINTS MORE DEMOCRATIC than the actual voters that turned out in 2008. So if you believe that NC Democratic turnout will exceed 2008 Democratic turnout by around 4 points, then I certainly understand why you have no problem with a D+15 poll sample. For people that do NOT believe NC Dem turnout will exceed '08, then  they would have a problem with this poll sample. What's so hard to understand about that? Why are people so people so down on assessing poll samples? It's not that hard to do when you can compare them to elections past.

Let me go slow so you understand my point, which is that since many registered voters in North Carolina vote Republican for President, you can't automatically jump to the conclusion that a Democratic sample gives the Democrat the edge. 43-44% is pretty much a floor for the President, it's not like the sample really inflated his numbers any.

[ajb]

I understand that you can't rule out a poll based on a subset of the data, and I haven't ruled any of these polls out. I'm merely pointing out a trend being followed by most pollsters - one that follows a 2008 turnout model of record-breaking high Democratic turnout. This site aggregates all of the poll samples out there: http://elections.huffingtonpost.com/pollster/party-identification (well, not all of them - excludes some important ones like YouGov, Mason Dixon, SurveyUSA). Based on the averages, most polls are sampling 6% more Democrats than Republicans. That's higher than '06 Dem turnout, & nearly the same as '08. To be fair, the PPP/Dkos polls over the last 2 wks have surprised me, as they reflect what most see as a realistic partisan turnout in 2012 (D+5 this week and D+3 last week), and I indicated that in comments at the time the poll was released. I'm not suggesting any polls are junk, nor will you find me doing that. Just cautioning that turnout means everything, which is what makes partisan ID is worth following.

Just remember that almost none of these pollsters weight for party ID. So there's no "trend being followed by most pollsters."

[krazen1211]

a few undecided blacks...poll still consistent with PPP's

And if I remember correctly, PPP's sample of D+14 is 3 points MORE Democratic than the actual voters that turned out in NC in '08.

You might want to take a look at PPP's final pre-election poll in NC in 2008.  They had a sample of D+11 (49-38)... and topline results of Obama 50, McCain 49. Pretty much dead-on.
By your line of reasoning, this might have been re-weighted back to around D+4, the biggest advantage the Democrats had had in NC since Carter.

It's probably a good idea, in other words, not to get too excited about partisan ID numbers in a poll taken almost six months before the election.

http://www.publicpolicypolling.com/pdf/2008/PPP_Release_NC_1103173.pdf

PPP also polled amendment 1 at a mere +16.

Better than the second most accurate pollster, which predicted a 61% to 32% defeat of the amendment

Well, actually, they ARE the second most accurate pollster. Survey USA had it at +20.

[Tender Branson]

This is already a likely voter poll. Didn't know that SurveyUSA would start with that so early.

For a likely voter poll, the sample is very good for Obama. And the overall topline is also what I would expect. NC is going to be tied for the next months.

[Miles]

This is already a likely voter poll. Didn't know that SurveyUSA would start with that so early.

For a likely voter poll, the sample is very good for Obama. And the overall topline is also what I would expect. NC is going to be tied for the next months.

Exactly.

We all know that NC will probably be decided by less than 5 points either way in November. All this debate about poll sample proportions is driving me crazy.