Oct 27 2008
Internal Polls Show PA A Toss-Up?
I am just not sure about the news that PA is in play, but if it is then Obama surely cannot win, even if he takes VA. But here are some stories confirming both GOP and Dem internal polls showing PA is in play. Story one:
Pennsylvania
The one blue state McCain hopes to grab will see a lot of attention from both campaigns. McCain is in Pottsville today while Obama visits Pittsburgh.
The poll of polls? Obama is up 51 to 41 percent. However, the McCain camp says internal polls show the race much closer.
Polls now show that he [Obama] is sealing the deal with these all important voters, and not just in Pennsylvania. As the economy has gone down hill, Obama’s numbers have gone up. Nationally, the reliable CNN poll of polls gives Obama an eight per cent advantage over McCain and in Pennsylvania Obama has a ten point lead over McCain. But polls in the “keystone state” have a reputation for being wildly misleading. Both Democratic and Republican internal polls last week showed the state near neck and neck. It’s true that Pennsylvania has voted for a Democrat for the last twenty years, but never decisively. No Democrat since 1964 has ever won the state with more than 51 per cent.
They definitely don’t want you to pay attention to leaked internal polls from the Obama camp that show McCain is within 2 points in Pennsylvania, a state that Obama lost by 10 points to Hillary Clinton in the Democratic primary.
Am I buying into the idea PA is in play? Not yet, but I have said this many times before (see here and here for example), polls that simply use party affiliation have never been right when it comes to the voting. Kerry barely won with 2.5% of the vote – but many polls rely on a 10-13% tilt towards the dems in their samples. Which means if PA votes like last time, it really is a tie right now.
“Am I buying into the idea PA is in play? Not yet,…”
I am. Hillary clobbered Obama in PA, even though the polls showed it close just before the vote. Obama regularly underperformed his poll numbers in the primary, and I think both a hangover from “Bittergate” and Murtha’s “western PA is racist” comment will come back to haunt him. Add to that the furor over the latest revelations of Obama’s redistributionist instincts and, yeah, I can think Pennsylvania is in play.
The data point I find most convincing: both campaigns are behaving as though Pennsylvania is a battle ground state.
I don’t think that means McCain has the election sewn up, not by a long, long way. But it’s certainly a good sign.
Obama was consistently underperforming 2% or more in Democratic Primary vote tallies compared to pre-voting Democratic primary polling.
I have no idea what that means on election day.
Not true of Obama’s underperformance of the polls in the primary. He just did not do good with the undecideds. However, in places where the black population was huge (NC, GA, VA) he outperformed the polls because black turnout was much larger than polled.
Currently, if the pollsters in those states are showing a huge black turnout in their models., then that is why some states go for Obama (like VA) or are close like (NC and GA). The question is, what will be the likely black proportion of the voting public in those states. Is there enough of a black population in these states (VA, NC, and GA) to push Obama over the top. The polls may well be projecting a very high black turnout and a very low white turnout.
It is all about turnout.
Mark N,
Please see:
http://www.thenextright.com/sean-oxendine/what-is-the-argument-for-why-mccain-will-win
Outside of the South – where he was buoyed by unusually high African American turnout that will be mitigated once the Republican vote is mixed in – Obama performed on average 2 points behind the spread the polls predicted. And look how frequently Obama gets what he gets! In AZ, CA, CT, KY, MA, NH, OH, PA, RI, TX (left in the data set, because AAs are a fairly small portion of the population here, so it isn’t “Southern†in the sense that the other states in the dataset are), VT and WV, his final result was three points or less above where he ended up in the poll average. That’s almost two-thirds of the total states in this dataset!!!
and
In other words, the undecideds are voters who are from demographic groups that are currently breaking toward McCain. And, incidentally, they are the people who are probably the least likely to be truly comfortable with a black President, even if they are convinced that they are not the least bit racist.
If you look at the map of the counties which were won in the primary Obama only carried a few of them and as usual his mostly urban dominated ones.
Many of the polling organizations tend to poll urban area for their samples which hits right into the strength of where is base is located and thus adds even more sample bias into the issue.
You can also go to CQ politics and look at the history for each congressional district making up the state and see just how much of a shift it would take to have the current poll numbers to be valid.
Also such a massive trend away from historic voting patterns would seem to require that all states would have a similar percentage shift in voter sentiment compared to historic norms.
The number of states that have shifts that vary all over the map suggests that the polling data simply can not be trusted.
The polls may well be projecting a very high black turnout and a very low white turnout.
That is meaningless. Kerry carried 90% of the black vote in 2004, Obama is projected to carry 95%, but remember that AA’s make up only 11% of the electorate. I just don’t think it is going to factor in.
What will be a factor is Ralph Nader. He was not on the ballot in 2004 in PA when Kerry carried it by 2.5%. He is this time around and will probably get 3% of the vote. That 3% is not going to come from the Republicans.
MerlinOS2-
Really good point. (Urban vs. Rural in polling…)
Makes you wonder whether Obama’s support among Democrats is as high as is being reported.
My point yesterday was turnout. AA’s made up 11% of the electorate in 2004 but with Obama will they make up 11% in 2008 or 12% or 15%, etc. Take your pick. It is not meaningless to a poll using the high end of the estimate above.
This a point that AJ and DJ should make when the blog on polls. All polls are to a large extent estimates. You take the raw data and process that through your estimates to come up with your final numbers. I see where Gallup is running with two different sets of estimates and assumptions. With estimation comes judgement. So to estimate a high AA turnout in VA, NC, and GA will by definition skew the poll numbers you produce. Whether they are valid estimates or not we will discover next Wednesday.