Replies

Here is my report for today:

Obama has now dropped .3% below the burnt toast level of 47%. [Insert usual disclaimers here about sample size, margin of error, etc.]

As usual, here is the reweighting of the top line numbers at the 2004 distribution of party ID (37R, 37D, 26I):

McCain 47.6
Obama 44.5
Und 7.9

These numbers point to a 3-4 win for McCain based on past election returns and exit poll data.

If, instead, I use the more Obama-friendly party ID distribution of the lower-turnout 1996 and 2000 elections (39D, 35R, 26I), I get this:

McCain 46.0
Obama 46.1
Und 7.9

These numbers point to a closer McCain win (50-48ish), based on the same exit poll data and election history.

Here is what I get when I do not reweight the poll for party ID, but do reweight it by area type (urban/suburban/rural) — most national polls (including TIPP) tend to undersample GOP-friendly rural voters, who have the highest “hang up” rate on pollsters).

Obama 46.7
McCain 44.2

Obama’s area type number is the same as the top line number, but McCain dipped by 0.4 points. But I’m not concerned. I think this is still the Halloween effect from Friday’s polling (which TIPP expressly acknowledged yesterday), plus a few more families with children in the burbs doing what families with children in the burbs do on Saturdays — getting out of the house. [Note: until learning about the significance of the “area type” demographic from McCain’s pollster a week or so ago, I never subscribed to the “weekend polls favor Democrats.” Now I do, although I suspect that it doesn’t skew the national numbers by all that much. Of course, a good pollster can compensate for this by re-weighting his polls by area type.]

A few more observations, again with the disclaimer that we are talking about small sample sizes within the subgroup. Take all of these comments with a healthy grain of salt.

McCain is now even in the Midwest (where many key battleground states are) with a whopping 12% of voters still undecided in that region.

This is potentially huge: McCain is now slightly ahead in the 25 and older group, with 11% of seniors still undecided.

McCain is now outperforming Bush among male voters and underperforming Bush with female voters. But he should close the female gap with the remaining undecided female voters and end up only slightly behind Bush. Reading this poll and the 2004 exit poll together, it is almost certain that the remaining women undecided voters supported Bush in 2004.

McCain is finally winning the Catholic vote, with another 11% in this slightly GOP-friendly demographic still undecided. Indeed, although this subsample is probably too small to leap to my next thought, I’ll blurt it out anyway: McCain has a chance to do better with Catholic voters than Bush did four years ago against Catholic Kerry. The Catholic vote will be especially important in Ohio, Pennsylvania, NH, and WI. It may also help McCain win a few more Hispanic votes in NV, CO, and NM.

One last parting comment: Oct. 29 (Obamammercial day) drops off of tomorrow’s result. I doubt that this will have any material effect on the poll itself.

Only one more of these to go! Yay.

This is very interesting. First, it gives Obama an eight point turnout advantage, which is six points too high.

Second, it oversampled the Philly area and slightly oversampled the Pittsburgh area as well; with reweighing the numbers become 50.0-44.4 (with the six point oversample). Actually, the numbers are even closer than that because Survey USA did not survey anyone in NW PA, which is a light red area.

Third, here is where it gets really interesting. Even with that extra six point advantage, Obama is performing barely at Kerry levels in the Pittsburgh and Central parts of the state, five points UNDER Kerry in the SW portion of the state (students of Obama’s huge PA primary loss to Hillary will start nodding their heads at this point), and only slightly better than Kerry in the NE part of the state. He may be in bigger trouble in PA than I thought.

Finally, McCain has three events in PA today and tomorrow: one in Delaware County in SE PA (57-42 Kerry in 2004), one in Scranton in NE PA (won 56-43 by Kerry four years ago, and one in Moon Township (Pittsburgh area, win by Kerry 53-46). I would surmise that McCain is going for the kill and he seems much closer than Bush ever did to getting it.

If this poll is right, and once you take a few more things into consideration, PA is not just razor close, it’s now tilting ever so slightly to McCain.