WWTBAM Bored

From the Opinionator blog at the NYTimes web site:

President Obama? Not so fast, says The New Republic’s Noam Scheiber. John McCain doesn’t need a “catastrophic external event” to defeat Obama on Election Day, Scheiber says.

“I happen to think Obama’s chances of winning are upward of 80 percent,” he writes at The Plank, his magazine’s political blog. He later adds, “But, truth be told, I can imagine a losing scenario that doesn’t involve outside events. It goes something like this: Obama wins all the Kerry states plus Iowa and New Mexico, giving him 264 electoral votes, then narrowly loses the rest of the red states where he’s currently competitive.”

Scheiber devises other plausible losing electoral maps for Obama. “Keep in mind that Obama loses if he wins all the Kerry states except Pennsylvania, even if he picks up Colorado, New Mexico, Iowa, and Virginia,” he writes. Here’s what Schieber, an Obama supporter, worries about:

1.) We’re at the point in the race when national trends may start diverging from trends in battleground states where McCain is still competing. After all, having an active campaign in a state makes a big difference. Just look at Michigan since McCain pulled out in early October. According to RCP, Obama is up 1.5 points nationally since October 2, but nearly 4.8 points in Michigan.

McCain is husbanding his resources for the absolute minimum number of electoral votes he needs to win, which means ignoring the national numbers and focusing on everything from Virginia on down the list of battlegrounds. There’s no reason to think he couldn’t lose the popular vote by 2-3 points but still win Virginia by 1.

2.) State polls seem to lag national polls, one-off polls tend to lag tracking polls, and polling averages (like the kind you find at RCP, Pollster.com, and FiveThirtyEight) lag any single poll by design. Which is to say, those state-level averages could easily be 5-7 days behind the current on-the-ground reality.

3.) While eight points is a lot to make up in two weeks, it’s not nearly as daunting over three weeks–well under half a point a day (.38, but who’s counting?). Or put differently: If, as some analysts believe, the momentum shifted somewhat toward McCain last week, then Virginia could actually be much closer than 8 points today. We just wouldn’t know it till next week.

4.) Throw in a Bradley Effect of even a point or two on top of that, and a few more costly Biden gaffes, and I don’t think Virginia is necessarily out of reach.


My only comment/question is, somehow you've got to factor in all the voting that has already occurred.

Thought this would brighten someone's day. And give them a reason to enter the QBPPC with their head held high.

nitrah55 wrote: My only comment/question is, somehow you've got to factor in all the voting that has already occurred.

In 2004, less than 1% of early voters indicated they would have changed their minds if they had waited until Election Day. The people who have already voted for Obama are those who probably wouldn't switch even if he announced he was granting Osama bin Laden political asylum in the US.

I do feel McCain has to do something to change the trajectory of this campaign, which is now looking eerily like Clinton-Dole in 1996 in which Dole made the rounds the last couple of weeks and the faithful whooped it up but there never was any movement.

I honestly wonder why the 527's have been sitting back. I would have thought they would have been bombarding the airwaves right now with Wright, Rezko, Ayers and company.

Investor's Business Daily has only had a tracking poll in the field for about two weeks, but they currently show Obama up 1.1% with 12% undecided. The margin of error is 3.5%

http://www.ibdeditorials.com/series13.aspx?src=POLLTOPN

Two things to point out. The sample size is 1000 voters which is somewhat lower than what Gallup and Rasmussen are using. Also, the number of undecideds is much higher (Gallup and Rasmussen are typically in the 2-4%). Polls that show a low number of undecideds usually employ a methodology where the pollster pushes the respondent as much as possible (i.e., well if the election was today you would have to make a decision, so who do you think that would be....). Needless to say, anyone who at first says he is undecided and then responds to this type of prodding is the least committed of voters, and if 10-15% of the voters (which would presumably be a much larger proportion of those who are actually going to vote on election day rather than early voters) are open to persuasion, then the election could break McCain's way.

One rule of thumb is that undecideds rarely break 50-50. They often break 75-25 or more in one way or the other, which turns a close race into a landslide (or vice versa).

silverscreenselect wrote:

nitrah55 wrote: My only comment/question is, somehow you've got to factor in all the voting that has already occurred.

In 2004, less than 1% of early voters indicated they would have changed their minds if they had waited until Election Day. The people who have already voted for Obama are those who probably wouldn't switch even if he announced he was granting Osama bin Laden political asylum in the US.

I do feel McCain has to do something to change the trajectory of this campaign, which is now looking eerily like Clinton-Dole in 1996 in which Dole made the rounds the last couple of weeks and the faithful whooped it up but there never was any movement.

I honestly wonder why the 527's have been sitting back. I would have thought they would have been bombarding the airwaves right now with Wright, Rezko, Ayers and company.

sss --

I think it's a couple of things.

1) Didn't both Senator McCain and Senator Obama discourage the 527s early in this election cycle?

I remember some group (MoveOn.org maybe?) getting a little pissy about Obama's comments.

And I'm almost sure McCain said something as far back as the spring (March, maybe) in an interview about what kind of campaign he wanted to run.

2) Maybe it's money.

There was a USA Today story which said that the 527s have raised something like 15% less this year than in 2004.

Now, that may tie in with having less time to do it, per the earlier McCain and Obama requests. I just don't know the timeline there.

From the Washington Post about a week ago.

http://www.washingtonpost.com/wp-dyn/co ... 01687.html

I also saw something on Huffington Post a while back (which I cannot now find) but it may have just been a riff on the WaPo story; the themes are the same.

I don't know, either. Like you, I figured there would be some 527 ads here.

Now it just may be too late.

And not JUST because of the remaining timeframe.

The reactions of GOP Senators like Olympia Snowe and Susan Collins to the robocalls and the poll data that shows that attacks aren't really working.

Maybe, the groups decided that WHATEVER attack is put out there moot at best and backfire at worst.

Look at the fundraising surge for El Tinklenberg in Minnesota after the comments by Rep. Bachman about anti-American members of Congress on TV. Close to a million bucks in a weekend and 72 hours from sure GOP seat to toss-up. Now the NRCC is pulling its support and putting into closer races.

I believe what could have made the difference for McCain here in Pennsylvania would have been a better ground game after the primary. It didn't really happen. All sorts of voter registration opportunities weren't capiltalized on, either.

What you have instead are a couple of homemade billboard signs (probably 3 feet by 5 feet) on a road in Chalfont, PA. They say, "Imagine a Palin Presidency."

There is no way on earth there should be signs like that in Chalfont.

silverscreenselect wrote:

nitrah55 wrote: My only comment/question is, somehow you've got to factor in all the voting that has already occurred.

In 2004, less than 1% of early voters indicated they would have changed their minds if they had waited until Election Day. The people who have already voted for Obama are those who probably wouldn't switch even if he announced he was granting Osama bin Laden political asylum in the US.

Actually, that's a pretty smart idea. Kind of along the lines of why Charles Manson spends the time between parole hearings trying to come up with something even shithouse rat crazier than he displayed at the last one. Just in case they were thinking of finally granting it.

mellytu74 wrote:What you have instead are a couple of homemade billboard signs (probably 3 feet by 5 feet) on a road in Chalfont, PA. They say, "Imagine a Palin Presidency."

There is no way on earth there should be signs like that in Chalfont.

The homemade sign thing reminds me of something I saw a while back. Some guy was apparently charmed by the oh so clever "nobama" bumper stickers, but too cheap to lay out a couple bucks for one. So he got him a pen and a paper and he made up his own little sign.

On the side of an SUV I see a white 8.5x11 sheet of paper taped to the window with the following written on it in pen:

NO

BAMA

I was trying to figure out what the dude had against the Crimson Tide. Then, it hit me. Ha!

As for polls (this has nothing to do with Mel's post but I don't want to make too many consecutive posts):

I think it's funny and/or interesting how people spend so much time poring over polls and yanking out their slide rules to try and make predictions and calculations and whatnot when:

A poll is only as good as the honesty of its answers. Why do people just assume that 96.5% of people actually give an honest answer? There are lots and lots of reasons to give fake answers to polls and questionnaires, especially when they have to do with sex, religion or politics.

A poll is only as good as the honesty of its answers.

As Mike Royko used to write: "For Christ's sake, lie to the bastards [taking the poll]!! It's none of their goddamn business."

ne1410s wrote:

A poll is only as good as the honesty of its answers.

As Mike Royko used to write: "For Christ's sake, lie to the bastards [taking the poll]!! It's none of their goddamn business."

Royko's column used to be syndicated in the local paper when I was a teenager. I always enjoyed reading him.

SSS, you may want to check out Nate Silver's www.fivethirtyeight.com post on this poll today. Take a close look at the 18-24 split, which has McCain capturing about 74% of the vote. This, of course, is ridiculous, but is (most likely) an artifact of the type of "likely voter" model they are using. Since to be a "likely voter" you must have voted in the last 2 elections (not just the last 2 presidential elections, but the last 2 national ones), this means the sample size for that demographic is abnormally small. Nate even speculates that, with a sample size of 98 (out of 1000 total) in that demo, the chances of the poll splitting the way this poll says they will is something like 5 billion to one against. Since almost every other poll has Obama leading in this demographic by 25 points or better, if you use that spread in the poll, it goes from a 1 point lead to a 4 or 5 point lead (which is what most of the other polls are saying).

It also appears that the IBD poll is skewed abnormally high in the south, which is primarily red-state country. While he doesn't go so far as to call this poll an "outlier," he strongly suggests that it's not really indicative of reality.

silverscreenselect wrote:Investor's Business Daily has only had a tracking poll in the field for about two weeks, but they currently show Obama up 1.1% with 12% undecided. The margin of error is 3.5%

http://www.ibdeditorials.com/series13.aspx?src=POLLTOPN

Two things to point out. The sample size is 1000 voters which is somewhat lower than what Gallup and Rasmussen are using. Also, the number of undecideds is much higher (Gallup and Rasmussen are typically in the 2-4%). Polls that show a low number of undecideds usually employ a methodology where the pollster pushes the respondent as much as possible (i.e., well if the election was today you would have to make a decision, so who do you think that would be....). Needless to say, anyone who at first says he is undecided and then responds to this type of prodding is the least committed of voters, and if 10-15% of the voters (which would presumably be a much larger proportion of those who are actually going to vote on election day rather than early voters) are open to persuasion, then the election could break McCain's way.

One rule of thumb is that undecideds rarely break 50-50. They often break 75-25 or more in one way or the other, which turns a close race into a landslide (or vice versa).

Mostly what I said here was what KT said before me.

I think they just undersampled young voters, got an abnormal result, and used that to extrapolate.

But the numbers 74 and 22 got me thinking. I wondered if those could narrow down the possible sample sizes.

It turns out there aren't too many whole numbers that can be broken down into whole-number fractions that round to 74% and 22%. The options are 23, 27, 46, 50, 54, 58, 65, 68, 69, 72, 73, 74, 76, 77, 78, 81, 82, and then 85 on up. I'm betting the sample is on the lower end of that spectrum.

Also, Mike said:

While he doesn't go so far as to call this poll an "outlier," he strongly suggests that it's not really indicative of reality.

I don't think "outlier" is a dirty word. It's a data point that doesn't fit in well with the rest of your data. All the other polls are saying it's a 4-8 point Obama lead, with margins of error of a few points. The MOE being a 95% confidence interval, we're going to have a poll every couple days outside those margins.

ToLiveIsToFly wrote:It turns out there aren't too many whole numbers that can be broken down into whole-number fractions that round to 74% and 22%. The options are 23, 27, 46, 50, 54, 58, 65, 68, 69, 72, 73, 74, 76, 77, 78, 81, 82, and then 85 on up. I'm betting the sample is on the lower end of that spectrum.

Also, Mike said:

While he doesn't go so far as to call this poll an "outlier," he strongly suggests that it's not really indicative of reality.

I don't think "outlier" is a dirty word. It's a data point that doesn't fit in well with the rest of your data. All the other polls are saying it's a 4-8 point Obama lead, with margins of error of a few points. The MOE being a 95% confidence interval, we're going to have a poll every couple days outside those margins.

I took statistics about 900 years ago, but I seem to recall that their definition of "outlier" has more negative connotations (somthing about the relative position to the mean), which is why I was trying to discount that. Your explanation is better, though.

As for your projected sample size, I think you're right about this one, it's likely on the small size, but I should point out that there's no reason it can't be any number, since the poll doesn't give ANY result more than 2 significant digits (and none to the right of the decimal point). It's likely they did some major rounding, as evidenced by the fact that the urban voters add up to 101%.

Ya know, I should really READ the posts before answering them. I totally misread your sample-size discussion, so just scratch that. And my apologies.

silverscreenselect wrote:

I honestly wonder why the 527's have been sitting back. I would have thought they would have been bombarding the airwaves right now with Wright, Rezko, Ayers and company.

I'm in Texas, so I don't see nearly as many political ads as people in Ohio, Florida, etc. I did donate money to BornAliveTruth.org since I figured that Obama's voting record on that bill would be more damning than any of the Wright, Rezko, Ayers stuff. But I guess letting babies die isn't as heinous as it used to be...

WheresFanny wrote:
On the side of an SUV I see a white 8.5x11 sheet of paper taped to the window with the following written on it in pen:

NO

BAMA

I was trying to figure out what the dude had against the Crimson Tide. Then, it hit me. Ha!

I have such a bumper sticker. Luckily, my car has yet to be egged, scratched or otherwise damaged. Although I am waiting for Bob or Nelly to hunt me down.

KillerTomato wrote:

ToLiveIsToFly wrote:It turns out there aren't too many whole numbers that can be broken down into whole-number fractions that round to 74% and 22%. The options are 23, 27, 46, 50, 54, 58, 65, 68, 69, 72, 73, 74, 76, 77, 78, 81, 82, and then 85 on up. I'm betting the sample is on the lower end of that spectrum.

Also, Mike said:

While he doesn't go so far as to call this poll an "outlier," he strongly suggests that it's not really indicative of reality.

I don't think "outlier" is a dirty word. It's a data point that doesn't fit in well with the rest of your data. All the other polls are saying it's a 4-8 point Obama lead, with margins of error of a few points. The MOE being a 95% confidence interval, we're going to have a poll every couple days outside those margins.

I took statistics about 900 years ago, but I seem to recall that their definition of "outlier" has more negative connotations (somthing about the relative position to the mean), which is why I was trying to discount that. Your explanation is better, though.

As for your projected sample size, I think you're right about this one, it's likely on the small size, but I should point out that there's no reason it can't be any number, since the poll doesn't give ANY result more than 2 significant digits (and none to the right of the decimal point). It's likely they did some major rounding, as evidenced by the fact that the urban voters add up to 101%.

Mike, you've made it clear you're not going to apologize to me about Billy Joel, and there's nothing else you have to apologize for (and yes, I know I'm quoting the wrong post).

It's been about the same amount of time since I took statistics, but my understanding of the term "outlier" goes like this: we keep talking about the MOE (or maybe it's just me) as having a 95% confidence interval that "reality" is within the MOE. What "reality" means here is that, if someone else did the exact same poll, on the exact same days, with the exact same methodology, they'd get a result within the margin of error. So 1 in 20 times, that won't happen. When it doesn't, one of the two polls (the original, or the re-polling) is an outlier. No pollsters use the same methods, but since just about everybody else seems to be coming into a different number range, it's likely that THIS poll is the outlier.

Now I feel like I'm just talking in a circle and repeating what everybody else is already saying.

ToLiveIsToFly wrote: Mike, you've made it clear you're not going to apologize to me about Billy Joel, and there's nothing else you have to apologize for (and yes, I know I'm quoting the wrong post).

Oh, very well....I'm sorry you have such bad taste in music.

ToLiveIsToFly wrote: It's been about the same amount of time since I took statistics, but my understanding of the term "outlier" goes like this: we keep talking about the MOE (or maybe it's just me) as having a 95% confidence interval that "reality" is within the MOE. What "reality" means here is that, if someone else did the exact same poll, on the exact same days, with the exact same methodology, they'd get a result within the margin of error. So 1 in 20 times, that won't happen. When it doesn't, one of the two polls (the original, or the re-polling) is an outlier. No pollsters use the same methods, but since just about everybody else seems to be coming into a different number range, it's likely that THIS poll is the outlier.

Now I feel like I'm just talking in a circle and repeating what everybody else is already saying.

LOL...I know exactly what you mean. But while you're making sense about the outlier, I still seem to remember that it's not just that something falls outside the MOE, but that it's so far from the mean as to make the results meaningless.

At any rate, I think we agree that this poll, in particular, appears to meet both definitions.

WheresFanny wrote: A poll is only as good as the honesty of its answers. Why do people just assume that 96.5% of people actually give an honest answer? There are lots and lots of reasons to give fake answers to polls and questionnaires, especially when they have to do with sex, religion or politics.

A poll is also only as good as those who don't answer. A lot of people just hang up on pollsters, either because they think it's another sales pitch or because they don't want to tell the pollster what they think (perhaps at some length, some polls can go up to 50 questions). If those people predominantly favor one candidate over another, that skews the eventual results.

mellytu74 wrote:

silverscreenselect wrote:

nitrah55 wrote: My only comment/question is, somehow you've got to factor in all the voting that has already occurred.

In 2004, less than 1% of early voters indicated they would have changed their minds if they had waited until Election Day. The people who have already voted for Obama are those who probably wouldn't switch even if he announced he was granting Osama bin Laden political asylum in the US.

I do feel McCain has to do something to change the trajectory of this campaign, which is now looking eerily like Clinton-Dole in 1996 in which Dole made the rounds the last couple of weeks and the faithful whooped it up but there never was any movement.

I honestly wonder why the 527's have been sitting back. I would have thought they would have been bombarding the airwaves right now with Wright, Rezko, Ayers and company.

sss --

I think it's a couple of things.

1) Didn't both Senator McCain and Senator Obama discourage the 527s early in this election cycle?

I remember some group (MoveOn.org maybe?) getting a little pissy about Obama's comments.

And I'm almost sure McCain said something as far back as the spring (March, maybe) in an interview about what kind of campaign he wanted to run.

2) Maybe it's money.

There was a USA Today story which said that the 527s have raised something like 15% less this year than in 2004.

Now, that may tie in with having less time to do it, per the earlier McCain and Obama requests. I just don't know the timeline there.

From the Washington Post about a week ago.

http://www.washingtonpost.com/wp-dyn/co ... 01687.html

I also saw something on Huffington Post a while back (which I cannot now find) but it may have just been a riff on the WaPo story; the themes are the same.

I don't know, either. Like you, I figured there would be some 527 ads here.

Now it just may be too late.

And not JUST because of the remaining timeframe.

The reactions of GOP Senators like Olympia Snowe and Susan Collins to the robocalls and the poll data that shows that attacks aren't really working.

Maybe, the groups decided that WHATEVER attack is put out there moot at best and backfire at worst.

Look at the fundraising surge for El Tinklenberg in Minnesota after the comments by Rep. Bachman about anti-American members of Congress on TV. Close to a million bucks in a weekend and 72 hours from sure GOP seat to toss-up. Now the NRCC is pulling its support and putting into closer races.

I believe what could have made the difference for McCain here in Pennsylvania would have been a better ground game after the primary. It didn't really happen. All sorts of voter registration opportunities weren't capiltalized on, either.

What you have instead are a couple of homemade billboard signs (probably 3 feet by 5 feet) on a road in Chalfont, PA. They say, "Imagine a Palin Presidency."

There is no way on earth there should be signs like that in Chalfont.

(I figure I can step outside of the lounge with a strictly non-partisan response)

About the 527's not playing such a large roll this cycle as compared to 4 years ago is that a lot of the ads that were played over and over again were free - played by various media outlets because of the claims in the ads somehow made them 'newsworthy'. Many of the famous ones, IIRC, had very little budget and only aired in a small number of markets, the news media, especially the 24 hour news cable networks would play them as part of the story. That gave the 527's much more exposure then they could afford to buy.

I'm guessing these networks are a bit more shy about giving exposure to these ads this go-around, maybe because they aren't as inflamitory or maybe they finally realized that they were totally played in 2004 and weren't going to be fooled by that tatic again to give them free airtime and if they want their ad to air, they would have to pay.

I don't have any data to back up my claim, just my faulty memory, but I'm pretty sure that most of the famous (or infamous) 527's got most of their air play for free and that doesn't seem to be happening this year - that is probably why they aren't having that big of an effect.

---Jay

Sir_Galahad wrote:

WheresFanny wrote:
On the side of an SUV I see a white 8.5x11 sheet of paper taped to the window with the following written on it in pen:

NO

BAMA

I was trying to figure out what the dude had against the Crimson Tide. Then, it hit me. Ha!

I have such a bumper sticker. Luckily, my car has yet to be egged, scratched or otherwise damaged. Although I am waiting for Bob or Nelly to hunt me down.

Why the hell would I want to do that?

I live in Kentucky. I can save the gas by doing that to a local instead.

The problem with polls, not just in this election but in general, is how to get a sample that's representative of the overall population in every respect that's relevent in terms of the election. That usually means weighting the raw votes according to a formula. If you interview 1000 people and 150 are black, but your model shows that the population of that state is 20% black, you don't just go out and call up until you get more black respondents, you overweight their answers appropriately, doing the same for age, sex, and party affiliation. Then you try to determine a likely voter model. The only way you can find approximately how accurate your results are is to compare your final results to the actual results and even that fails to take into account late shifts in voting (even if a poll is released election day, it usually contains results dating back to Friday or Saturday before the election. So they fix things for the next election and wind up running into new problems, just like the people who keep tweaking the BCS formula every year so it comes up with the after-the-fact "right" teams for the last year's game.

Although we identify pollsters as having a Democratic or Republican bias, no pollster wants to be seen as merely a marketing tool for a party, because they are in business and the vast majority of their business comes from other types of opinion polls (about products, general demographic trends, movies and TV shows, etc.). Marketing departments who hire polling firms don't want to pay big bucks for someone who is just going to tell them what they want to hear. So they do try to be accurate, but it's not an accurate science.

If either or both candidates thought the election was over, they would be spending time this week in Oregon, Minnesota and Georgia rather than Florida, Ohio, and Pennsylvania. The odds don't favor McCain, but there are enough odd polling results and indications from early voting that it's not going to be the cakewalk Obama supporters would like. There are two realistic scenarios for McCain to win, both assuming that states like NC, FL, MO, IN and NV break in his favor. Either he wins PA or he wins one of CO and VA. Neither of those scenarios is that farfetched. There are signs in a number of key states like FL and NC that the early youth vote is lagging behind expectations and Obama may be yet the latest Democrat to go down believing in the youth vote (and it's been his claims about the youth vote that have caused a number of the polls to skew their sampling).

A lot of the feedback I'm hearing from Obama supporters is increased nervousness about this entire situation.

silverscreenselect wrote: There are two realistic scenarios for McCain to win, both assuming that states like NC, FL, MO, IN and NV break in his favor. Either he wins PA or he wins one of CO and VA. Neither of those scenarios is that farfetched.

I think your electoral math is wrong for the "or he wins one of CO and VA" part. In the McCain-doesn't-win-PA part of the scenario, you're assuming Obama wins all of the Kerry States, right? That's 251. Not even you think Iowa or New Mexico are in play, right? At least neither one of them is in your list of states above. So that's 263. If McCain wins all the other states - he pulls out NC, FL, OH, MO, IN and NV - but he wins only one of CO and VA, he gets 266 Electoral Votes if it's Virginia and 262 if it's Colorado. If you win less than 270 EVs and your last name isn't Bush, you don't become President.

silverscreenselect wrote:If either or both candidates thought the election was over, they would be spending time this week in Oregon, Minnesota and Georgia rather than Florida, Ohio, and Pennsylvania. The odds don't favor McCain, but there are enough odd polling results and indications from early voting that it's not going to be the cakewalk Obama supporters would like. There are two realistic scenarios for McCain to win, both assuming that states like NC, FL, MO, IN and NV break in his favor. Either he wins PA or he wins one of CO and VA. Neither of those scenarios is that farfetched. There are signs in a number of key states like FL and NC that the early youth vote is lagging behind expectations and Obama may be yet the latest Democrat to go down believing in the youth vote (and it's been his claims about the youth vote that have caused a number of the polls to skew their sampling).

A lot of the feedback I'm hearing from Obama supporters is increased nervousness about this entire situation.

The latest news is that Obama will win even if he loses all of the "battleground" states.

ToLiveIsToFly wrote:

silverscreenselect wrote: There are two realistic scenarios for McCain to win, both assuming that states like NC, FL, MO, IN and NV break in his favor. Either he wins PA or he wins one of CO and VA. Neither of those scenarios is that farfetched.

I think your electoral math is wrong for the "or he wins one of CO and VA" part. In the McCain-doesn't-win-PA part of the scenario, you're assuming Obama wins all of the Kerry States, right? That's 251. Not even you think Iowa or New Mexico are in play, right? At least neither one of them is in your list of states above. So that's 263. If McCain wins all the other states - he pulls out NC, FL, OH, MO, IN and NV - but he wins only one of CO and VA, he gets 266 Electoral Votes if it's Virginia and 262 if it's Colorado. If you win less than 270 EVs and your last name isn't Bush, you don't become President.

Bush beat Kerry 286-252 (technically 251 since one Kerry elector voted for John Edwards). If Obama adds IA-7 and NM-5, that would make it 274-264. PA has 21 electoral votes, VA 13 and CO 9. If Obama flips either VA or CO he wins, unless McCain flips PA in which case, even if Obama takes both VA and CO, it's 273-265.

Not likely, but still possible.

Essentially taking PA allows McCain to lose two or three smaller states and still win.