While Barack H. Obama pulled into a bigger (but still catchable) lead in most of the national polls, John S. McCain has suddenly started improving in the more critical state polling, bizarrely enough. Ohio is now back to being a toss-up state, with a new Mason-Dixon poll showing McCain leading for the first time in a long, long time, after nearly a week of the average leaning towards Obama.
Whoops, correction: The national polls did not tick up for Obama today; I spoke too soon. When the rest of today's tracking polls came in, Obama's already small lead shrunk 0.7% down to 5.3; this is based upon Rasmussen (holding steady), Gallup traditional (Obama drops 2 points from yesterday), Zogby (Obama up one tick), and IBD (Obama down two ticks). We're still waiting for the newest release of the Battleground poll.
One other interesting point on the national polling: Rasmussen has a ludicrous (in my opinion) turnout model where Democrats will outperform Republicans by 6.5%! In 2004, the gap was about 1.5% on election day; in 2006, it was 6.1%. But there are two related points to note:
First, what Rasmussen is measuring is not turnout but rather the party affiliation claimed by respondents in a separate poll of adults. In other words, at best, Rasmussen is measuring total party registration (self reported) -- not what percent of each party will turn out. Thus, when they weight their polling on that basis, they're making the stealth turnout assumption that all the first-time registered Democrats and youth Democrats and such are just as likely to vote as older Democrats or Republicans who have voted every election for many years, that 6.5% more registered Democrats directly translates into 6.5% more votes for Obama. This is a questionable projection, to say the least.
Second and even more intriguing... Rasmussen predicts a 6.5% Democratic advantage in turnout and adjusts its polling accordingly; but even so, they have Obama up by only 5%. Aren't they tacitly admitting that McCain will do significantly better in winning over Democrats than Obama will in turning Republicans to the dark side?
Now let's get back to the state polls released today...
Virginia is on the bring of becoming a toss-up (it's been leaning Obama for a while), and would already be except for one CNN poll; Nevada, while still technically leaning Obama, would also be a toss-up, except for one single-day poll I've never heard of -- Suffolk -- from a week ago.
Colorado is achingly close to being a toss-up as well; the most recent Denver Post/Mason-Dixon poll has Obama ahead by only 5 points, and Obama's overall lead is only 5.5.
If we change those states to toss-ups on the RCP electoral map, Obama drops below the magic 270 for the first time in a couple of weeks -- a stunning improvement for McCain. If the most recent polls are accurate, that means that John McCain can now win the election by only winning enough of those toss-ups, without having to take a single state where Obama has a significant lead right now.
But that's not all... in Pennsylvania, where John McCain and Sarah Palin have been campaigning heavily (against the advice of well-meaning Democrats, who have been advising McCain that he can't win there, so he should pull out), there are five polls in the RCP average; one of them, Marist, is clearly an outlier (Obama +14, twice as big a lead as the next nearest poll). Take Marist out of the mix, and Obama only leads by 5.3% in Pennsylvania -- making that state as close as makes no difference to a toss-up as well.
If we make Pennsylvania a toss-up, that leaves Barack Obama at only 243 in the electoral count -- rather, 242, since McCain is very likely to win that one electoral vote in Maine decided by a Republican-leaning district.
I have always assumed that McCain will win every state that is already a for-real toss-up in the RCP average right now, since he's a closer... and in the primaries, Obama underperformed his polling in nearly all the later races against Hillary Clinton. Thus I give McCain Florida, Georgia, North Carolina, Indiana, Missouri, North Dakota, Montana, and Arizona. Those are all states that went to George W. Bush in both the 2000 and the 2004 elections; McCain has improved his standing in each of those states recently, so he has momentum; and in fact, McCain currently either leads by more than 1% or is tied (less than 1% lead for one or the other nominee) in all of those states except Florida, where he trails by only 3.3 points, within the margin of error (I don't count the addlepated Los Angeles Times poll).
Everything then comes down to four states: Nevada, Colorado, Ohio, and Pennsylvania. All except the last are states George W. Bush won twice. Obama must win two of them -- and the correct two -- to take the election away from McCain; he has to win one of either Ohio or Pennsylvania, and then he must win Colorado. If McCain wins both Ohio and Pennsylvania, or if he wins one of them plus Colorado, then he wins the race, 270 to 268.
I suspect McCain is going to win Ohio, Colorado, and Nevada, giving him a narrow but comfortable victory of 275 to 263... but he might sweep all four on a reasonably good night, making it a more convincing 296 to 242.