5) Where Are the Overnights? When McCain pollster Bill McInturff released his now-infamous memo he made it clear that his spirits had been buoyed by a remarkable night of polling on Monday. In fact, the version sent around the campaign he wrote, "Pls pray for another night of tracking like the one I had last night!!!"

Well, since that release, he's traded broadsides with Stan Greenberg as to its accuracy. And I took him to task, as well. But here's the thing - all of that has focused on numbers that were produced on Monday night. And, as McInturff himself notes, those numbers were a sharp departure from what he'd been seeing. In polling, we have a technical term for this - outliers. If McInturff had been able to replicate those results on subsequent nights, much less improve upon them, I expect we would have seen still-more memos, or evidence of it in his responses. But we haven't. So we can safely dismiss McInturff's original numbers as aberrational.

4) And Speaking of Outliers... The much hyped Zogby tracking poll, which showed a 1-point McCain lead in its overnight sample on Saturday, reported a 10-point Obama lead in its Sunday overnight sample. This is why most pollsters won't break out results on a nightly basis - 400 voters, with no multiple-night callbacks, simply isn't large enough sample. 

Zogby's explanation sure is entertaining, though: 

So what happened to give McCain a one-point lead in the one-day polling on Friday? It was a day of consolidation for him, too. He had been losing support among key groups and began to regain some of his own base. He now leads by 21 points among NASCAR fans, 9 among investors, 6 among voters in armed forces households, and 2 among voters over 65 years old.

Instead of blaming random variation, Zogby doubles down, confidently asserting that McCain had been "losing" support, then had a magical day of "consolidation," before presumably hemorrhaging support again. And he proves it by citing numbers from a motley collection of presumptively friendly subgroups. If it takes guts to report results based on a sample size of 400, breaking out armed forces households takes some serious cojones. With fewer than 3 million adults residing in households with a member of the active-duty military, guard, or reserves, these voters can't comprise more than 3% of Zogby's total sample. So when he says McCain maintains a six-point lead, he means something like 33 voters to 31. But Zogby's not done:

Remember, as I said yesterday, one day does not make a trend. This is a three-day rolling average and no changes have been tectonic. A special note to blogger friends: calm it down. Lay off the cable television noise and look at your baseball cards in your spare time. It is better for your (and everyone else's) health.

The continued presumption that the poll is reflecting actual changes on a daily basis, but that these changes may not represent a "tectonic shift" unless maintained for several days, is breathtaking. But what I really love is the swipe at his most trenchant critic, Nate Silver. When he's not blogging on polling at FiveThirtyEight.com, Nate's day job remains as a co-author of Baseball Prospectus. I'm sure that if Nate wrote a little less about polling, it would be better for Zogby's health, but he's doing the rest of us a world of good.

3) Black Voters: Perhaps the most important demographic group in this election is also the most reliably Democratic - Black voters. In the early voting, they turned out at rates that defied predictions - 35% of voters in Georgia, 36% in Louisiana, and 26% in North Carolina. Each of those numbers far exceeds their percentage of registered voters, not to mention eclipsing all past turnout records. As a demographic group, they're voting fairly monolithically for Obama, and have put a number of reliably red states in play.

Polling, as a science, tests its models retrospectively. When voters act in unprecedented fashion, the models struggle to keep up. Pew, for example, loosened its voter screen just for black voters - it projects them at !2% of the electorate, up from 9% four years ago. (That's one reason why Pew has consistently shown some of the largest leads for Obama.) I'm sympathetic to the impulse, but applying different screens to different populations amounts to guesswork. 

The problem for pollsters is that black voters haven't been reporting attitudes that are hugely different from past elections, but they have been exhibiting radically different behavior. Since polls use the former to predict the latter, they're running into all kinds of trouble. Gallup's latest release confirms this, reporting that Black voters are more likely to have been contacted by the Obama campaign, and to express a higher likelihood of voting - but on gauges like registration and interest, Black voters and White voters report virtually identical increases. That's despite the very good objective data on black registration this cycle, from the various and several states, which shows that black registration has actually increased far out of proportion to their share of the population.  

The Gallup release also contains this priceless line:

In Gallup's most recent update on likely voters, blacks constitute 11% of both the expanded and traditional likely voter group, higher than the 8% representation in Gallup's final estimate of voters from 2004. If these trends continue through the election next Tuesday, black turnout rates this year may approach or match turnout rates among whites.

That's Gallup's own special way of confessing that, until recently, its likely voter models had been screening out many of these black voters. Now that they tell Gallup they've already voted, though, they score high enough on Gallup's models to make the cut, and their percentage of the total is rising. But here's the key point. In 2004, when Gallup's final numbers put them at 8%, black voters actually accounted for somewhere between 9% and 11% of the total. And this year, with Gallup pegging them at 11%, the write-up is projecting an increase. If they match raw turnout rates among white voters, as the release suggests, they'll account for at least 12% of the total, in line with Pew's figures. Every point that black turnout increases produces (roughly) a 0.9% swing toward Obama. So if polls assume that turnout is at 11%, and it surges to a perfectly plausible 13%, that's a 1.8% shift. And it's amplified by the concentration of black voters in states like Virginia, North Carolina, and Georgia that are suddenly in the balance, where a proportionate increase would seal an Obama win. 

2) The Keystone State: The fashionable thing to do, among angst-ridden Democrats, is to fret about Pennsylvania. Nate Silver has already done his level-best to talk these folks down off the ledge. And just as he finishes, SurveyUSA releases a new poll that has Obama up just seven points, 51-44%. Since its last poll had Obama ahead 53-41%, that's a big deal. But before you leap, you should note that this latest poll has Obama up 80-16% among black voters, down from 89-10%. So that's almost a two-point swing right there - and unless you believe black voters are abandoning Obama, you can safely discount it. McCain now does better among whites and non-college graduates. But even as he's gained ground among Republicans, and actually lost a little among Independents. And that, I think, is the key to understanding the 'tightening' in Pennsylvania. As McCain ramps up the coded appeals and the fear-mongering, he's bringing home some of his core supporters, who had wavered toward Obama or considered not voting. So his numbers are rising. But he's progressively less likely to win over moderates or independents, and is shoring up Obama's support among his own core voters, so Obama's support isn't actually shrinking. Pollster's neat graph nicely illustrates this dynamic.

Here's the essential point: There's a reason McCain waited until the final days to unleash his nastiest, most divisive message. It's a desperate ploy. He always knew that these attacks would shore up his support and increase his numbers in the polling. But his campaign also understood that to be a Pyrrhic victory - his core support isn't enough to win the state, and these attacks sink his final hope of winning over the moderate swing voters. So the undecideds - mostly conservatives and Republicans who didn't like either candidate - are coming back to McCain. But in the process, he's alienating enough voters to seal Obama's victory. In short, Obama is now less likely to win Pennsylvania by double digits, but perhaps more certain than ever of carrying the state. Ironic, no?

1) Forty-Eight Hours: The polls open in forty-eight hours, and the only question left is the margin of Obama's win. How many electoral votes will he carry? Which red states will turn blue this year? How many senators and representatives will travel to Washington on his coattails? 

Is that hubris? Presumption? Am I tempting the fates? No, not really. Perhaps a third of the votes have already been cast, and we have a fairly good idea how they break down. We've got reams and reams of polling data - this isn't 1948 - and the polls have continued right down to the wire. We don't know all of the details, but we can be certain of the headline.

There are reasons good and bad why most observers refrain from calling elections in advance. The risk of humiliation. The danger of discouraging voters from turning out, so corrosive to our democracy. The tightness of most national races, rendering them too close to call. Superstition. The thrill of suspense, and the ratings and readers it garners. 

But for better or worse, I'm calling this race. Nothing in life is certain; it's too high a standard to apply. But I can say, beyond a reasonable doubt, that Barack Obama will win the election on Tuesday. And that's good enough for me.

If you've enjoyed this, please share it with other readers by clicking the 'recommend this' link. You can find more analysis on my blog, or subscribe by clicking "Follow Me" on the right. As always, I welcome your comments and corrections, and thank you for your feedback.