Friday, November 02, 2012

Election 2012: Joe Scarborough Fails Statistics Forever

On August 6th of this year, the Mars Science Laboratory touched down in Gale Crater. It was a feat of celestial navigation that's hardly had any equal. Not only because it was able to hit such a small target after flying three hundred fifty million miles, but because of this:

Entry into Mars' atmosphere had been timed with such exquisite precision that the Mars Reconnaissance Orbiter was in position to take a picture of Curiosity floating down under its parachute. Modern guidance and navigation techniques are sufficiently precise that even after a flight of ten months and hundreds of millions of miles, one spaceship can look over its shoulder and snap a photo of another. Part of the mathematical tool kit that makes it all possible is Bayesian analysis, a statistical method for reducing unknown sources of error. The same kind of analysis can be performed on any data, including statistical data, and Nate Silver does this with polling data over at FiveThirtyEight ... and, for what it's worth, it's a mathematical tool kit that Joe Scarborough has gone on public record describing as "a joke."

Mind you, I enjoy his program greatly, and my wife watches it every morning. But Joe, if you don't understand what Bayesian analysis is or how it works, could you kindly shut the Hell up about it?

We're down to the last four days, my friends. It's the two-minute warning. Either man could win at this point, obviously, but the probabilities are beginning to narrow down. Nate Silver's analogy of Barack Obama being ahead by a field goal is an apt one. If he can get a first down, he can run out the clock. About all Romney can hope for is for him to fumble, or throw a pick-six. And it's important to note that either of those two things could happen. It's all coming down to a handful of states, and even a handful of counties within those states. With fewer than a hundred hours to go, every second counts.

And now, over to the part of the program that Joe hates: the numbers. As usual, my data sources are Intrade, FiveThirtyEight, and Pollster. Incidentally, I'll answer a question, in case you're curious: why those three? I like Intrade, because it captures that whole "wisdom of crowds" thing. And it was freakishly accurate last time. Until it gets something important horribly wrong, it's in the mix. And I like FiveThirtyEight, because it uses techniques I learned during my time as a GN&C weenie in graduate school. Pollster, on the other hand, uses a more traditional "poll of polls" approach. If three different methods reach more or less the same answer, then it's probably a pretty dependable answer. Anyway, for the last time this season, here they are, current as of Friday afternoon.

From Intrade:

Barack Obama (D): 66.9%, 290 EV (+3.1%, +14 EV)
Mitt Romney (R): 33.1%, 244 EV (-3.2%, -14 EV)

From FiveThirtyEight:

Barack Obama (D): 81.1%, 303.3 EV (+6.7%, +7.9 EV)
Mitt Romney (R): 18.9%, 234.7 EV (-6.7%, -7.9 EV)

From Pollster:

Strong D: 237 EV (+/- 0)
Lean D: 44 EV (+4)
Tossup: 66 (+11)
Lean R: 0 (-15)
Strong R: 191 (+/- 0)

There's no good news here for a Republican. The closest thing to good news for Romney is that he's got a firm floor of 191. That hasn't budged in months. But his "lean" support has bounced into and out of the "toss-up" column all along. Now, I think that a fair bit of that will end up in his court. But from that firm floor of 191, he needs all of the toss-ups, plus at least 13 from Obama's "leaning" column. And I don't think he's going to get all of the toss-ups. Some of those are going to bounce Obama's way, and that's all gravy anyway, since if all his "strong" and "lean" support proves out, that's all he needs.

What Mitt Romney Must Do: He's got to get every last one of his people to the polls. Every. Last. One. And even that might not be enough. But that's what it comes down to, now. His campaign has to execute, and get their people in to vote.

What Barack Obama Must Do: See above, except that he's working from a slightly better position. As I said earlier, all he needs is a first down and he can run out the clock. His main enemy now isn't Romney, it's Complacency. He can still snatch defeat from the jaws of victory, if he works at it. Execution, and a strong ground game, those are the keys.

And The Winner Is... Unemployment is down. The housing market is entering a real recovery. Gas prices are stable. Incumbent Presidents almost never lose unless the economy is totally dead, the international situation has gone south, or both. While the economy isn't what anyone would call good, it's better now than it has been, and that's good enough for the advantages of incumbency to work their magic. So, I'm going to make my semi-official prediction: Barack Obama wins re-election, with 303 votes in the Electoral College. The popular vote will probably be 51-49, or something very close to it. It's a closer race than last time, and Mitt Romney has run a pretty good race for someone who has always come across as a collaboration between the MIT Artificial Intelligence Lab and Disney's Animatronics. But this was always Obama's race to lose, and in that first debate, he sure gave it the old college try. If you can still find someone willing to take the bet, I'd take 2-1 odds, and I'd take 300 for the over-under in the Electoral College.

It's too late now to vote early. But, it's never too late to vote often! I'll check in Tuesday night, and see how things turn out.

No comments: