Sunday, November 04, 2012

Lies, Damn Lies, and Nate Silver

I somehow managed to spend way much time this morning reading about the recent flap over the New York Times blogger and poll-geek Nate Silver's prediction that Obama has a 77.4 percent chance of winning the election on Tuesday. This, predictably, has raised the ire of numerous commentators (mostly those in the Romney camp) who called into question the validity of the number. And that, in turn, has raised the ire of Silver defenders, who have taken pot shots back at the pot shot takers and accused them of many egregious things, such as not being good at math.

Paul F. Campos's piece in Slate is a good example of the "learn your math, dummy" counter-reaction, and he closes it by raising the spectre of the Monty Hall Problem, an old counter-intuitive chestnut. I hardly think this is fair, since the Monty Hall Problem is a sort of optical illusion of logic that fools all sorts of smart people the first time they tackle it, and it's hardly a good litmus test of someone's basic statistical literacy. The statistical concept here is a much more fundamental one.

It's a fun sort of thing to while away a Sunday morning pondering, but in the end, does Silver's prediction even really matter? Should Romney supporters take a look at the numbers and, disheartened, fold up the tent?

Of course not. What Silver's model is saying is, based upon the way the poll numbers look right now and past performance of those polls, if you ran 100 elections, Obama would win 77 of them. But that also means Romney would win 23, so hope is by no means lost. If your favorite football team was down one touchdown going into the 4th quarter and you knew in similar situations the trailing team lost 77% of the time, would you leave the stadium and quit cheering? Of course you wouldn't.

And, what's more, those polls can still continue to change over the last few days leading up to the election. How much of a move in the polls themselves does it take to adjust that 77.4% likelihood down to 65%? To 55%? To a coin flip?

To my mind, Silver's predictions shouldn't change the behavior of Obama or Romney supporters one way or another. If you are in the Obama camp, you wouldn't want to take the foot off the gas, any more than either of those football teams wants to slack off in the 4th quarter. If you're in the Obama camp, it means this election is by no means a sure thing. Better buckle down and get that number close to 100%. If you're in the Romney camp, it means it isn't over yet. One strong last minute drive and those numbers can totally change around . . .

Hmm . . . I just checked back at Nate Silver's blog. The morning he updated Obama's chances to 85.1%. Not looking too good for the trailing team . . .

In any event, come Tuesday, I predict I am 100% likely to be glad the whole thing's over.


No comments:

Popular Posts