Tuesday, November 6, 2012

Reply hazy, try again later

There's a new game in town: hold the pundits accountable for their election predictions. Various pundits are now busy compiling lists of predictions of other pundits so, I guess, they can do a "neener, neener, neener" tomorrow morning. Here are a few of them:

Ezra Klein at Wapo.

Flora Zhang at CNN.

Adam Pasick at New York Magazine.

Pasick's list is particularly complete and easy to manage. But the truth of the matter is that no pundits will actually be held "accountable." How could they be? They're just offering their best guess (or prayer). Still, some will look and feel foolish tomorrow, at least for a few days, particularly those who predicted landslides. Dick Morris predicted a Romney win by over 100 Electoral Votes. If Obama wins--by any margin--Morris will be hard-pressed to explain his fantasy. By the same token, Jim Kramer of CNBC has predicted an Obama win by over 300 Electoral Votes, assuming I guess that Obama will win Texas and Alabama (not really). He stands to look foolish even if Obama wins.

But the meat and potatoes guy for these prediction is now Nate Silver at the New York Times FiveThirtyEight blog. He had it at 332 Obama, 206 Romney at the time Klein wrote his piece. And that prediction was echoed by many other pundits on the Left--like Markos Moulitsas at the DailyKos--though Silver has since scaled it back to 313 Obama, 225 Romney, with a 90% chance of an Obama victory. Silver also predicts a 2% margin of victory in the popular vote for Obama.

I like Silver's blog; I read it regularly and think he is quite gifted at statistical analysis, so I think it's quite possible things go exactly as he predicts. That said, neither Silver nor any other pundit has a crystal ball. And the art of polling has become--truly--an art. The objective truth of any finding in any political poll is open to debate, as pollsters have become overly concerned with their own internals in order to justify smaller and smaller sample sizes.

And there are a lot of holes in their methodologies, not the least of which is the existence of people like me who make it a point to never respond to a poll. My existence automatically increases the margin of error of every poll that attempted to solicit my opinion. Not by much, true, but then I'm not alone in this regard. I know that too. How many--what percentage of the population--are there out there who follow the same path? Who knows? Because there are also people who may respond to one or two polls, but then get annoyed and balk at all future polls.

Other problems inherent in their methodologies include multiple assumptions about demographics, assumptions that are necessary but still represent "best guess" situations. But my intention isn't to run down polling, to argue that they are all wrong or that even particular ones are. My point is that predictions based on polling numbers--like Silver's--may be perfectly accurate with regard to the data in those polls, but the margins of error in the polls may not be anywhere close to what they are assumed to be, meaning such predictions are not actually based in reality. If the margins of error are just a tad larger than what is assumed, the polls are practically useless, since the numbers are as close as they are. There's nothing dishonest or underhanded about it; it's one of those things many--on both sides of the aisle--just don't want to accept.

What does all of this mean for the Election? It means that a Magic Eight Ball may be as good a predictor as anything else:

Who will win today's Presidential Election?

Cheers, all. And don't forget to vote.

No comments:

Post a Comment