In the Washington Post, Michael Gerson writes a rant against Nate Silver. He also takes a lot of potshots at political science, too.
The current mania for measurement is a pale reflection of modern political science. Crack open most political science journals and you’ll find a profusion of numbers and formulas more suited to the study of physics. In my old field of speechwriting, political scientists sometimes do content analysis by counting the recurrence of certain words — as though leadership could be decoded by totaling the number of times Franklin Roosevelt said “feah” or George W. Bush said “freedom.”
There is a certain amount of political science that is overly wed to a quantitative method. I remember one paper that utilized a three page long methodology and utilized an army of grad students as coders. Its main finding was that the Daily Kos had a liberal bent. No, really.
By being overly wed to questions that can be answered with quantification, major questions in political science go unanswered. "Who benefits?" is the basic question in political science. Sometimes that answer requires going into the field and talking to people.
I think that not enough political scientists look at their projects with a critical eye. "Who gives a shit?" should be a question that some academics should ask more often.
But that's neither here nor there, because that's not really what Gerson is saying. He's just saying that we shouldn't listen to Nate Silver with his bad news, because numbers are bad. Now, that's just silly.
Apt. 11d 
That first Gerson piece is a horrible mess. It’s like he put two different editorials in a text randomizer. He shifts from political science to political punditry applying criticisms of one to the other. Also, he doesn’t understand the role of statistics in either the social or natural sciences, but that isn’t exactly rare.
LikeLike
If the assumptions contained in those polls — on the partisan composition of the electorate or the behavior of independents — are wrong, it is the failure of pollsters, not of statisticians such as Silver. Note to recent college graduates: Strongly consider a profession in which one is right, by definition, 100 percent of the time. It beats poker.
This argument correctly points out that if the polls are wrong, that doesn’t mean Silver is wrong. But it fails to recognize how Silver could be wrong. In particular, if Silver is predicting that there is an 80% likelihood of things happening, and they happen 100% of the time, then he is “wrong” even if the stuff that he predicts actually happen.
As a baseball fan, I feel like I read all of these arguments 10 years ago, substituting “the Oakland A’s” for “Barack Obama.”
LikeLike
The Gerson piece was a hot mess (and not helped by the general perception that it was written only because of the polarity of Silver’s predictions). But, I think there are some meaningful questions to ask about the current interest in polling predictions.
The first is the concern that the predictions actually influence outcomes. There’s strong behavioral evidence that they could. I don’t have a solution to that concern, because, really, forbidding Silver from thinking and writing about what he thinks about the election is pretty much constitutionally impossible. We should talk about it, though.
The second is the concern that the polling can be used for evil, if it accurately predict outcomes. For example, after the 1st debate, Laura opined “meh” but then didn’t think it would influence people. The polls suggested that the debate did influence the polls — that they were picking up something (though we can’t be sure of the mechanism). What about when polling can accurately predict the success of an campaign and the ads it produces? What if polls effectively predict the success of straightforward lies? The ability to use polls to predict success could change the way the election game is played in destructive directions (of course, lots of people think that’s happening already, but increasing accuracy of polls will make the problem worse).
The science here is essentially amoral, but that does mean it can be used for evil.
LikeLike
Gerson on 11/5:
On the eve of the election, Nate Silver — baseball forecaster, online poker wiz, political handicapper — placed President Obama’s chances of returning to office at 86.3 percent. Not 86.1 percent. Not 87.8 percent. At 86.3 percent
Jonah Goldberg on 11/7
On any given day, Silver might have announced that — given the new polling data — “the model” was now finding that the president had an 86.3 percent chance of winning. Not 86.4 percent, you fools. Not 86.1 percent, you Philistines. But 86.3 percent, you lovers of reason.
http://www.nationalreview.com/articles/332766/critique-nate-silver-s-pure-reason-jonah-goldberg
Which is worse — writing your own mindless garbage, or mindlessly regurgitating someone else’s?
LikeLike