Wednesday, May 13, 2009

There's Sweaty, Abject Flailing -- And Then There's This

A quick timeline:

- Mickey, in the throes of who-put-what-in-which-aging-hippie-when-and-who-knew-about-it passion, tries to connect his tawdry fascination to some actually relevant topic by speculating without any evidence (and *against* Nate Silver, which is dangerous territory for Mickey) that Edwards' presence in the primary hurt Hillary Clinton.

- Using charts and math and stuff, others point out that this is crazily unlikely.

- Mickey then updates:

5/11 Update: Mark Blumenthal's calculation today is pretty thorough. His New Hampshire numbers seem especially devastating to the idea that Hillary would have benefitted from Edwards' absence. But, as Blumenthal notes, you can never respond conclusively to a conjecture that 'the whole dynamic of the race would have changed.' ... You could also speculate that Edwards' N.H. supporters lied to the pollsters Blumenthal cites--i.e. they were really non-black voters of the sort who would never have voted for Obama. (If only there were a name for this "effect.") ... I'd still be interested in what Nate SIlver's model shows--if I remember, it assumed that voters ethnicity (along with other demographic factors) was hugely predictive--suggesting Hillary might have picked up a lot of Edwards' white support in early primaries, no matter what a) "second choice" polls showed or b) what those voters told pollsters later in the race when Edwards finally dropped out. ...

Let's unpack:

(1) Because nobody can ever respond conclusively to conjecture, all counterfactuals are valid.

So, even though there is no evidence to support Mickey's counterfactual, there could never be *conclusive* evidence support *any* counterfactual (absent a DeLorean and a flux capacitor), therefore Mickey's is just as plausible as any other? Works for me!

(2) Mickey's counterfactual is plausible because of ... wait for it ... the Bradley Effect! Which caused people to lie to pollsters about their *second choice*! When they've already said they are voting for the white guy! But they then elevate the black guy over the white woman, so the pollster doesn't think they're racist!

It's a good thing that the Bradley effect hasn't been largely debunked over and over by the very expert cited repeatedly by Mickey in this post. Otherwise, Mickey might look foolish here.

[Say, wouldn't the Reverse Bradley Effect have *further* boosted Obama in Iowa and the other caucus states absent Edwards? Shouldn't that factor into Mickey's calculations? -- ed. If Mickey has calculations beyond "Outdated Assumptions Of White Guilt + Arrow's Theorem = Anything Is Possible!!!", I'll eat a stuffed moose.]

And finally:

(3) Nate Silver also might have kinda maybe said something about ethnicity or demographics or something predicting voting patterns -- who has time to look it up, anyway? -- so whites vote for whites, right?

Except, you know, Silver clearly used his model to interpret polling data rather than to ignore it entirely. So close!


But, seriously, the way Elizabeth Edwards avoided admitting mistakes, shamelessly cherry-picked evidence, and spun facts to her advantage was truly ridiculous.