Yes, at heart I’m a political race-horse junky, but — like my fig leaf subscription to Entertainment Weekly masking my inner celeb gossip whore — I read only good political race-horse coverage like FiveThirtyEight. Nate Silver gives great chart, you’ve never heard better statistical pillow talk and his naked wonkery ain’t bad either. In other words, I subscribe to read the articles. No really.
Anyway, I wanted to point out something Nate said that I think has broader application:
Making things candidate neutral is partially a marketing decision — I can’t imagine how much yelling and screaming there would be if I said “let’s give McCain 2 bonus points because he’s a Republican” or “let’s give Obama 2 bonus points because the economy stinks”.
In his model for predicting the outcome of the November election using a mix of polling data and historical regressions, he doesn’t have candidate specific variables. He does this so he doesn’t, or I should say his methodology doesn’t ((He’s made it pretty clear he’s supporting Obama.)), appear biased.
He does this at the expense of biasing his predictions. This is because we know Democrats, for example, tend to have biased high polling numbers or the incumbent party polls poorly relative to what actually happens on election day.
Forgetting for a second that Nate seems to be throwing out variables that would improve bias in McCains favor and ignoring the fact he’s leaving out variables that favor Obama that don’t appear to be “candidate specific” (e.g. why is the health of the economy candidate specific?), I think this “unbiased appearance” bias is quite common in social sciences. Its just the ommitted variable bias where the omitted variable is purposefully omitted for “marketing” reasons.
IQ isn’t one of the Mincerian variables, for example, because it would appear biased to include it in wage regressions. This is strange given we’re just trying to figure out what causes variation in wages.