Did righty Silverbating prevent them from getting over the top?
November 7, 2012 15 Comments
There’s another column I fully expect to see from all the lefty pundits, which is the idea that the righty ‘rejection of data’ and hatred for all things ‘reality-based’ is what caused them to lose the election. So I figured I’d get it out of the way by writing it myself.
Missed in the righty obsession with whether Silver’s model headline-percentage was ‘right’ is the fact that there’s a lot of other useful info to be gained from analyzing polls and likely electoral-college outcomes in the way that sort of model does. For example, on the sidebar Silver featured something he called a ‘Return on Investment Index’, which I gather measured the relative probability that one voter in a given state would determine the outcome. In my spreadsheet I had thrown in a column for something similar (which, dimensionally, should probably if nothing else differ from Silver’s ROI numbers by the states’ relative populations): a state by state ‘delta’ / leverage, or the answer to the question: if a state’s Obama polling lead increases by 0.1%, how much does that increase his probability of winning? And how much does that increase the Expected # EVs he would win? For example here’s what it looked like the last time I had calced it:
I mean, I dunno. I don’t claim these (or Nate’s) numbers are perfect. But these are the sorts of basic analytics that, y’know, I figure a campaign might want to look at. You could do very basic, obvious things with such numbers, like, I dunno, use the table to figure out where to target ad buys?
Say the Ohio leverage is 3x the Iowa leverage. Maybe that tells you to to spend 3x as much money in Ohio as Iowa. So you do. Then you look at more polls. The Ohio poll #s moved by X and the Iowa poll #s moved by Y. That tells you how much each $100k you spend moves the poll by, which (via the leverage or ROI or whatever) tells you how many ‘expected EVs’ that $100k bought you – maybe not at all, maybe some, maybe more in Iowa than Ohio, maybe vice versa – so you adjust your $cost/0.1% move ratio per state on that basis, which (via the leverage or ROI, which you are constantly recalculating/updating in the face of new data) helps you further make ad-buying decisions. Repeat. Feedback. OODA loop. Moneyball. Call it whatever you like, but there’s more to this sort of ‘whiz kid’ analysis than simply spitting out a headline Probability To Win and then arguing/whining over whether it’s ‘right’. To be clear (because I hear this criticism coming), none of this has to mean you make yourself a slave to any of these numbers. Perhaps you have some extraneous reason to believe a state is extra important, or not important at all. Fine. If you reject one of these numbers on solid extraneous reasons, or even on your ‘hunch’, then fine. You’re the boss. But it still could be an aid to have a daily report of these numbers in your hands to refer to, no?
Righties who spent all their time saying Nate Silver is wrong just might not have grasped any that.
Now, my assumption had always been that just because rank and file righties were Silverbating all the time, and the blowhard Michael Barones of the world were pulling predictions out of their butts based on “fundamentals”, didn’t mean the actual Romney campaign was being so stupid. Surely, I figured, somewhere in the Romney campaign was some 20something guy, with a spreadsheet, a spreadsheet very much like mine in fact, or perhaps (though less likely) Nate Silver’s, doing the same sorts of calculations to help inform the campaign’s decisions as I’ve just sloppily outlined above. Surely. Right?
But what this blog post presupposes is, maybe not.