Tuesday, August 5, 2008

Princeton Election Consortium

This is likely only going to interest hardcore election nerds, like myself, but here goes:

FiveThirtyEight kinda sucks. For those that don't recognize the name, it's an electoral projection site that tries to predict future election results based on a complex model constructed from poll numbers(state and national), demographics, various arbitrary weights, and Nate Silver's blood glucose level each Tuesday. OK, OK... that last one isn't true... but it might as well be. He has constructed an incredibly overwrought model that is nearly impossible to interpret... it could be be the Wizard of Oz, for all its black box-ishness... which is why I don't link to him, despite the fact that it seems like something I'd be all over. I do like his commentary, but I find his model ponderous, and thus don't use his stuff to try and support/knock down any arguments.

Well, lo and behold, Kevin Drum alerts me that there is someone else who shares my qualms(plus more!)... and he knows statistics... and he's doing his own meta-analysis!

I will warn you, he's not doing "predictions"... he's doing "what if the election was held today?" musings, but with statistical rigor. He's also entirely dependent on state polling, which, right now, is somewhat sparse... though this will turn from a weakness to a strength as we approach the election.

Regardless, Princeton Election Consortium gets my meaningless endorsement as THE place to go if you want to know the state of the race, but you want more than "he's winning".

No comments:

Post a Comment