Credit Where Credit Is Due

Throughout the latter stages of the 2008 U.S. election I became a frequent visitor to a website called Five Thirty Eight - Electoral Projections Done Right. (They take their name - 538 - from the fact that there are currently that many electoral college votes available in the presidency election.) Jimmy Hinckley was the first person to tell me about this amazing resource, although he wasn't the last, as knowledge of its existence seemed to get around. I was checking it several times a day by the time the final couple of weeks leading up to Nov 4th rolled around.

The site's authors are polling junkies, apparently, as well as statistical nerds of the highest order (and I mean that in a very positive way). They were running Monte Carlo simulations of thousands of scenarios based on the large array of polling information that was coming out daily in the run-up to Nov 4th. And their focus wasn't just on the race between Obama-Biden and McCain-Palin, but also included data on each of the Senate races (approximately 1/3 of all Senate seats end their 6-year terms every 2 years) and the Congressional seats (all of which have 2 year terms, I believe). Every day there were new observations and insights to be found there, and my growing optimism toward Obama's chances were largely driven by placing a bit of faith in what I was seeing at FiveThirtyEight.com. (Vicki was absolutely convinced that her country wouldn't elect a black man, and both of us were thrilled to have that theory be proven wrong!)

But what really blew me away was just how accurately they predicted the results of the election, especially in terms of which states would go which way, for president. I don't know if they'll keep these graphics up indefinitely, but if you click on the link above, check out the two U.S. maps at the top right of their site. The first shows the actual results, with blue states representing those that voted Obama-Biden and the red states being those that went the other way. Below it is the map with the projections by the site owners themselves, finalized on the 3rd. I believe that, out of 49 states and one district (D.C.) that they made a call on (they decided Missouri was too close to call, and it was one of the last to have a declared winner), they only got one wrong (Indiana)!! Think about how many traditional red states went blue this year, not to mention how many weren't declared until several days later (on account of being so close) and yet they nailed all but one of them! (Also: Nebraska splits its 5 electoral college votes into 3 groups, and one of those 3 went Dem whereas 538 had predicted that all 3 would go Rep).

They also did very well on projecting how the Senate seats would play out, as well as the Congressional results... all thanks to the collection of an incredible amount of polling data from around the country and the smarts to run it all through simulations to see what the most likely outcomes were. I suspect that these guys have now made a name for themselves and may be going places as a result...