For All You Numbers Geeks

I came across in interesting blog the other day, www.fivethirtyeight.com. It's run by a statistician and avowed Obama supporter, but the analysis is not biased. Instead, the anonymous publisher of the site weights and aggregates other published polls, then runs the numbers 10,000 times to get a probabilistic take on election outcomes. His method was quite successful in predicting the results of the Indiana and North Carolina primaries, with numbers more accurate than most of the "big" polls. On the site, there is an interesting model involving youth and minority turnout. You should check it out yourself, but the basic premise is that if Obama can increase significantly the youth and black turnout--supposedly his strength--he can trounce McCain in the general election. If he can't, if the turnout in 2008 mirrors the turnout in 2004, then Obama will likely lose the general election.