Election Probability

Ok, I know I'm a geek. After all, my first degree was in Mathematics. Which both makes me trust and distrust statistics. I guess I trust them way more than intuition, but am well-aware of their limitations. I think this election cycle is far less predictable than usual making polling even more problematic. The primary causes for the polls increased level of unpredictability: (1) the increased voter turnout runs ramrod over previous models of who are "likely" voters causing the polling institutions to grasp at straws to define this, (2) the Bradley effect and (3) perhaps related to the Bradley effect, there is a reasonable theory that many undecided voters just don't want to admit that they're not voting for the black guy and will disproportionately break for McCain. The completely detached, logical side of me will be watching election night results for purely academic reasons while the other part of me is having a complete emotional breakdown.

Anyway, I came across the 538 website, which I've never read before. It definitely feeds that academic side of me. If interested, go to their site and look at the very interesting graphs and maps. I've pulled a couple of them out so that I can make comments.

1. This first one is obviously the result of simulations of election results. Compare how many read bars versus blue and how small those read bars are. Those red bars represent the likelihood of an Obama loss appear really small. Now for the bad news, simulations like this necessarily must make models imitating how voters think and behave. If you think polling in this election cycle is very difficult, making good simulations is impossible.

2. This second graph is presumably based on the same set of simulations as above and of course should not be trusted for the same reasons. But for fun, the findings are astounding.

• Chances the McCain will win popular vote: just 5.48%!
• In 2000 and 2004, Bush's path to the White House was to win Ohio and Florida. The chance that even if McCain can win both of these (he's currently trailing in both) that Obama will still win: 73.7%! Damn!