The election in 2000 was statistically exactly balanced. 2004 was pretty close, too. At the time, I guessed that some trend must be driving the elections to be so close. Here is an article from Slate that supports that idea. There are a few links at the bottom to others with the same thought, including a more interesting one from David Levy. Turns out it was first formulated in 1929.
It's an example of experimental mathematics.