Let me point out an observation about the election results: you can fly 3,000 miles from northern Idaho to south Florida and not cross a single county that went for Kerry.
I love to admit I’m wrong. That means I’m the kind of person who can learn --- and I do learn. My prediction of a landslide for Bush was wrong.
I’ve been forced to look at the voting data carefully and understand why I was wrong.
There is one incidental and relevant fact that I didn’t comment on. It turns out the popular majority is a poor predictor of the winner because the Electoral College has been changing over 14 years as our demographics change. We now know that a victor from the pure urban states needs greater than a 1% popular victory to win and a victor from the rural states needs a greater than 3% popular victory to win the Electoral College. Within that 4% point spread the Electoral College is a random outcome generator.
My predicted landslide did not happen. The reason relates to a totally new phenomenon that I didn’t take into consideration: 527s. Five-twenty-sevens are the new McCain-Feingold political vehicles that spent nearly half a billion dollars to get out the vote. The Republican 527s succeeded in getting 20% more votes for Bush in 2004 than in 2000 (over 9 million additional votes). The Democrat 527s did nearly as well.
What is the effect of 527s? The effect is to blur the outcome, to create a closer race without changing the popular voting results.
This blurring effect is visible when you watch a vote count where provisional votes are added to an existing vote total or where absentee votes are added to a vote total. Almost always the added votes merely make the resulting totals closer. There are rare exceptions, very rare.
The exceptions are interesting and to the point. The reason added votes don’t change an outcome is because the additional votes are from the same population as the regular voters and don’t have a significantly different viewpoint.
There are two exceptions that I know of. One is where an unusual last minute event effects the election results (the absentees weren't exposed to the opinion changing event) and the other is where the absentee voters have a uniquely different experience such as absentee military voters in Iraq or Americans in Israel voting absentee.
My Bush landslide prediction was wrong. I now understand that the underlying data was correct but it was washed out by the massive 20% extra voter turnout caused by the novel introduction of 527s that brought along a half a billion dollars of voter turnout money.