The Naked Dollar has speculated that there is no one in America who voted Republican four years ago who will be switching their vote to Obama this time. I've asked over 100 people if they know of such a person, and the answer is always no. If you DO know anyone like this, by all means leave a comment, and describe their motivation as best you can.
Of course, we all know lots of people going the other way. Obama's approval numbers are down substantially with every one of his major support constituencies.
Obama won by 6 1/2 points. This means that if 3 1/2 voters out of 100 change their vote, Romney wins. Hard to imagine that won't happen, but leave that for a moment.
For fun, let's assume no one changes their vote. Obama wins by 6 1/2, right?
No. This assumes the makeup of the electorate is static, and it never is. Simply, the demographics of every election are different. So, here's a statistical experiment: how will the election turn out if no one changes their vote, but the people who show up, as broken down by party affiliation, are the same as 2010, not 2008.
It took a while to find the data I needed, but eventually I dug it up. The results of our hypothetical election: Obama 51.7%, Romney 48.3%. Margin: 3.4%. Now were down to fewer than 2 people who need to switch their votes.
The problem for Obama is that lots of people are likely to switch their votes. Just look what happened to the independent vote in '10. Obama won independents 56-44, but in 2010 the outcome flipped to 44-56 (for Republicans). If there's a result close to this in 2012, Romney wins with relative ease. Current polling suggests that independents will go for Romney, and, further, Republicans are far more fired up than Dems to vote.
Note: this analysis is only of the popular vote, which, of course, is not what counts. Interestingly, it looks like Romney may have a slightly tougher time with the Electoral College than with the popular vote. It's not beyond imagination that he wins one and loses the other.