demographics is a lot less interesting in politics than, say, cable news makes it. i just saw someone assert that the election will turn on the vote of blue collar white men. but let's try to get this straight. let's say you're trying to explain an election in which candidate a beats candidate b at 54 to 46% (just for simplicity).
any demographic segment that constitutes , say, 10% of the elecorate that votes largely for one of the candidates can, with exactly equal propriety, be represented as having been decisive. this could be black folks, but it could be people between 35 and 45 years old, or people with graduate degrees, soccer moms, left-handers, blondes, what you like for your little account. now probably left-handers are split. but that of course just makes them even more decisive: a potential battleground. the fact that they, or white blue collor men, are split, is no less decisive and significant than that they go one way or another. had they not been split, they would have decided the election. therefore the split decided the election. none of these can be "the" explanation because they are all to a precisely equal degree an explanation. if barack could have split off the left-handers, or people with violet eyes, he would have won. in other words, this variety of analysis flows to any taxonomy you like, and then the idea that we're detecting the categories of race, religion, age, income, education shows much more about how we split people up - about our prejudices, say - than about anything substantive about an election.
here's a paradigm case: blame nader's 1.5% of the vote in florida for gore's loss of the state in 2000. well, any 1.5 percent segment of the population that didn't go to gore is exactly equally responsible: those who voted for buchanan, for instance, or practitioners of voodoo. the idea is that had these people not voted for nader, they would have voted for gore. then again, had the voodoo practitioners not voted for buchanan, they would have voted for bush. in other words, we also make a series of presumptions on which only a vote for one of the major candidates is sensible, in which counterfactuals are cast in terms of a two-party system etc.
all of these ideas are not things we detect: they are a priori categorizations that shape narratives. the methodology, in other words, is fundamentally non-empirical, though it all comes in the quasi-scientific numbers of the polling industry.

Comments

[Rik Little]

Arent you forgetting about the Electoral College, Delagates, Super Delegates, the court system, the media who give us all "facts" which are owned by multinational corporations, and all the bribes, racketeering, uncorroborated mathematical spin machines and outright evil fascist agendas which only leave us with propositions like the Ralph Nader question to argue about? Wouldn't it be cool if after the election it was revealed to the American public that Barack was in fact White, Hillary was in fact transgender or John McCain was in fact a HAL 9000 robot? Maybe then the populace would question authority.

ok. this is my whole point: any 1.5% of the population who didn't vote for gore equally caused him to lose. for example, had the black vote been unanimous...

[Henry]

If I had lived in Florida in 2000 and favored Nader, I would nevertheless have voted for Gore because I would have known that an overwhelming majority of people would vote for Gore or Bush and that, consequently, a vote for Nader would be at best wasted and at worst would cause Gore to lose. You might reply that my "knowledge" that most people would vote for Gore or Bush was not knowledge, but a presumption. True, but one could say the same thing about my "knowledge" that the sun will rise tomorrow. Granted, if a large number of people had not presumed that an overwhelming majority of people would vote for Gore or Bush, and that, consequently, a vote for Nader would not necessarily be wasted, then Nader could have been elected. By the same token, if my grandmother had balls she could have been my grandfather.