Replies

The problem with it is that it assumes an unpolarized electorate. In other words, the probabilities in the range of 50-60% in favor of Bush are artificially low when weighted in this manner. In a polarized electorate (even if only slightly so), probabilities tend to truncate, not distribute evenly as the math here assumes.

Could you please explain what you just said in English?

;^)

Let me try it this way, and perhaps some semantics explanation is in order. When I think of polarization I'm cheating because I'm thinking of it in two different ways. One way to think of it, as most people probably would, is that it means that a population is evenly divided between 2 or more choices. But another way to think of it is in terms of voter commitment: How strongly do those polling in favor of Bush or Kerry feel about their choice, or, what are the odds that their choice today will be reflected on November 2.
Let's examine the first definition. Assume you have a highly polarized electorate BUT there is a small percentage of undecideds. In most elections, even though the general population may be highly polarized, the undecideds are NOT. They usually end up disproportionately favoring one candidate over the other. This is precisely why they are undecided in the first place: they are malleable to that candidate which speaks best to the middle. One or the other candidate will usually win out in that regard. This, right now, favors Bush and the probabilities should be weighted to reflect this; i.e. they should be truncated, not evenly distributed between 0 and 100%.
But the other sense of polarized is much more telling. If you show today that 51% of the electorate in Ohio will vote for Bush, hypothetically, then it is NOT necessarily the case that Bush has a 51% probability of winning Ohio. Why? Thinking up an ideal, extreme example may help you see this. In reality, it's not that extreme, but the extreme example can help us see the in-between reality. Sooo, assume Ohioans are absolutely committed to Bush: they'll never change their minds no matter what happens between now and November 2 (many voters are indeed like this nowadays). Then the odds of Bush carrying Ohio are exactly 100%!!!, not 51%. In other words, the probabilities have truncated. In reality, since all Bush supporters are not like that, the true probability lies somewhere between 51 and 100%.
What the data you have is actually doing is assuming that the polls can be infinitely in error. In reality, they only have about a +/-4 percentage point error range. Consequently, the weightings are just weightings of how much we believe polls, not who's going to win.

In reality, since all Bush supporters are not like that, the true probability lies somewhere between 51 and 100%.

In the interests of fairness and completeness, I should add that this same analysis applies to the Kerry side. But really what is happening is that the candidates are 'trading off' voters from each other's solid camp. This 'trading' follows the same rules as it does for undecideds; i.e. it will usually be disproportionate (and right now favors Bush).