Broadly speaking, I share Matt's model of the electorate:
These are the sort of margins that suggest anything could've been the culprit. A couple hundred votes in Florida, a couple thousand in Ohio, and Democrats win. The losses weren't the clear consequences of a great economy or a popular Republican leader or a particular confluence of circumstances. Indeed, given the backdrop of 9/11, the elections have actually been substantially closer than one might expect. But losses require some sort of change, and so what's happened is that the call for change has also come on the margins: The substantive beliefs of a party are protected by the powerful, but style can be thrown under the bus. So things like a candidate's penchant for pedantry, or his "elitist" bearing, became the acceptable culprits. The ideas remained basically the same. By 2008, however, the policy environment had shifted dramatically, and the Democrats had been out of power for almost a decade, and so the base elected someone who represented a more substantial break with the past. If Obama racks up a big win, however, it will be largely because the underlying factors (shitty economy, Bush fatigue) moved the core electorate, not because stylistic factors (great speeches!) moved the marginal electorate.
the majority of voters are voting as blind partisans. Of the rest, most are being driven by the macro factors (shitty economy, sick of Bush) or purely by issue salience (vote Republican when I care about national security, vote Democratic when I care about the economy) or other such things. And yet, few people like to say that kind of thing. And this is where the campaign comes in.The main impact of campaign attacks, I think, is not to actually change anyone's mind but rather to familiarize everyone with the talking points of the side they agree with. In 2000, voters who valued "experience" turned out to favor Al Gore strongly. In the 2008 campaign, I think it's clear that voters who value "experience" will favor John McCain. That's not, however, because there's some coherent bloc of "experience" voters who shifted loyalties -- it's because "experience" was a Democratic talking point in 2000 and it's a Republican talking point in 2008 so people change which candidate attributes they value. In 2004, you could find a lot of Democrats who thought John Kerry military service proved important things about his fitness for office, whereas in 2008 Republicans are more likely to say that about John McCain.It's easy enough to complicate that picture, but broadly speaking, that's about right, and about what the social science suggests. What's odd about the last decade or so in American politics has been that elections have been decided on the margins. As this quickie graph shows, the 2000 and 2004 elections were extremely close by recent historical standards. And Democrats actually won one of them:
These are the sort of margins that suggest anything could've been the culprit. A couple hundred votes in Florida, a couple thousand in Ohio, and Democrats win. The losses weren't the clear consequences of a great economy or a popular Republican leader or a particular confluence of circumstances. Indeed, given the backdrop of 9/11, the elections have actually been substantially closer than one might expect. But losses require some sort of change, and so what's happened is that the call for change has also come on the margins: The substantive beliefs of a party are protected by the powerful, but style can be thrown under the bus. So things like a candidate's penchant for pedantry, or his "elitist" bearing, became the acceptable culprits. The ideas remained basically the same. By 2008, however, the policy environment had shifted dramatically, and the Democrats had been out of power for almost a decade, and so the base elected someone who represented a more substantial break with the past. If Obama racks up a big win, however, it will be largely because the underlying factors (shitty economy, Bush fatigue) moved the core electorate, not because stylistic factors (great speeches!) moved the marginal electorate. 
