> Seen in this light, the breakdown of voting preferences would
> look considerably different, and would probably fit in more
> with Bartels more than Franks. My impression is the "new
> working class", big users of the internet and heavily
> represented on lists like this one, identifies more with the
> liberal left than with the Republicans.
I do not think this is right. Using broad catch-all categories like "working class" always carries the danger of over-generalization, especially when diversification seems to carry the day. I think such a broadly defined concept of class misses or misrepresents tremendous diversity that really matters politically: occupational identity (cf. Michael Hoover's excellent posting on the 'economist class', or more broadly, Pierre Bourdieu, _Homo Academicus_), regional culture (cf. midwest vs, northeast), urban/suburban/rural divide, union membership, gender gap, ethnic divisions and identities or even cognitive predisposition (the Lakoff's authoritarian father vs nurturing parent thing).
My problem with this whole line of argument is that it centers on semantics instead of the substance. The substance is that the majority of the US population has repeatedly expressed their electoral preference for conservative-leaning politicians, and that left-leaning candidates do not even lose honorably - they fail to gain support that would put them over the margin of statistical error which is 5 percent (it is a figure of speech, the concept of statistical error does not apply to populations). Nader, who is not particularly left-of-the-center got less than half percent of the general popular vote, Kucinich that might qualify as a leftist, got less than 4 percent in the primaries - i.e. among those who are already supposed to be leaning towards the "left." The matter of fact is that the majority of the US population leans to the right, or as Doug says, it is confused and when it doubt it leans to the right. Whether they are classified as 'working class' or some other class - albeit I am at loss thinking what that other class would be (certainly not the capitalist/investor class) - seems to me like a semantic argument that is emotionally charged but does not explain much.
Speaking more broadly, attempts to predict voting behavior from broadly defined socio-demographic characteristics of individuals seem overly deterministic to my taste. Such predictions might have worked in the past, when knowledge, norms, and expectations were transmitted mainly by social channels that were defined by these socio-demographic characteristics, i.e. we could meaningfully talk about "Irish American consciousness" Italian American consciousness, "Polish American consciousness" "blue collar worker in New York consciousness" etc. Then there was the "machine politics" -often intertwined with socio-economic and ethnic identities that directly influenced voting. Today, however, the importance those channels has been greatly reduced and it is practically confined to narrow niches. The majority of the population, at least those who vote, do not seem to be overly (if at all) affected by them. Therefore, socio-economic factors that tended to define them are poor predictors of what people think or do politically.
I think what influences people the most today are institutions, especially organized religion (at least in this country), government, the media, and organized business, and to a somewhat lesser extent educational and cultural institutions, civic organizations like unions, environmental groups, political interest groups, etc. Therefore, institutional influences seem to be better predictors of electoral outcomes and individual socio-demographic characteristics. Since most of the influential institutions in the US tilts to the right, and there is a real dearth of left-leaning institutions, especially among those most influential ones - that explains why the electoral outcomes also tilt to the right. If we had more left-winning institutions like Europe does (e.g. national unions, national Labor Party, etc.) then the left-winning candidates would have a much greater chance in electoral politics, instead of being swiped under the margin of error.
Consequently, I find little comfort if someone manages to demonstrate that statistically speaking, the group arbitrarily defined as 'working class' has a somewhat greater probability of voting on a particular type of a political candidate. Such arguments tend to beg the question 'so friggin' what?' The point is not to explain in a comfortable way but to change. Knowing the class status of the voters will do little to change their voting behavior - that change can be produced only by a massive expenditure of financial and cultural resources (from research programmes to massive media campaigns) that only institutions have. Therefore, the real issue is not studying individual voters to death, but building left-leaning institutions. "If you build, they will come" and vote for political candidates anointed by these left-leaning institutions. I once thought that founding the Labor Party was an effort in that direction and eagerly became a card-carrying member, until it turned out be yet another flop of the US trade unionism.
Wojtek