One of the few recent examples of political scientists engaging with broader public debates was the disagreement of Larry Bartels with Thomas Frank over whether the non-Southern white working class had started voting for Republicans. The debate ended up being in part an argument over how best to define the working class: through looking at people’s education, income, or some combination of the two. David Brady, Benjamin Sosnaud and Steven Frenk have recently published a piece arguing that neither segmenting by education or income really captures the working class properly, and that political scientists should learn from sociologists, by adopting a measure that defines the working class as people belonging to a specific set of occupational groups. On the basis of this measure, they come up with some interesting findings. Their arguments don’t support Frank’s underlying thesis about why the male white working class has gone Republican (they don’t find evidence that religion and wedge issues played a role), but they do find contra Bartels that the effect isn’t a simple product of regionalism. In their words:
The White male working class has moved suddenly and massively towards the Republican Party since 1992. Our dating of this transition roughly coincides with Frank’s identification of the early 1990s as the point when the White working class began to dealign from the Democratic Party (Frank, 2004, 91, 98). In sharp contrast to Bartels, we contradict claims that the White working class shift towards Republicans is isolated to the South. Moreover, we demonstrate that his proxies for class are not adequate and that theoretically justifiable measures of class are essential. Ultimately, at least for men, our study supports Frank’s claim that the White working class has dealigned from the Democratic Party.
Their results also suggest that there is a significant divergence between white working class men and women – while the former are clearly more likely to support Republicans, the latter, if anything, are more likely to support Democrats. I’d be interested to see a public response from Bartels, if he’s so inclined.
(Via an email from David Glenn)
{ 7 comments }
Henry: Can you provide a reference or, even better, a link to the Brady et al. piece?
http://www.soc.duke.edu/~brady/web/wwc.pdf
I would be interested to see this same analysis broken down by regions of the country, if not states (rather than South and non-South). I wonder in which states this pattern is stronger and in which it is weaker, and then what the characteristics of those states are (policy-wise).
Based on a very quick reading:
The conceptualization and measurement of class offered here seems interesting and useful, though I’m unclear how one would know whether it is more or less useful or appropriate than other measures for addressing this or any other specific political question.
I’m not entirely sure what to make of the statistical analyses. There are no interactions between class status and time, which would be the obvious way to test the propositions about the changing role of class. The year-by-year coefficients in Table 2 bounce around a lot, and it isn’t obvious which (if any) changes are significant. There also seem to be errors in the reported results — odds ratios greater than one with negative Z-scores (though perhaps I am just confused about what the authors are reporting).
Having said that, the general empirical pattern seems to be one of off-setting effects, with white working-class men generally being more Republican and white working-class women generally being more Democratic. For example, for Model 7 in Table 3 (covering 1996-2004, the period in which the class effect is supposed to have kicked in), being a white working-class male apparently made one 17% more likely to vote Republican while being a white working-class female made one 16% more likely to vote Democratic.
If that is right, it is mystifying to me why the authors choose to interpret the evidence as supportive of Frank’s thesis. Readers have interpreted his argument in many different ways, but this is the first time I’ve heard anyone suggest that it should be applied specifically to white working-class men but not to white working-class women (or, at least, that movement in a Republican direction among men should be interpreted as support for the thesis without regard to off-setting movement in a Democratic direction among women).
More broadly, for what it is worth, what seems most interesting to me about Frank’s argument is not the claim that working-class white voters (however defined) have become more likely to vote Republican, but the claim that they have done so specifically in response to Republican appeals to cultural conservatism. The analysis presented here isn’t really designed to shed light on that proposition, but insofar as it does the evidence seems to be to the contrary.
P.S. I just noticed that Henry’s link leads to the original, unpublished version of my paper. The version published in the Quarterly Journal of Political Science (also posted on my Princeton webpage) was extensively revised in response to Frank’s critique. Oddly, that fact seems to have escaped many people (including, most recently, Michael Tomasky in the New York Review of Books).
So we need a definition of “working class” that can be broken down into racial categories, as well as regionally. Should then “working class” be compared/contrasted to/with “middle class” (which would also require definition) to determine whether the former is included, partially or wholly, in the latter? If only partially, what “class” might the remainder fall into?
I read the piece last night. Two things bothered me:
1. While the authors were quite explicit (in the appendix) as to what occupations they included in both the “broad working class” and “restricted working class”, I was left cold by the theoretical arguments upon which their criteria were based. Perhaps it’s simply that I’m not a sociologist.
2. No mention (at least that I could find) was made of the size, relative to the elctorate as a whole, of either definition of the working class. My assumption is that, using both definitions of the working class, the number has decreased as a percentage of the total electorate. It is almost certainly true with the more restrictive definition. To me, the ever-decreasing impact of the “working class” as an electoral group is at least as interesting as shifts in their voting patterns.
Comments on this entry are closed.