Paul Krugman posts a telling graphic from Gelman, et. al. showing that the correlation between income and voting behavior has generally grown stronger over time (the dot is the correlation, the whiskers are the range of uncertainty):

incomevoting.png


Krugman comments that "the conventional pundit wisdom about the relationship between class and voting" -- namely that there's less class polarization than there used to be "is, literally, the opposite of the truth." The difficulty is that there's a lot of ambiguity about how we should define class. Fortunately, the best article on this controversy was written by me. Krugman, following Larry Bartels, wants to define the "white working class" as being composed of white people in the bottom third of the income distribution (which, note, is considerably less than one third of all white people). Dissenters from this view make some good points:

Gopoian and Whitehead point out that “only one-third of the Bartels voters were actively doing paid work,” a fact that undermines the “working” half of the working-class label. What's more, “of those who were working, nearly half were under the age of 30,” a category that would include such non-obvious members as several 20-something Ivy League–educated members of the Prospect's staff.



In short, the low-income whites who Bartels finds to be strong backers of the Democratic Party have a marked tendency to be retirees or students and even those who are working tend to be very young. The alternative definition of "white working class" is "white people who don't have a bachelor's degree." Under that definition of white working class, the white working class does, indeed, support the Republican Party. However:

The education-based definition of the working class comes with problems of its own. Using the education criterion, almost two-thirds of white voters, and a significantly larger portion of the overall population, get defined as “working class,” arguably making the group too large to target politically in a meaningful way. The median household income of non– college-educated whites was $47,500 in 2004, slightly above the national median. Consequently, the working-class category of those without four-year college degrees ends up comprising a rather miscellaneous group, lumping together people living below the poverty line with many reasonably well-off people. Indeed, college dropout and richest man in America Bill Gates is considered working class under this standard. One outlier hardly disproves a theory, but according to the NES fully 29 percent of voters have some college education but no degree, slightly outnumbering those with a bachelor's degree or more. The “some college” group was, according to 2004 exit polls, the educational cohort in which Bush achieved his best performance. Thus, the conservative inclinations of the educationally defined working class are largely attributable to the sentiments of its best-educated members.



The moral of the story, in my view, is that we need better data. With a sufficiently large data set and adequate statistical tools, it should be possible to try to prize apart the influence of age, income, and educational attainment on voting as separate factors. But as things stand, the picture looks very murky. One major takeaway, though, is that people need to write and talk more carefully about the oft-neglected "some college" crowd. This is a much larger proportion of the population than educated professionals tend to realize, and it's their conservative political views that mainly drive the right-leaning voting habits of the entire non-college block. Since I feel like most pundits don't realize that "some college" status is so common, they also don't realize what occupations "some college" people are doing, or really have a clear picture in their heads of who these people are even though their political views are the cornerstone of a major trump in contemporary political journalism.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.