On Oct 21, 2008, at 12:37 PM, SA wrote:
> Bartels' methodology is deeply questionable. His definition of the
> "white working class" is whites in the bottom third of the income
> distribution, which comes to a bit more than one fourth of whites.
<http://www.princeton.edu/~bartels/kansasqjps06.pdf>
In an earlier version of this review I interpreted those passages as suggesting that the people Frank had in mind were people with low incomes. Thus, I proposed to "follow Frank's lead" (and the earlier statistical work of Stonecash 2000) by categorizing voters on the basis of economic status, using the terms "low-income" and "working- class" interchangeably to refer to people with incomes in the bottom third of the income distribution in each election year. (In 2004, those were people with family incomes below $35,000.) I showed that, contrary to Frank's assertions, white voters in this group had not become less Democratic in their voting behavior or less conservative in their views about economic or social issues. Nor could I find any evidence that they cared more about social and cultural issues than about bread-and-butter economic issues.
As it turns out, that isn't what Frank meant at all. Indeed, in a critique of my earlier paper, he calls this definition "A primar y error . . . a mistake so basic that it effectively negates his entire effort" (Frank 2005, 2). Mostly, the evidence of error seems to be that this definition does not confirm the significance of working- class conservatism demonstrated in "book after book" by sociologists and historians, poll results, fretting by Democratic strategists, and gloating by conservative newspaper columnists.
Frank's solution to this dilemma is to redefine the conflict at the core of his book not in material economic terms, after all, but precisely as "a fight between the ignorant and the educated." "[T]here is an easy way to truly 'follow Frank's lead,' " he now writes (2005, 6): "use educational attainment as a proxy for class." The sole textual basis he offers for this alternative definition is a passing reference in his book to the fact that people living in the "hard-core right-wing parts" of Johnson County "are probably less likely to have college degrees" (2004, 104). Earlier in the same paragraph he wrote more definitely that those areas have "lower per capita incomes," and later in the same paragraph he contrasted them with "more affluent suburbs." Nevertheless, we are now led to understand that income is irrelevant and that the lack of a college degree is the key desideratum for membership in Frank's version of the working class.
The textual interpretation seems strained, but since it is Frank's book I am inclined to accord him the benefit of the doubt. Thus, the analyses presented here "truly 'follow Frank's lead' " by focusing on the political attitudes and behavior of white voters without college degrees.
While it seems fruitless to quibble about who is really in the working class, it is important to be clear about what we are talking about. The potential for confusion is illustrated in a 2005 New York Times column by David Brooks entitled "Meet the Poor Republicans." Brooks writes that "we've seen poorer folks move over in astonishing numbers to the G.O.P." In support of this assertion Brooks notes that "George W. Bush won the white working class by 23 percentage points in this past [2004] election." The 23-point margin refers to white voters without college degrees – precisely the definition of the What's the Matter with What's the Matter with Kansas? 205 white working class now proposed by Frank.1 But are these really "poorer folks"? Poorer than Brooks and Frank, yes. Poor by the standards of ordinary Americans, not really.
Even in 2004, after decades of increasingly widespread college education, the economic circumstances of whites without college degrees were not much different from those of America as a whole. Among those who voted, 40% had family incomes in excess of $60,000; and when offered the choice, more than half actually called themselves "middle class" rather than "working class." Meanwhile, among working- class white voters who could even remotely be considered "poor" – those with incomes in the bottom third of the national income distribution – George W. Bush's margin of victory in 2004 was not 23 percentage points but less than two percentage points.
Over the entire half-century covered by my analysis the mismatch between Frank's definition and his concern for "the poor," "the weak," and "the victimized" (2004, 1) is even more striking: white voters without college degrees were actually more likely to have incomes in the top third of the income distribution than in the middle third, much less the bottom third. However, Frank himself now seems curiously uninterested in such material economic distinctions, or in the political behavior of the unlucky members of his working class who are not earning middle-class or upper-middle-class incomes. His only reaction to the finding that tens of millions of white voters in the bottom third of the income distribution have become significantly more Democratic over the past half- century is to dismiss as "well-known to poll-readers everywhere [the fact] that society's very poorest members tend to vote Democratic" (2005, 3). Apparently Frank has little interest in meeting the poor Democrats.
HAS THE WHITE WORKING CLASS ABANDONED THE DEMOCRATIC PARTY?
Frank portrays a broad shift in the political loyalties of traditionally Democratic working-class voters as crucial to the development of the Republican Party's new "dominant political coalition" (2004, 8). He depicts "sturdy blue-collar patriots," "small farmers," "devoted family men," and "working-class guys in midwestern cities" all contributing their share to this "panorama of madness and delusion" (2004, 10). In his critique of my earlier paper he adds that "Working-class conservatism . . . is an important, if not the pre- eminent reason for the continuing electoral weakness of the Democratic Party" (2005, 2). Figure 1 charts the voting behavior of Frank's white working class in presidential elections over the past half-century, using survey data from the American National Election Studies.2 The solid line in the figure shows the Democratic share of the two-party vote in each election. There are obviously considerable fluctuations in Democratic support from one election to the next as recessions, wars, and good and bad candidates come and go; thus, it would be a mistake to attach too much significance to gains or losses in any single election. An important advantage of the NES data is that they allow us to distinguish these short-term shifts from longer-term trends.
The dotted line in Figure 1 summarizes the long-term trend in Democratic support over the past 14 presidential elections among white voters without college degrees. That trend is clearly downward; the cumulative decline in Democratic support amounts to almost six percentage points. In this respect the data are completely consistent with Frank's account. However, a look beneath the surface of Figure 1 suggests three crucial qualifications.
First, even this gradual decline in Democratic support depends critically on the assumption, implicit throughout Frank's account but never stated, that he is writing about the white working class. The distinctiveness of white political behavior in contemporary America and the overwhelming focus on whites in the recent literature on class politics and voting behavior makes this limitation expedient. Nevertheless, it is worth bearing in mind that ignoring non-whites produces a distorted picture of the broader political situation. Indeed, including non-whites without college degrees in the analysis would be sufficient to reverse the overall trend in Figure 1, producing a two-point increase in Democratic support among the working class as a whole over the past half-century.
Second, although Frank's account may lead readers to imagine that the class distinctions reflected in having or not having a college degree have become more significant in recent elections, a comparison of the voting behavior of whites with and without college degrees suggests the opposite. That comparison is presented in Figure 2. Clearly, the overall pattern is one of convergence rather than divergence. Through the first half of the period there was a fairly large and consistent difference in voting behavior between white college graduates and other white voters. (The average difference in Democratic support between the two groups was ten percentage points, and the only difference smaller than eight percentage points was in the Republican landslide of 1972.) Since 1980, however, there has been no consistent difference in voting behavior between whites with college degrees and whites without college degrees. From this perspective, class (as now defined by Frank) has become much less politically relevant over the past half-century.
On the other hand, class "in the material, economic sense" has become much more politically relevant over the same period. That fact should be evident from Figure 3, which compares the voting behavior of the poorer and more affluent segments of Frank's white working class.3 Here, the general pattern is the opposite of the one in Figure 2. While the electoral significance of educational attainment has declined substantially over the past half-century, the electoral significance of income differences has increased substantially. Through the 1950s and 1960s there was virtually no difference in Democratic support on the basis of income; the average Democratic vote shares among the lower-, middle-, and upper-income segments of Frank's white working class were 48%, 52%, and 48%. (Since almost 90% of white voters in this era counted as "working class" by Frank's definition, the figures are virtually identical for the entire white electorate.) However, since the 1970s there has been a large and fairly consistent gap in partisan preferences between richer and poorer white voters regardless of whether or not they happen to have college degrees. Since 1976, Democratic presidential candidates have received 50% of the votes from the lower-income segment of Frank's white working class, 43% from the middle-income segment, and 35% from the upper- income segment. (The corresponding Democraticvotesharesfrom the lower, middle, and upper thirds of the white electorate as a whole are 51%, 44%, and 37%.)
The pattern of income polarization in Figure 3 is consistent with Stonecash's (2000, 118) finding that "less-affluent whites have not moved away from the Democratic Party and that class divisions have not declined in American politics," and with McCarty, Poole and Rosenthal's (forthcoming, chapter 3) finding that income has become an increasingly strong predictor of Republican partisanship and presidential voting since the 1950s. In the white working class, as in the electorate as a whole, net Republican gains since the 1950s have come entirely among middle- and upper-income voters, producing a substantial gap in partisanship and voting between predominantly Democratic lower income groups and predominantly Republican upper income groups.
The voting behavior of Frank's white working class in the 2004 election suggests that, if anything, the partisan divergence between its richer and poorer segments is continuing to increase.
...
Whereas the rows of Table 1 differentiate Frank's white working-class voters on the basis of income levels, the columns present separate tabulations of changing Democratic support for the South and for the rest of the country.4 These separate tabulations suggest a third and even more striking lacuna in Frank's account of the decline in Democratic support among white working-class voters over the past half- century. Focusing on the overall trends, in the first row of the table, we see that the Democratic presidential vote share has declined by almost 20 percentage points among southern whites without college degrees. Among non-southern whites without college degrees it has declined by one percentage point. That's it. Fourteen elections, 52 years, one percentage point.
The remaining entries in the table provide similar comparisons of southerners and non-southerners within each income segment of Frank's white working class. In every case, we see a similar 20-point gap between the South and the rest of the country. Among the most affluent segment, the difference is between a substantial ten-point Republican shift outside the South and a massive 32-point Republican shift in the South. Among the least affluent segment, a ten-point Republican shift in the South has been more than counterbalanced by an 11-point Democratic shift in the rest of the country. (On the other hand, within each region we see a similar 20-point difference in the shifts observed among voters in the top and bottom thirds of the income distribution; the economic and regional trends are largely independent and both quite powerful.)