Chris Johnston

Larry Bartels continues a conversation regarding Jonathan Haidt’s recent claims regarding “working class” voting in American politics. You can read the back and forth for yourself. See also Gelman’s many posts on this topic which are cited in the Bartels article and appear on his blog. If we take Haidt to be claiming that low income people tend to vote Republican, then, as Gelman has pointed out, and Bartels expands, this is factually incorrect. Rich Americans still vote Republican at higher rates than low income Americans.

The problem I have with this conversation is that by focusing so narrowly on the question of whether low-income individuals vote at 50% or above for the Republicans, or whether rich and poor have reversed (a la Thomas Frank), we lose sight of the very important contributions being made to the study of mass politics by people like Haidt and John Jost and others. The spirit of Haidt’s comments is that moral divisions in the American mass public matter a lot in contemporary politics. I fully understand that Bartels and Gelman see their task as correcting factually incorrect statements about aggregate distributions, and thus it is not their responsibility to delve deeper, but I fear the unintended consequence of this is to delegitimize Haidt’s and others’ broader theoretical approaches, which I think have a lot to offer the field. Consider, for example, the swipe in Bartels’ P.S. at the end of his post:

“P.S. While Haidt’s new post defends the premise of his claim about working-class voters, he tones down the substance of his earlier argument about the crucial political role of moral values, writing that his critics “are probably all correct when they say that economic concerns played a stronger role in recent electoral shifts than the sorts of moral/cultural issues that I and Thomas Frank were talking about.” I would just add that that seems to be the case regardless of whether the (white) working class is defined by income (as in the unpublished paper of mine cited by Haidt) or education (as in the published version of the same piece).”

My current book project with Howie Lavine is intimately related to these debates. Given this, it seems within the broad outlines of the purpose statement of this website to expand this conversation with some data. Please understand that I am not trying to say Bartels and Gelman are wrong; merely that the types of dynamics Haidt identifies should be taken seriously, and in a more nuanced fashion.

Let’s take a look at the 2008 American National Elections Study. The study included oversamples of Black and Latino Americans. Black Americans voted for Obama at very high rates, and there is not enough variance in voting to consider these citizens further. Given the oversample, I will turn first to White Americans, and then consider self-identified Latino Americans. To operationalize the moral/cultural divide in American politics, I will rely on a scale constructed from four pair-wise comparisons of values. For each, the respondent is asked to decide which value is “more important for a child to possess.” The comparisons include “obedience or self-reliance,” “curiosity or good manners,” “being considerate or being well-behaved,” and “independence or respect for elders.” Obedience, manners, well-behaved, and respect for elders are considered “conservative” value responses, and I sum these items such that higher values indicate greater moral/cultural conservatism. This operationalization is very much in line with Haidt’s moral foundations theory of ideological conflict which emphasizes differences in the subjective importance of loyalty and authority as important to understanding liberal-conservative differences. It should be noted that this scale has been utilized in the past to measure the psychological construct of “authoritarianism.” Personally, I hate this term authoritarianism. It implies that people choosing these “conservative” responses (which are typically the majority by the way) are bad people. Since I do not believe this is true, I reject this term, and will instead utilize the term “childrearing values” below. Just remember that higher values mean a more “conservative” moral/cultural outlook.

So…let’s take a look at the data. I broke income and moral/cultural values into three categories. For income, it is a tertile split. For values, because the distribution is a little wacky, I divided the variable into the lower and upper 20% so that comparisons can be made between moral/cultural “liberals” and “conservatives.” The y-axis in all graphs to come represents estimated % Republican of the two-party vote.

So, how much does income matter for Whites Americans? It matters of course. 42% of the bottom third voted Republican, while 60% of the upper third voted Republican. How do moral/cultural divisions compare as a predictor of vote choice among Whites? It turns out they are an equally good, if not superior predictor. 33% of the “moral liberals” voted Republican, while 60% of the “moral conservatives” did. So moral considerations, at first glance matter at least matter as much as income.

Ah, you say, but this is just tapping into a regional dynamic, namely, the South voting Republican at higher rates than the non-South (see Bartels). A reasonable hypothesis, but wrong. One way to examine this possibility is in a fully specified regression model which includes “South” as a control. I can tell you this does not change anything. Another way is to look at the South and non-South separately. What does this look like?

Ah, you say, but this is, as Gelman has pointed out, a conflict of the rich. It is post-materialism. Those who have a bunch of money argue about non-material things, while those struggling vote their material interests. As Gelman states:

“Those conservative moral interests seem a lot more compelling to people who make a lot of money than to people who are just getting by. Or, to flip it around, liberal moral interests seem much more salient if you’re making less than $75,000 a year.”

Yes, again, clearly income matters. But is cultural conflict isolated to the rich? A reasonable hypothesis, but wrong. Let’s look at Whites with less than 75K in income separately. There is no indication that cultural conflict matters only for rich Whites. 28% of the cultural liberals with less than 75K in income voted for McCain, while 60% of cultural conservatives in this bracket voted for McCain. What about a more realistic division for low income, Whites with a household income less than 40K? 26% of cultural liberals voted for McCain, and 56% of cultural conservatives did. Again, a very large gap.

Another interesting dynamic, one which I will expand on in the coming months, is that political “sophistication” increases the extent to which cultural divisions affect mass preferences. To get enough data here, I will look at Gelman’s “less than 75K” White citizens. If we average across all these individuals, cultural liberals again voted for McCain at 28% and cultural conservatives at 60%. We might expect that increasing political sophistication would increase the extent to which these citizens vote their material interests. After all, cultural divides are “gut” conflicts, and thus “easier.” So, what does this distribution look like for the upper 40% of political sophistication? Now 24% of cultural liberals vote McCain while 70% of cultural conservatives do. Again, these are noisy estimates given data limitations, but the pattern is pretty clear. This sophistication dynamic is very important, and constitutes a critical part of my current project, and I will try to develop this further in future posts. I should also note that in fully specified regression models, it is sophistication, not income or education that is the key moderating variable (if you were wondering).

Ah, you say, but the measure you are using taps the construct authoritarianism (boo! You shouldn’t use that term), and this is related to racism, and so this is just an idiosyncrasy of the 2008 election. A reasonable hypothesis, but wrong. The data in 2004 are a bit sparse, and the distribution of childrearing values a bit more annoying, but we can still break values into the bottom 30% and upper 38%, and look at Whites voting for G.W. Bush over Kerry. Looking first at the income tertile split, it matters not at all. About 60% of both lower and upper income categories reported a Bush vote in 2004. What about moral/cultural divisions? 42% of cultural liberals voted Bush, while 71% of cultural conservatives did. What about cultural divides for those people making less than 75K? 43% and 70%. What about less than 40K? 44% and 71%.

Ah, you say, but you are only focusing on Whites. I have not examined Black Americans because there is not enough variance in the vote. But the 2008 study has an oversample of Latino Americans. Do moral divisions matter for these citizens as well? Here I am looking at the bottom 26% and upper 29% of childrearing values for Latinos. While, as expected, Latinos voted Democrat at much higher rates overall (about 77% in these data), moral divisions matter substantially. Moral liberals among self-identified Latinos voted for McCain at 18%, while moral conservatives voted for McCain at 33%. Again, an important gap. Moreover, the sophistication effect above holds for this group as well. For the upper 30% of sophisticates, the numbers for cultural liberals and conservatives are 18% and 40%, with admittedly sparse data.

Ah, you say, this is all well and good, but economics still matters more than culture! How could you think otherwise in such a contemporary environment of economic turmoil!? I completely agree that most American citizens think the economy is the most important issue, and economic preferences show the strongest associations with the vote and with partisanship. It is the purpose of Howie and I’s book to make sense of the fact that cultural divides seem to matter a lot, both in terms of popular narratives and in the data, while economics remains the most important concern of most citizens. So a response to this last critique will have to wait. I hope to continue posting on this issue over the coming months as our book begins to take shape!

For now, what to say? What I am not saying is “Larry Bartels and Andrew Gelman are wrong.” They are factually correct. What I am saying is that the single-minded focus on showing the errors in psychologists understanding of aggregate voting distributions has obscured the importance of the theories from which these statements emerge. This conversation is frustrating to me, because, as I hope I have shown, people like Jonathan Haidt and John Jost have ideas that need to be taken seriously if we are to understand mass politics on a deeper level.

Yes, lower income citizens vote Democrat at higher percentages than they vote Republican. But changes in moral/cultural orientations are associated with large changes in the propensity to vote Republican. This is an important part of contemporary American politics, and has consequences for the overall distributions of votes for the two parties, but is obscured by this narrow debate (admittedly instigated by Haidt making incorrect claims) on whether low income citizens pass or do not pass the 50% threshold.