Ideological sorting, racial resentment, class coalitions, and partisan polarization in the United States
Michael Lind argues in the New Class War (Random House: 2020), that 'the managerial overclass' has come to dominate culture, economics, and politics since the 1970s, and imposed an ideology of 'technocratic neoliberalism' across the West. In doing so, they have gutted intermediary institutions and thereby disempowered the working masses. This has led to an intense class confrontation that is driving the rise of 'demagogic populism'. Ezra Klein argues in Why We're Polarized (Simon & Schuster: 2020), that the major political parties, which were ideologically and demographically scrambled at midcentury, have increasingly sorted themselves into hostile camps with little overlap; that there is a vicious feedback loop whereby elites and masses polarize each other along party lines; with the result that partisan conflict intensifies over time. The result is gridlock, political dysfunction, and an increasingly escalating partisan conflict that is threatening the sociopolitical fabric of the United States.
In what follows, we shall map some of these developments and document the importance of the class axis in comprehending them. Noam Chomsky conjectured in one of his hundred books, perhaps it was in Manufacturing Consent (Pantheon: 1988), that elites are always more indoctrinated than the masses. In what follows, we shall have occasion to test this hypothesis. All the data we analyze below is from the cumulative data file of the American National Election Study (ANES) 1948-2016. This is a large (N=59,944) survey of Americans with data going back to midcentury, thus allowing us to interrogate diachronic and synchronic patterns simultaneously.
As we have argued previously, class is best captured by education attainment; and that the two diplomas provide the structural parameters of American class society. In what follows, we define the white working class as non-Hispanic whites with only a high school diploma, and the white middle class as non-Hispanic whites with at least a college degree. The white working class constitutes the dominant strata of American society. See Figure 1.
Recall that class is passed on at your parents dinner table. Class reproduction is so strong that parental educational attainment is strongly correlated with the educational attainment of the respondent's spouse, even after controlling for respondent's education attainment. See Figure 2.
We use the Partisan 7-point scale (1=strong Democrat, …, 7=strong Republican) to chart political polarization. Figure 3 displays the diachronic pattern in party identification by class. We can see that the white working class has increasingly migrated to the GOP. More recently, since 2000, the white middle class has moved towards the Democratic Party. The movement of the white working class towards the GOP is sustained and much more pronounced.
Another way to recover this pattern is to look at the proportion of the populace that identifies strongly with a party. The bottom-right panel of Figure 4 shows that a trivial fraction of nonwhites are strong GOP partisans; the bottom-left panel shows that an increasingly fraction of whites as a whole (now about 20 percent) identify as strong Republicans. But this overall pattern conceals discordant movements by class. The top panels show that, beginning in the late-1970s and accelerating after 2000, the white working class has increasingly identified strongly with the GOP; while college-educated whites have, since 2000, decreasingly identified as strong Republicans.
What is driving these patterns? Here we examine two explanations. First, as Klein argues, ideological sorting of the parties may have increased the salience of ideology, thus driving partisanship. Simply put, now that a vast ideological gulf separates the two parties, those who identify as Liberal have good reason to become Democrats; conversely, those with think of themselves as Conservative find a natural home in the GOP. In order to test this hypothesis, we project the Partisan 7-point score on the Liberal and Conservative thermometers (in both, a score of zero means who you hate the group, while a score of 100 means you love the group). We do this separately for each year for which we have data. If ideological sorting is what is driving partisanship, the percentage of variation explained and the size of the fixed-effects should increase over time. This is precisely what we find; as Figure 5 shows.
Another way to test this hypothesis is to use the 7-point ideology scale (1=extremely liberal, … , 7=extremely conservative), instead of the Liberal and Conservative thermometers as predictors, and a dummy for strong GOP identification as the response in a logit model. Figure 6 displays the fixed-effects of ideology on the probability that the respondent identifies as a strong Republican. We do this separately for elites and masses. We can see that Chomsky's conjecture cannot be ruled out. Since the early-1990s, ideology has become an increasingly strong predictor of GOP partisanship; and this pattern is considerably more pronounced for college-educated whites.
The second hypothesis, argued most forcefully by Michael Tesler, seconded by Klein, and at the heart of the Racial Resentment Hypothesis that has attained a virtually hegemonic position among the professoriate and the scribes, says that racial resentment has driven working class whites into the arms of the GOP and many liberal college-educated whites into the arms of the Democratic Party.
We test the second hypothesis, together with the first, in two ways. First, we compute Spearman's rank correlation coefficients for ideology and partisan score; and partial correlation coefficients for a racial resentment score and partisan score, controlling for ideology. Our racial resentment score is computed as a mean of responses to four questions. Specifically, we average the degree of agreement or disagreement, as appropriate, with the following statements: (1) Conditions Make it Difficult for Blacks to Succeed, (2) Blacks Should Not Have Special Favors to Succeed, (3) Blacks Must Try Harder to Succeed, and (4) Blacks Gotten Less than They Deserve Over the Past Few Years. We do this separately by year and class. Table 1 displays our estimates. Table 1. Spearman correlation with Partisan score. White Working Class Ideology P Resentment |Ideology P 1986 0.239 0.000 -0.026 0.683 1988 0.324 0.000 0.029 0.573 1990 0.265 0.000 0.100 0.155 1992 0.324 0.000 0.066 0.145 1994 0.342 0.000 0.101 0.046 1996 0.466 0.000 1998 0.271 0.000 2000 0.451 0.000 0.213 0.037 2002 0.482 0.000 2004 0.552 0.000 0.085 0.269 2008 0.500 0.000 0.044 0.495 2012 0.572 0.000 0.194 0.000 2016 0.634 0.000 0.127 0.008 Positive correlations indicate strength of Republican identification. Estimates in bold are significant at the 5 percent level. Source: ANES, 1948-2016. White Middle Class Ideology P Resentment |Ideology P 1986 0.604 0.000 0.086 0.260 1988 0.652 0.000 -0.029 0.627 1990 0.626 0.000 0.050 0.517 1992 0.678 0.000 0.182 0.000 1994 0.747 0.000 0.105 0.052 1996 0.731 0.000 1998 0.702 0.000 2000 0.669 0.000 0.167 0.022 2002 0.713 0.000 2004 0.776 0.000 0.228 0.000 2008 0.806 0.000 0.206 0.001 2012 0.799 0.000 0.195 0.000 2016 0.797 0.000 0.162 0.000 Positive correlations indicate strength of Republican identification. Estimates in bold are significant at the 5 percent level. Source: ANES, 1948-2016. We find that the ideology has become increasingly predictive of partisanship across the elite-mass divide; although the strength of the predictive relationship is stronger for the elites. This provides strong support for Chomsky's conjecture and Klein's claim about the increasing salience of ideology in driving partisanship. By contrast, the evidence for the racial resentment hypothesis is weak and recent. It is only in the past decade that the correlations become significant for the working class. The relationship is stronger and older (since 2000) for college-educated whites — providing further support for Chomsky's conjecture.
Note that the correlations displayed above are symmetric — they do not tell us whether the pattern is being driven by the left or the right. In order to get to the bottom of this, we test the hypothesis a second time by estimating logit regressions separately for the GOP and the Democrats; again by class and year. Table 2 displays the fixed-effects of the logit regressions with strong identification with the GOP as the response. Table 3. Fixed-effects for logit model: GOP. White Working Class slope coefficient std error Resentment Ideology Resentment Ideology 1986 -0.10 0.40 0.173 0.129 1988 0.00 0.54 0.150 0.102 1990 0.36 0.48 0.198 0.142 1992 0.03 0.55 0.136 0.096 1994 0.12 1.37 0.150 0.211 2000 0.82 2.18 0.492 0.698 2004 -0.28 1.43 0.227 0.278 2008 0.24 1.68 0.185 0.223 2012 0.57 1.51 0.106 0.112 2016 0.19 1.91 0.164 0.183 Response is strong Republican identification. Intercept not shown. Predictors are normalized to have mean 0 and variance 1. Estimates in bold are significant at 5 percent. White Middle Class slope coefficient std error Resentment Ideology Resentment Ideology 1986 -0.32 1.66 0.281 0.363 1988 -0.27 1.91 0.174 0.280 1990 0.62 1.25 0.326 0.293 1992 0.41 1.49 0.211 0.209 1994 0.22 2.85 0.194 0.426 2000 0.34 1.99 0.380 0.449 2004 0.32 2.75 0.329 0.496 2008 0.34 2.95 0.213 0.422 2012 0.14 3.63 0.129 0.280 2016 0.51 3.39 0.203 0.319 Response is strong Republican identification. Intercept not shown. Predictors are normalized to have mean 0 and variance 1. Estimates in bold are significant at 5 percent. As we can see, the evidence is extraordinarily strong for the theory that ideological sorting has increasingly driven GOP partisanship. The effect of ideology of partisanship is considerably more pronounced for college-educated whites than the white working class — in accordance with Chomsky's conjecture. Meanwhile, the evidence for racial resentment as a driver of GOP partisanship is very weak. The fixed-effects are small and only significant for a single year for both classes (2012 for the white working class; 2016 for the white middle class).
Table 3 displays the same for strong identification with the Democrats as the response. Table 4. Fixed-effects for logit model: Democrat. White Working Class slope std error Resentment Ideology Resentment Ideology 1986 0.07 -0.20 0.156 0.116 1988 -0.02 -0.21 0.136 0.082 1990 -0.18 -0.40 0.162 0.123 1992 -0.15 -0.33 0.104 0.072 1994 -0.30 -0.28 0.147 0.185 2000 -0.04 -0.35 0.290 0.379 2004 -0.43 -0.95 0.246 0.296 2008 0.01 -0.65 0.163 0.168 2012 -0.44 -0.91 0.097 0.115 2016 -0.56 -1.02 0.144 0.151 Response is strong Democrat identification. Intercept not shown. Predictors are normalized to have mean 0 and variance 1. Estimates in bold are significant at 5 percent. White Middle Class slope std error Resentment Ideology Resentment Ideology 1986 0.00 -1.96 0.290 0.419 1988 -0.21 -1.46 0.247 0.307 1990 0.18 -1.75 0.295 0.377 1992 -0.35 -1.24 0.250 0.248 1994 -0.11 -3.35 0.278 0.613 2000 -0.37 -1.24 0.341 0.362 2004 -0.77 -2.67 0.395 0.588 2008 -0.25 -2.62 0.181 0.333 2012 -0.69 -2.79 0.145 0.243 2016 -0.54 -2.33 0.185 0.255 Response is strong Democrat identification. Intercept not shown. Predictors are normalized to have mean 0 and variance 1. Estimates in bold are significant at 5 percent. Again, the evidence for the racial resentment theory of polarization is weak and recent. Although it must be noted that it is stronger than that for the GOP — providing indirect support for my hypothesis that Boasian antiracism is a driver of the breakdown of elite-mass relations and the attendant abandonment of the Democratic Party by the white working class. Meanwhile, ideology is a strong driver of Dem partisanship — particularly for the elite, as Chomsky's conjecture would lead us to believe.
Figure 7 shows the fixed-effects for racial resentment by class and party. The black dotted lines are 95 percent confidence intervals. They largely straddle the X axis (red, dotted), meaning that the slopes are statistically insignificant. Interestingly, the slopes are significant for Dems in 2012-2016, suggesting that polarization driven by race is much more recent than hitherto believed. Again, the results provide evidence in support for my hypothesis on Boasian antiracism as a driver of class confrontation and political polarization.
Figure 8 shows the fixed-effects for ideology by class and party. We see that the effect of ideology on partisanship has increased across the board, although it is more pronounced for elites than masses. These results are congruent with the correlation coefficients reported above. The effect is in no way confined to either party or class. Compared to 1986-1994, all fixed-effects for the second period for which we have data, 2000-2016, are dramatically larger in absolute value. Ideological sorting indeed.
The evidence marshaled above provides strong support for the general interpretation offered by Ezra Klein — although he completely ignores the class axis beyond recycling tropes about working class racism. We find that ideological polarization has increased dramatically across the board. However, it is much more pronounced among college-educated whites relative to the working class. The patterns again point to the early-1990s as a crucial inflection point. Partisan polarization comes to a boil over the past decade. Meanwhile, the evidence for racial resentment as a driver of partisan polarization is weak; and in as much as there is evidence of it at all, it is much more recent and not confined to the GOP. Instead, we find much stronger evidence that white antiracist Democrats have become much more partisan over the past decade. Moreover, the evidence for the hypothesis that racial resentment has made the white working class strong GOP partisans is practically nonexistent.
This is part of my larger project to diagnose the breakdown of elite-mass relations in the United States in particular, and in the West in general. If you find this research agenda important and intriguing, please get in touch. Now that social-justice warriors at Columbia History have conspired to oust me (the case is now with the Supreme Court of New York), I am heading to Arizona to work with Professor Noam Chomsky. We have some projects lined up. But I will try to get him interested in this and other projects of relevance to our present predicament. I'm looking for funding for my work with Noam as well as more permanent research/teaching positions. Let me know if you have any leads for the Policy Tensor!