Three-Fifths a Racist: A Typology for Analyzing Public Opinion About Race

At least this is the historical root of liberal democracy, born out of the bloodshed of the wars of religion, and largely confirmed by subsequent experience. “Reasonable disagreement” and “public reason” are terms of art in political philosophy. I intend them in roughly the sense made famous by Rawls (1993).

Caution is warranted because it can turn into arbitrarily dismissing people, recklessly playing the thought police, or ignoring the merit of arguments of people we don’t like. However, even if much caution is warranted in acting on my argument, the conceptual point is still valid. A political theory companion piece, Neblo (n.d.) “Motive Matters: Liberalism & Insincerity,” (http://​polisci.​osu.​edu/​faculty/​mneblo/​papers.​htm) develops the implications of this line of reasoning for contemporary liberal democratic theory with special reference to Rawls’s theory of a free-standing political conception.

These descriptions are merely summary tags for more elaborate theories. I discuss their subtleties as necessary below. A large number of terrific scholars have contributed to this debate. However, Sears and Kinder are the central Symbolic Racism theorists. Henry and Sears (2002); Kinder and Sears (1981, 1996); Kinder and Medelberg (2000); Sears (1994); Sears and Henry (2003, 2005); Tarman and Sears (2005). Symbolic Racism has also been called The New Racism and Racial Resentment. While some of the variations are important for some purposes, my argument is more general in form. Therefore, I will use the original name to refer to the genus. Principled Ideology is my own label. Sniderman has been a staunch defender of Principled Ideology theory. Sniderman et al. (2000); Sniderman and Carmines (1997); Sniderman et al. (1997); Sniderman and Piazza (1993); Sniderman and Tetlock (1986); Kuklinski et al. (1997). Bobo developed the Group Conflict model. Bobo (1983, 1998, 2000). Sidanius’s Social Dominance Theory is probably most similar to the Group Conflict model, though distinct in certain respects. Sidanius and Pratto (1999); Federico and Sidanius (2002a, b). For a more general discussion see also Hurwitz and Peffley (1998); Sear et al. (2000); Stoker (1996); Wood (1994). Remarkably, this list barely scratches the surface of this literature.

Because my sample does not contain enough non-white protocols for separate analyses, “opinion about race” refers to the opinions of whites.

Charles Glock and his collaborators (Apostle et al. 1983) also used a typological approach to study attitudes about race. I was unaware of Glock’s work when I began this study, so the results are not tightly comparable. However, one can still discern some changes over time. To the extent that the typologies are comparable at such a remove, my sense is that three of Apostle et al.’s (1983) main categories (the Geneticists, Supernaturalists, and Radicals) have become quite marginal, while the Individualists and the Environmentalists are still quite recognizable.

I used a cluster sample approach for recruitment, selecting five cluster sites to reach a diverse range of people in greater Chicago based upon criteria likely to affect their views on race. Given the significant resource constraints, the goal was to get a kind of rough stratification by site proxy so that Q-types linked to ascriptive and associative characteristics would have a chance to express themselves. The geography of greater Chicago is very important for its racial politics. The first cluster surveyed the patrons of a place of business on the southwest side of the city. This site was chosen to sample blue-collar ethnic whites living mostly in urban “border” neighborhoods (i.e., segregated white neighborhoods bordering segregated African-American neighborhoods with a history of racial conflict). The next site surveyed the employees of a business establishment in the southwest suburbs. Subjects here varied a bit more in class, and lived further away from any majority African-American neighborhoods. The third site sampled subjects from patrons of a place of business in a southeast side neighborhood with a long history of (relatively) stable racial integration, and a fairly educated, professional group composition. The fourth site sampled subjects from a social organization on the city’s north side consisting mostly of younger professionals. The fifth site sampled subjects from the staff of a high school in the northern suburbs of Chicago. I also made an unsuccessful attempt to get an adequate number of subjects from other regions of the country using acquaintances as proxies, but decided to concentrate on doing a reasonable job of approximating the demographics of Cook county Illinois, rather than doing a very poor job of approximating a national sample. Obviously the lack of a national random sample limits my ability to infer that I have captured the full range of Q types, and their rates in the general population. In future work, I would like to use these data to develop a dramatically shortened survey that could be used on a national probability sample. Subjects were not compensated. The survey was paper and pencil. It was anonymous by default (using a code blinded to the investigator to link pre and post), though there was an option for subjects to identify themselves to the researcher for purposes of doing face to face, semi-structured, follow-up interviews. At each cluster site the researcher made an appeal for respondents, and administered the survey to those present and willing to do it on the spot. There was also an option to take the survey home and return it to the researcher via mail or an intermediary.

For the complete list of race questions see “Race Appendix” accessible at: (http://​polisci.​osu.​edu/​faculty/​mneblo/​papers.​htm).

I did not use Q-methodology in the narrow sense of having subjects sort stacks of items and such. I merely used Q rather than “R” (standard attitude/trait) factor analysis. See McKeown and Thomas (1988). Thus my analyses do not suffer from some of the problems associated with many Q studies: I have a larger and non “purposeful” sample, use standard questionnaire format, avoid ipsative data, etc.

My use of exploratory techniques here might make some readers uneasy. I share their general concern about such techniques, so I should explain why I think that they are scientifically justified and practically useful in the present context: (1) my main hypothesis is that unit heterogeneity problems were bedeviling progress in this debate (which can be tested at a general level without specifying the precise form of the heterogeneity a priori); (2) there was a profusion of theoretically plausible factor structures from which I could have specified a secondary hypothesis, but no good empirical or theoretical reasons to choose among them. Therefore, I was faced with either: (a) picking blindly from among them a priori (which would have been incredibly inefficient), or (b) running a long series of confirmatory tests (which destroys their meaning, rendering them effectively exploratory, and less transparently so); (3) my follow-up interviews with high loading subjects for each factor served as a kind of validity check on the empirically derived typology; (4) any deviation from the true structure of unit heterogeneity would bias against the many significant results on the tests I run below to adjudicate between the Symbolic Racism, Principled Ideology, and Group Conflict interpretations of race politics. Thus, those tests simultaneously provide some evidence for the construct and criterion validity of the typology; (5) since the general heterogeneity hypothesis receives considerable support (below), the factor structure derived here can now help generate theoretical reflection and empirical guidance for formulating new and more detailed hypotheses to test in the future.

I experimented with different dimensional solutions and rotation methods to make sure that the five factors that I extracted were reasonably robust and stable. In earlier version of these analyses I had worked with a Varimax rotation because it is the most widely used method in general, but I settled on an Oblimin rotation on theoretical grounds. Varimax stipulates orthogonal factors, whereas Oblimin allows for oblique factors. It certainly seems plausible that the various “ideal types” could share aspects of each other’s frames. As it happens, the factor inter-correlations are modest (between −.154 and .308), and the same basic five factors are recognizable when compared with the Varimax rotation. It is worth noting again, in this context, that the regional nature of my sample precludes me from inferring that there would not be additional types were I to have a broader sample.

Q methodology uses a combination of three criteria to determine an item’s ability to characterize the different clusters of people: (1) whether the item is distinctively associated with that specific factor; (2) the group’s mean level on the item; and (3) the group’s mean on the item relative to the other groups’. Such criteria are not typically applied mechanically, but rather are deployed by the researcher qualitatively to build an interpretation.

In labeling this group “Principled,” I take no position on whether the principles to which they appeal are ultimately adequate or correct. I simply want to argue that PC’s are within the bounds of public reason—i.e., that racial liberals cannot simply claim racism to avoid engagement with PC’s. For present purposes, I remain agnostic as to which side should or would carry the day in debate conducted according to the constraints of public reason.

Wong and Bowers (1997) found negative results for the Symbolic Racism theorists’ specific formulation of racial resentment as subtle anti-black affect fused with harsh individualism. My results on the specific formulation were more equivocal. Though RR’s absolute affect toward blacks was not particularly low, they rated them lower than other groups (see below). Similarly, if one thinks that individualism is a value “required to be successful in America,” then RR’s bluntly attribute blacks lagging fortunes to its absence. Feldman and Huddy (2005) find that SR functions differentially by a subject’s ideology. Their results are consistent with mine insofar as ideology is implicated in my typology.

At this point, I want to recognize that my use of the name may seem subject to the same accusations of tendentiousness that the Principled Ideology theorists level against the Symbolic Racism theorists. Many of the items that characterize this group are not demonstrably false, racist, or deserving of the pejorative connotations associated with “resentment.” Below, however, I will demonstrate that subjects with this answer pattern hold less defensible attitudes.

I ran a seventh validity test not reported here (but available from the author) because doing so properly would require a substantial digression on developmental theories of moral reasoning. I find that RP’s and PC’s score somewhat higher (and OR’s quite low) on a measure of one’s propensity to grasp and deploy principled moral reasoning to concrete moral dilemmas, even controlling for ideology and education.

Because I use a convenience sample, the strong frequentist interpretation of “statistically significant” obviously does not apply. As is common, I use the term as a shorthand for characterizing the relative size of the intra-sample differences.

The survey was administered in 1998, before Jackson’s own sex scandal.

I thank one of the anonymous reviewers for pushing me on this point, and helpfully complicating my thinking on it.

In “Motive Matters: Liberalism and Insincerity” (http://​polisci.​osu.​edu/​faculty/​mneblo/​papers.​htm) I try to pick up where they leave off and assess whether subjects who score high on racial resentment are necessarily expressing “an antipathy based on a faulty and inflexible generalization.” (108). A detailed conceptual analysis of the kind that Kinder and Sanders begin, but do not carry through, is the most important next step in further resolving this debate. Doing so would involve integrating overt normative arguments, which all parties de-emphasize because of an accusation that Symbolic Racism theorists were allowing their political commitments to interfere with their scholarly judgment. See Tetlock (1994) and Sears (1994). Whatever one’s view on that particular accusation, this debate has clear normative content, and trying to sweep it under the rug is counter-productive, even from a strictly scientific point of view, and certainly from a broader intellectual perspective. I develop these ideas as a scholarly contribution to discussions in political theory, but also to make any influences on my scientific judgment subject to transparent critique.

The original item contains the “to succeed” phrase. Doing so introduces an ambiguity in whether the question should be read as an equal opportunity or outcome question. While I decided to test this independently, Schuman (315) has since made the same point and calls for assessing whether it affects response patterns. The results for this variation and the dependability variation in the worker experiment are somewhat complex, so for purposes of space in this paper, I had to cut their presentation. However, I hope to discuss them in an empirical piece focusing on reasoning processes about race.

To test for both fatigue and “overload reactivity” in my own data, half the sample was given the reverse question order. Comparisons of the early and late questions for each split half look very similar, suggesting that fatigue did not cause greater randomness, nor increased structuration from overload/reactivity.

Thus, they might try to defend themselves by shifting the argument to one of fact, claiming that their generalization is not so faulty understood ceteris paribus.

I plan on trying to disentangle these effects more thoroughly in future work.

Note that, PI theorists can still sustain their claim that many resenters’ opposition to racially progressive policies is over-determined, which is normatively important. See Neblo (n.d.) “Motive Matters: Liberalism and Insincerity.” http://​polisci.​osu.​edu/​faculty/​mneblo/​papers.​htm

A general population sample might also uncover further types, which would be especially plausible with the addition of Southern whites.