the troubling implications of factor analysis for democracy (with notes on Adorno)

Human beings have latent characteristics, factors that we cannot directly observe or ask individuals to report but must infer from many observations. For example, you cannot reliably assess students’ knowledge of American history by asking them one question or by inviting them to say how much they know about the topic. The standard method is to ask them many questions about varied topics in US history and derive one or more scores from all this data. Similarly, we typically ask many questions or use many observations to assess a person’s extraversion, racial bias, performance on the job, or even likelihood of voting next November.

One common method for inferring the latent variables from many direct observations is factor analysis, invented by Charles Spearman in 1904 and prevalent in psychology since then (Fabrigar et al., 1999).

In the study of politics, factor analysis is often used to infer latent variables from people’s opinions about political issues and about related topics, such as morality, economics or social identities. This method yields genuine insights. However, to the degree that it explains the phenomena of public opinion and political behavior, it has three troubling implications.

First, some people may have anti-democratic traits that emerge as latent variables, whether or not they would admit to opposing democracy. That was a finding of the classic 1950 work, The Authoritarian Personality by Adorno, Frenkel-Brenswik, Levinson, & Sanford. Given a long questionnaire about a wide variety of topics, some people scored high on an F-scale (“f” for fascism), meaning that they were latently authoritarian, although most would have denied it.

A comparable finding emerged from John R. Hibbing and Elizabeth Theiss-Morse’s Stealth Democracy: Americans’ Beliefs about How Government Should Work (2002). Many Americans apparently believed that political disagreement was a sign of corruption and would prefer government by disinterested elites.

And the currently very influential Moral Foundations theory of Jonathan Haidt finds that many people display a latent variable of Authority, which sounds at least potentially undemocratic, especially if it is a predominant factor for an individual.

Second, if people have latent stances about ordinary political matters, but those stances vary, we are likely to polarize and not to be able to resolve our disagreements, which lie below the surface and may be difficult to shift. Hibbing and Theiss-Morse write, “Deliberation will not work in the real world of politics where people are different and where tough, zero-sum decisions must be made. … Given the predilections of the people, real deliberation is likely to make them hopping mad or encourage them to suffer silently because of a reluctance to to voice their own opinions in the discussion” (207).

Similarly, Haidt writes, “I began to see that many moral matrices coexist within each nation. Each matrix provides a complete, unified, and emotionally compelling worldview, easily justified by observable evidence and nearly impregnable to arguments from outsiders” (Haidt, p. 125). (Here he uses the metaphor of a matrix from the 1999 movie of that name, but factor analysis is his main method.)

Third, if we have latent characteristics that explain our concrete ideas and beliefs, we are not the self-conscious, self-critical learners and participants in self-governance that we might imagine ourselves to be. Haidt and team say, “Individuals are often unable to access the causes of their moral judgments” (Graham et al, 2011, p. 368).

I think that latent psychological characteristics exist, and the search for political ones has yielded insights. However, we should be alert to the implications of this method and careful about overstating its significance.

Given a set of data on a roughly related topic, you are going to be able to find factors. You can’t know a priori how much of the variance the factors will explain, but they will explain some of it. This method may help you to understand your dataset, but it does not reveal that people actually possess latent characteristics. That is a presumption of the method, not a finding of it. [NB: This paragraph applies to exploratory factor analysis, not to confirmatory factor analysis. The latter, when conducted appropriately, tests a hypothesis that has been generated for other reasons.]

Further, the statistical output does not demonstrate that the factors you found are foundations, causes, explanations, or reasons for the directly-observed data. All you have found is that you can model each measured variable as a function of the other variables and an unobserved variable, with some error. The interpretation of that finding requires some other basis, such as a substantive theory of human psychology.

Finally, the whole approach of deriving factors from a questionnaire or an observational checklist (or “big data”) is only one way to study human beings. It should be compared to other methods, including the sensitive interpretation of their explicit speech and their intentional actions and interactions. Such comparisons are especially important if we are trying to draw broad, meta-conclusions about whether people are capable of deliberative self-governance.

All of these issues arose in The Authoritarian Personality, which combined factor analysis of questionnaire data with the psychoanalysis of selected subjects. Although the investigators came either from quantitative, largely Anglophone, positivist research or from Continental critical theory, they shared a premise: people may not know what they really want or believe, but we can find out by digging into their unconscious. In this study, extensive interviews, in what the authors describe as “clinical” settings, were used to suggest survey questions and then helped to interpret the output of factor analysis.

On the team was Theodor Adorno, the great Frankfurt School theorist. Almost forty years later, his colleague R. Nevitt Sanford recalled that “Adorno was a most stimulating intellectual companion. He had what seemed to us a profound grasp of psychoanalytic theory, complete familiarity with the ins and outs of German fascism and, not least, a boundless supply of off-color jokes” (quoted in Gordon, p. 39). I find Adorno’s critical reflections on the study even more timely and interesting than his jokes.

In The Authoritarian Personality, the authors make their theory vivid and concrete by presenting portraits of two pseudonymous subjects, both Republican-voting, college-educated, white men in their twenties. Larry is reluctant to categorize people into groups or make assumptions about individuals based on their demographic characteristics. He is interested in a wide variety of people, is self-critical, and overtly opposes discrimination. In contrast, Mack quickly categorizes people into groups he sees as homogeneous, he views their underlying traits as inescapable, and he assumes that the groups to which he belongs are in zero-sum conflict with the others. He is “pre-fascist” or susceptible to authoritarian politics.

Now consider the behavioral scientists who categorize individuals like Mack and Larry as “highs” or “lows” on measures like the f-scale, and who trace such differences to “deep-lying trends in [the] personality”–trends of which the individuals are unaware (Chapter 1). These scientists sound much more like Mack than Larry.

This similarity raises at least two possibilities:

  1. Mack is right. People do fall into discrete, homogeneous, and contending groups that are determined by underlying, unchosen factors. Mack is wise to categorize people and to assess whole groups critically. It’s just that the category of people we should be concerned with are the authoritarians (including Mack), not the groups that Mack dislikes, such as Jews.
  2. The Authoritarian Personality reflects some of the same problematic social conditions that gave rise to Mack. Authoritarians and scholars of authoritarianism manifest the same tendencies because they are both influenced by the same circumstances.

I think Adorno makes the second argument. Under his own name, he contributes a chapter of “Remarks” to The Authoritarian Personality. Here he describes a shift from the “free competition and market economy” of the late 1800s, which prized and actually enhanced individuality, to the “mass society” of the mid-1900s (Kindle loc. 1417, 1453). “It is not accidental that Freud’s theory was conceived during the second part of the nineteenth century, when individuality as a social category was at its height” (1420). In keeping with his time, Freud emphasized the importance of the individual’s family, biography, and inner life for explaining idiosyncratic outcomes. But now “our society is … on its way to become one and whole, leaving less and less [sic] loopholes for the individual and tolerating less and less nonsocial, individual realms of existence” (1448). “The overwhelming machinery of propaganda and cultural industry” make individuality impossible and mold us into groups” (1463). Or again:

The whole pattern of present-day culture is molded in such a way that it takes care of the masses by ‘integrating’ them into standardized forms of life which are built after the model of industrialized mass production, and by actually or vicariously satisfying their wants and needs. … Populations are treated en masse because they are no longer “masses” in the old sense of the term. They are manipulated as objects of all kinds of social organization, including their own … [1455-6]

In private notes not included in The Authoritarian Personality, Adorno elaborated:

Our high-scoring subjects do not seem to behave as autonomous units whose decisions are important for their own fate as well as that of society, but rather as submissive centers of reactions, looking for the conventional “thing to do,” and riding what they consider “the wave of the future.” This observation seems to fall in line with the economic tendency towards gradual disappearance of the free market and the adaptation of man to the slowly emerging new condition. Research following the conventional patterns of investigation into public opinion may easily reach the point where the orthodox concept of what people feel, want, and do proves to be obsolete (quoted in Gordon, pp. 44-5).

Adorno implies that factor analysis works–it explains our world–because the social and political system has destroyed the autonomous, self-critical individual. People like Larry, who still try to think freely and treat others as free individuals, are naive about their real conditions. “Modern society is a mass society” (1453).

The co-authors of the book advocate “increasing the kind of self-awareness and self-determination that makes any kind of manipulation impossible.” They advocate a true education in the Kantian sense: Bildung. Its ideal is “the rational system of an objective and thoughtful man”–becoming at least as emancipated as Larry, and much more so than Mack.

I assume that Adorno signed on to this chapter because he shared his colleagues’ ideal. Note that despite Adorno’s left-radical roots, this ideal (like Adorno’s lament for the “decay of individuality brought about by the decline of free competition and the market economy” [1417]) makes him sound like a classical liberal in the tradition of de Tocqueville and Mill. The difference, I think, is his profound pessimism about returning to a liberal society under modern conditions.

In any case, this is the main point I want to draw out: To infer unobserved characteristics from human beings’ concrete statements about morality and politics can yield insights, but it also implies a view of people as incapable of self-governance. It thus aligns the researcher with anti-democratic or illiberal research subjects. This is a matter of degree, and cautious use of factor analysis is often helpful. Maybe we can even emancipate people by revealing what they latently believe so that they can criticize those beliefs. But if you think that factor analysis will yield truly definitive insights about public opinion, then your view of the world is akin to Mack’s. Adorno would say you are simply a realist. I think you might be overestimating the power of the method.

See also: who wants to deliberate?; Moral Foundations theory and political processes; structured moral pluralism (a proposal); and Habermas and critical theory (a primer). Citations: Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological methods, 4(3), 272.Graham, Jesse, Nosek, Brian A., Haidt, Jonathan, Iyer, Ravi, Koleva. Spassena, & Ditto, Peter H. 2011. Mapping the Moral Domain. Journal of Personality and Social Psychology, 101:2; Haidt, Jonathan. 2012. The Righteous Mind: Why Good People Are Divided by Politics and Religion; Gordon, Peter E. “The authoritarian personality revisited: Reading Adorno in the age of Trump.” Boundary 2 44, no. 2 (2017): 31-56.

This entry was posted in Uncategorized on by .

About Peter

Associate Dean for Research and the Lincoln Filene Professor of Citizenship and Public Affairs at Tufts University's Tisch College of Civic Life. Concerned about civic education, civic engagement, and democratic reform in the United States and elsewhere.