A Psychological Profile of the Alt-Right 1

A Psychological Profile of the Alt-Right

Patrick S. Forscher1 and Nour S. Kteily2

1Department of Psychological Science, University of Arkansas, 2Department of Management and Organizations, Kellogg School of Management, Northwestern University.

Author notes

Data and materials for this project can be found at https://osf.io/xge8q/

Conceived research: Forscher & Kteily; Designed research: Forscher & Kteily; Collected data: Forscher & Kteily; Analyzed data: Forscher & Kteily; Wrote paper: Forscher & Kteily; Revised paper: Forscher & Kteily.

Address correspondence to Patrick S. Forscher, Department of Psychological Science, University of Arkansas, 216 Memorial Hall, Fayetteville, AR, 72701 (Email: [email protected]).

A Psychological Profile of the Alt-Right 2

Abstract

The 2016 U.S. presidential election coincided with the rise the “alternative right” or “alt- right”. Although alt-right associates wield considerable influence on the current administration, the movement’s loose organizational structure has led to disparate portrayals of its members’ psychology, compounded by a lack of empirical investigation. We surveyed 447 alt-right adherents on a battery of psychological measures, comparing their responses to those of 382 non- adherents. Alt-right adherents were much more distrustful of the mainstream media and government; expressed higher Dark Triad traits, social dominance orientation, and authoritarianism; reported high levels of aggression; and exhibited extreme levels of overt intergroup bias, including blatant dehumanization of racial minorities. Cluster analyses suggest that alt-right supporters may separate into two subgroups: one more populist and anti- establishmentarian and the other more supremacist and motivated by maintaining social hierarchy. We argue for the need to give overt bias greater empirical and theoretical consideration in contemporary intergroup research.

Keywords: politics; extremism; intergroup relations; dehumanization; prejudice; alt-right

A Psychological Profile of the Alt-Right 3

A Psychological Profile of the Alt-Right

The 2016 U.S. presidential election broke with orthodoxy on numerous counts. One of its most surprising features was the fact that won the Republican nomination — and subsequently the U.S. presidential election — despite flouting both conservative orthodoxy and U.S. political norms forbidding offensive speech targeting minorities.

Trump’s election coincided with the rise of a political movement, the “alternative right” or “alt-right”, that took an active role in cheerleading his candidacy and several of his controversial policy positions (Schreckinger, 2017). Although the movement embraced Trump enthusiastically, the precise nature of its membership and goals remain unclear. Nevertheless, associates of the alt-right appear to wield considerable influence within Trump’s administration (Schreckinger, 2017), dramatically increasing the movement’s reach and political power. Moreover, some alt-right associates have shown a willingness to use aggressive behavior in pursuing their aims, including violence at political rallies (Berkeleyside editors, 2017) and doxxing of political opponents (i.e., publicly releasing sensitive personal information on the internet; Broderick, 2017). Beyond the significance of the alt-right’s rise for theories of intergroup relations, the movement’s growing political influence and the seeming willingness of some of its adherents to use extremist tactics towards their objectives highlight the practical need to better understand the psychological roots of alt-right support.

Some of the opaqueness of the alt-right’s motivations follows from the movement’s largely decentralized structure; whatever formal organization it has exists primarily online (Caldwell, 2016; Bokhari & Yiannopolous, 2016; Lyons, 2017; NPR Staff, 2016; Southern Poverty Law Center, 2017). This has led to divergent understandings of their psychology. On one end of the spectrum are portrayals characterizing the movement as primarily driven by anti- globalist and anti-establishment sentiment (Guardian style editors, 2016; Bokhari & Yiannopoulos, 2016; Daniszewski, 2016). On the other end are portrayals characterizing the movement as driven by anxiety about threats to the status and power of US-born Whites (Lyons, 2017; SLPC, 2017; Caldwell, 2016; Daniszewski, 2016; see also Gest, Reny & Mayer, 2017), with some outlets explicitly labeling the movement as White supremacist (NPR staff, 2016; Armstrong, 2017).

These different views of the alt-right make distinct predictions about its primary goals. If the alt-right is primarily anti-establishment, one would expect its adherents to focus on transferring power from perceived elites to perceived non-elites. If the alt-right is primarily supremacist, one would expect its adherents to focus on protecting and promoting the political interests of Whites and other favored groups. These characterizations need not be mutually exclusive; populist and supremacist subgroups may exist within the alt-right, and a given person may identify with the movement for both populist and supremacist reasons — perhaps especially if they view Whites as a victimized group whose plight has been ignored by elites (Craig & Richeson, 2014; Norton & Sommers, 2011).

We sought to clarify the psychology of people who identify with the alt-right movement by comparing the characteristics of supporters and non-supporters. We assessed an array of psychological constructs, capturing a wide range of potentially relevant distinguishing characteristics. Based on journalistic and academic investigations of the radical right, we

A Psychological Profile of the Alt-Right 4 speculated that populism and might be important features of alt-right support and ensured that we assessed associated psychological constructs. For populism, these included concern about the gap between elites and non-elites, general attitudes about the economy, and trust in various media outlets. For supremacism, these included desire for collective action on behalf of Whites, Social Dominance Orientation (Pratto, Sidanius, Stallworth, & Malle, 1994), blatant dehumanization (Kteily et al., 2015), and the motivations to express prejudice toward Blacks (Forscher et al., 2015). We use “populism” and “supremacism” broadly — the former to mean suspicion of elites and mainstream institutions, and the latter to mean a belief that some groups are superior to others and need their interests protected.1 Finally, we measured characteristics, such as the Dark Triad and self-reported aggression, that might capture a disposition toward extremism.

Although we did not have firm hypotheses due to our study’s exploratory nature, we had a modest expectation that alt-right adherents would prefer groups that have been historically advantaged in US society (e.g., men, Whites), derogate both disadvantaged minority groups (e.g., Blacks, Muslims) and political outgroups (e.g., Democrats), distrust mainstream institutions, and report high levels of extremist behavior.

Method

All data and materials can be found at https://osf.io/xge8q/.

Alt-right sample. We recruited our participants using mTurk and set a recruitment goal of 400-500 per sample, which we believed was large enough to obtain accurate effect size estimates for most psychological variables (Schönbrodt & Perugini, 2013). Our mTurk posting specifically asked for alt-right participants, and we offered participants $3 for their participation.

To ensure that this sample genuinely identified with the alt-right, we added identification probes both at the beginning and end of the survey (with the latter probe appealing for honesty and assuring that compensation was not contingent on their response; see Chandler & Paolacci, 2017). If a participant answered that they were not a member of the alt-right at the beginning of the survey, they were routed out and a cookie was placed in their browser preventing re- responding. If the participant indicated they were not alt-right at the end of the survey, they still received payment but their data were not included in analysis.

978 people viewed the online consent form, 682 identified as alt-right in the first probe, 552 made it past the initial demographic questions, 492 reached the second probe, and 480 (97.6%) identified as alt-right on this second probe. We eliminated 36 extra responses: 17 because they were not recruited through mTurk,2 14 because we determined that they were unlikely to be true members of the alt-right from their free response data (specifically, when asked to describe the movement, they copied text directly from the two top Google hits for “alt- right”: Wikipedia and the Southern Poverty Law Center), 3 because we determined that they

1Note that our measured constructs map imperfectly onto these definitions and hence are imperfect approximations of them. 2Initially, we supplemented our alt-right recruitment through , which resulted in 17 responses. Because we met our recruitment goals efficiently using mTurk recruitment alone, we decided to eliminate non-mTurk responses to ensure comparable recruitment methods across samples.

A Psychological Profile of the Alt-Right 5 were less-complete duplicates of other responses in the dataset, and 2 because they both completed the survey very quickly (6 minutes or less) and gave repetitive survey responses. This left 447 participants for analysis (34.0% female, 66.0% male; 93.3% White, 2.9% Black, 4.9% Hispanic, 1.8% East Asian, 1.1% South Asian, 1.1% other). 74.7% reported voting for Donald Trump, 5.6% for Hillary Clinton, 4.5% for Gary Johnson, 0.7% for Jill Stein, 2.9% for another candidate, and 11.6% reported not voting.

Comparison sample. A few weeks after recruiting the alt-right sample, we recruited a sample of participants who did not identify with the alt-right through the same sampling source, mTurk. We did not identify any specific inclusion criteria in our mTurk advertisement. Participants were paid $2 for participating (alt-right participants received a slightly higher incentive because we assumed they would be more difficult to recruit).

To ensure this sample did not identify with the alt-right, we placed an identification probe at the end of the survey. 513 people viewed the online consent form, 411 completed the initial demographic questions, and 405 reached the identification probe, of whom 390 (96.3%) did not identify as alt-right. We eliminated 8 extra responses: 7 because they had duplicate responses in the dataset, and 1 because this person both completed the survey very quickly (6 minutes or less) and gave repetitive survey responses. This left 382 participants for analysis (49.7% female, 50.3% male; 80.9% White, 11.0% Black, 6.3% Hispanic, 5.8% East Asian, 1.3% South Asian, 0.8% other). 17.5% reported voting for Donald Trump, 55.8% for Hillary Clinton, 5.2% for Gary Johnson, 2.1% for Jill Stein, 3.4% for another candidate, and 13.6% reported not voting.

Procedure. All measures described below were presented in blocks of similarly-themed scales. The order of the blocks was randomized, as was the order of the scales within blocks and, unless otherwise noted, the order of questions within scales. We reduced sets of items that did not already have a previously validated factor structure3 through exploratory factor analysis with the alt-right sample (the focus of our investigation), using scree plots of principle components to decide the number of extracted factors followed by oblimin rotations to decide the variables that loaded on each factor (see https://osf.io/xge8q/ for the scree plots and factor loading matrices).

The survey started with four free response questions to allow the participants to describe their thoughts in their own words. For the alt-right sample, these were “In one or two sentences, how would you describe the alt-right ideology?”, “Why do you choose to identify with the alt- right movement?”, “What are some common misconceptions about people who identify with the alt-right movement?”, and “What are your thoughts when people claim the alt-right is racist?” For the comparison sample, these were “In one or two sentences, how would you describe your political ideology?”, “What are your general views about the alt-right?”, “What parts, if any, of the alt-right movement are you sympathetic to?”, and “What are your thoughts when people claim the alt-right is racist?” All answers to these questions are available at https://osf.io/xge8q/.

In our description below, we omit the following measures: (1) scales measuring the attribution of animalistic traits to White and Black people due to overlap with the ascent items

3These included ascent, self-reported aggression, attitudes towards the economy, perceptions of disadvantage, concern about political issues, police attitudes, media trust, and race-based collective action.

A Psychological Profile of the Alt-Right 6

(which we had for a broader array of groups) and to avoid overweighting dehumanization in our cluster analyses (described below); (2) Whites’ beliefs about the goals of Black people because it was an exploratory measure included for another project; and (3) single items measuring the evaluation of President Trump’s job performance, perceptions of Trump as an ally of the alt- right, and satisfaction with the direction of the country because they are single-item measures with unknown reliability.4 We note that, as is common with factor analyses, some subjectivity is required in labeling different factors, and it can in certain cases be challenging to identify one label that captures the full conceptual scope of a given factor. We therefore acknowledge that some of our labels for the constructs may be imperfect. A summary of our scales, including scale anchors, the number of items per scale, sample items, and scale reliabilities is shown in Table 1.

4Descriptive statistics for these measures, and all measures prior to the construction of aggregates through exploratory factor analysis, are shown at https://osf.io/xge8q/. As also shown there, our cluster analysis results are robust to the inclusion of the omitted single-item measures. A Psychological Profile of the Alt-Right 7

Table 1. Survey measures.

A Psychological Profile of the Alt-Right 8

Friendships. Participants were asked to list the names or nicknames of up to five friends. They then rated each friend using 1 to 7 Likert scales on the degree to which they felt close to the friend, the degree to which the friend’s morals matched their own, and the degree to which the friend identified as a member of the alt-right (for the comparison sample, this last question asked the extent towhich each friend shared the participant’s political views). We created averages of each subscale representing their perceived closeness to and moral match with their friends, as well as the degree to which their friends shared their political views.

Personality characteristics. We measured both moral foundations (Graham et al., 2011) and Dark Triad traits (Jones & Paulhus, 2014).

Moral foundations are theorized to derive from innate psychological mechanisms that can be modified by the social environment. These foundations consist of equality, fairness, loyalty, authority, and purity, which are all measured using the Moral Foundations Questionnaire (MFQ). The short form of the MFQ measures each of these foundations with 4 items (i.e., 20 items total), half of which ask participants to rate the relevance of various considerations to their morals on a 0 (not at all relevant) to 5 (extremely relevant) scale, and half of which ask participants to rate their agreement with various statements on a 0 (strongly disagree) to 5 (strongly agree) scale. We only used the agreement items to conserve survey length (i.e., 2 items per foundation). Only the purity subscale yielded acceptable reliability, and we therefore present results only for this subscale.

The Dark Triad traits consist of , Machiavellianism, and psychopathy, each of which is associated with callous, manipulative behavior (Jones & Paulhus, 2011). The short form of the Dark Triad scale (Jones & Paulhus, 2014) measures each trait using 9 items using a 1 (strongly disagree) to 7 (strongly agree) scale. We further shortened these to 4 items each and averaged all items for analysis.

Intergroup orientation. We measured Social Dominance Orientation (SDO; Pratto, Sidanius, Stallworth, & Malle, 1994) and Right-Wing Authoritarianism (RWA; Altemeyer, 1988).

SDO assesses individuals’ beliefs about the desirability of hierarchy between groups. We used six items from the SDO7 scale (Ho et al., 2015), including all four of the short-form SDO-D sub-dimension (reflecting active support for the domination of some groups by others) and two from the short-form SDO-E sub-dimension (reflecting opposition to equality), each of which is measured on a 1 (strongly disagree) to 7 (strongly agree) scale. The two remaining SDO-E items were left out due to experimenter error. We examined the overall SDO scale (i.e., as one dimension) for our purposes here.

RWA is an individual difference composed of conventionalism, submission to authority, and authoritarian aggression. We assessed RWA with 12 items taken from Altemeyer’s (1981) scale.

Motivations to inhibit and express prejudice. These scales consisted of the motivations to express prejudice toward Blacks (Forscher, Cox, Graetz, & Devine, 2015), and

A Psychological Profile of the Alt-Right 9 the motivations to respond without prejudice toward Blacks (Plant & Devine, 1998). Each motivation is further subdivided into an internal (value-driven) motivation and an external (social) motivation, resulting in four subscales: The internal motivation to express prejudice (IMP), and the external motivation to express prejudice (EMP), the internal motivation to respond without prejudice (IMS), the external motivation to respond without prejudice (EMS). Each scale is measured using 5 items on a 1 (strongly disagree) to 9 (strongly agree) scale. We shortened this to 3 items each.

Dehumanization scales. We measured blatant dehumanization of various groups using the ascent dehumanization measure (Kteily, Bruneau, Waytz, & Cotterill, 2015). This scale asks people to rate how ‘evolved’ they perceive people or groups to be using a diagram shown in Figure 1. This diagram depicts the purported biological and cultural evolution of humans from quadrupedal human ancestors. People use a 0-100 slider to decide where a person or group falls along the continuum established by the silhouettes in the image, with a score of 0 corresponding to the quadrupedal human ancestor and a score of 100 corresponding to a modern human. Higher scores therefore indicate humanization, lower scores dehumanization.

Figure 1. Image used for the ascent scale anchor points.

We assessed humanity attributions towards a broad array of targets. On the basis of exploratory factor analysis among the alt-right sample, we created three subscales corresponding to (1) the humanity attributed to targets favored by the alt-right (Americans, Europeans, Swedes, White people, Donald Trump, Republicans, Christians, men); (2) religious and ethnic groups targeted by the alt-right (Arabs, Muslims, Turks, Mexicans, Nigerians, and Blacks); and (3) political opposition groups (Hillary Clinton, Democrats, feminists, Republicans who refused to vote for Trump, journalists).

Self-reported aggressive behavior. We adapted items from the Pew Research Center (2014) to measure the self-reported frequency of online and offline name-calling, physical threats, harassment, and making statements because others find them offensive. We also measured two online behaviors that have no clear offline equivalents: making private information about a person public without their consent (i.e., doxxing) and sharing memes intended to offend others. All self-reported behaviors were measured on a 1 (not at all frequently) to 7 (very frequently) scale. On the basis of exploratory factor analysis, we created

A Psychological Profile of the Alt-Right 10 separate subscales assessing harassing behavior (online and offline threats, online and offline harassment, doxxing) and intentionally offensive behavior (online and offline name-calling, online and offline offensive statements, meme-sharing).

Attitudes about the economy. We adapted four questions from a survey by Pew Research Center (2016). Two of the questions asked the participant to assess each of their personal economic situation and that of the national economy using a 1 (very bad) to 4 (very good) scale. The second asked the participants to rate whether they expected their personal and the national economic situations to get worse or improve using a 1 (worsen a lot) to 5 (improve a lot) scale. On the basis of exploratory factor analysis, we created separate subscales measuring current and future evaluations of the economy across the personal and national dimensions.

Perceptions of disadvantage. We asked participants to rate the degree to which they thought 10 groups were advantaged or disadvantaged in society using a 1 (strong disadvantage) to 5 (strong advantage) scale. On the basis of exploratory factor analysis, we created two subscales assessing perceptions of disadvantage affecting allies and ingroups (i.e., White people, men, Republicans, and the alt-right) and that affecting adversaries and outgroups (i.e., Black people, Muslims, Hispanics, women, and immigrants).

Concern about political issues. We asked participants to rate the extent to which they perceived 12 issues to be a problem in the using a 1 (not at all a problem) to 7 (a big problem) scale. On the basis of exploratory factor analysis, we created four subscales assessing concern about discrimination against ingroups and allies (i.e., discrimination against Whites, discrimination against men), concerns about corruption and wealth inequality (i.e., government corruption, the gap between the Washington elites and the common folk, the gap between the rich and the poor), security-related concerns (i.e., crime, illegal immigration, Islamic terrorism), and concerns about issues and groups prioritized by liberals (i.e., discrimination against Blacks, discrimination against women, climate change).

Police support. The police support questions consisted of 3 items that asked people to agree or disagree with statements that either expressed support for the police or suspicion that police are racist using a 1 (strongly disagree) to 7 (strongly agree) scale. We averaged items together such that higher numbers indicate greater support.

Media trust. The media questions asked the participants to rate their trust in 22 news media outlets, which we sampled from a report by Pew Research Center (2014) and supplemented to capture a broad range of ideological lean and media format. Participants rated the extent to which they perceived each outlet as trustworthy using a 0 (not trustworthy at all) to 100 (extremely trustworthy) scale. Participants were allowed to select a checkbox labeled “Don’t know” if they were not familiar with a particular media outlet. On the basis of exploratory factor analysis, we created separate subscales representing trust in “alternative” media outlets (i.e., Fox News, the Drudge Report, InfoWars, , Rush Limbaugh, Sean Hannity, Glenn Beck) and more “mainstream” outlets (i.e., CNN, the Economist, NYT, Washington Post, BBC, the Wall Street Journal, ABC, CBS, NBC, and MSNBC. Although they loaded on the same factor, we removed ratings of Buzzfeed, the Rachel Maddow Show, and the Huffington Post because they did not seem as mainstream as the other outlets in the cluster. Note that the “alternative” media grouping contains Fox News, which is arguably mainstream (at least in terms

A Psychological Profile of the Alt-Right 11 of the size of its viewership) but tends to take a substantially different editorial line than other mainstream television outlets and may thus be seen by participants as aligning with the rest of the “alternative” grouping. In addition, the “mainstream” factor contains some outlets, like the Economist and the Wall Street Journal, that lean conservative but, because they tend to emphasize economic rather than social , tend not to advocate for the same social policies as the outlets in the “alternative” cluster (Benkler, Faris, Roberts, & Zuckerman, 2017).

Race-based collective action. This category included two scales, one measuring support for collective action on behalf of the interests of White people (e.g., “I think there are good reasons to have organizations that look out for the interests of Whites”), and a second measuring opposition to the Black Lives Matter (i.e., BLM) movement. Support for collective action on behalf of Whites was measured using 5 items, and the scale assessing opposition to BLM consisted of 3 items. All items used a 1 (strongly disagree) to 7 (strongly agree) scale.

Results

Alt-right descriptive statistics. As shown in Table 2, in absolute terms, the alt-right sample reported low levels of trust in the mainstream media (M = 28.65, SD = 22.27) and moderate levels of trust in alternative media (M = 45.23, SD = 24.90). The alt-right sample also reported levels at or near the scale midpoint of dark triad traits (M = 3.73, SD = 1.11), social dominance orientation (M = 3.98, SD = 1.57), Right-Wing Authoritarianism (M = 4.28, SD = 1.31), and ascent dehumanization of both religious/ethnic outgroups (M = 63.14, SD = 30.61) and the alt-right’s political opposition groups (M = 60.19, SD = 31.44).

Although the mean levels of these latter variables were all near the scale midpoints, they are nevertheless striking considering how the variables are scaled. For example, the means of 63.14 and 60.19 on the ascent scale place the religious/national groups and political opposition groups just above the primitive, stoop-necked human ancestor in Figure 1 and are just 10 rating points above Americans’ ascent ratings of ISIS (M = 53.53, Kteily et al., 2015), a group with which the US is engaged in violent combat. The mean of 3.98 on social dominance place the alt- right participants directly between disagreement and agreement on items like “Some groups of people are simply inferior to other groups” (social dominance), a full rating point above the level that typifies a nationally representative sample of Whites (Ho et al., 2015).

From a similar perspective, the mean reported levels of some of the other variables are “high” even though they are far from the scale maximums. This is the case for the motivations to express prejudice (IMP: M = 3.74, SD = 1.57; EMP: M = 3.07, SD = 2.19), self-reported offensive behavior (M = 2.69, SD = 1.68) and harassment (M = 1.98, SD = 1.44). These mean levels imply that people in the alt-right sample only slightly disagree with statements like “Avoiding interactions with Black people is important to my self-concept” (IMP) and “I avoid interactions with Blacks because of pressure from others” (EMP), and reported doing things like “Online, physically threatened another person” (harassment) and “Offline, made a statement because others find it offensive” (offense) somewhat frequently. Levels of desire for collective action on behalf of White people (M = 5.06, SD = 1.62) and opposition to collective action on behalf of Black people in the form of Black Lives Matter (M = 5.57, SD = 1.36) were particularly high, reflecting support for statements like “I think there are good reasons to have organizations A Psychological Profile of the Alt-Right 12

Table 2. Descriptive statistics of the alt-right and comparison sample. “Estimate” reflects the difference in means across samples (Malt-right – Mcomparison), “Lower” reflects the lower bound of the 95% CI, “Upper” reflects the upper bound, and “d” reflects the standardized mean difference. The means are color coded such that red represents the scale minimum, white represents the scale midpoint, and green represents the scale maximum. The mean differences are color coded such that red represents a third of the maximum possible negative difference, white represents no difference, and green represents a third of the maximum possible positive difference.

A Psychological Profile of the Alt-Right 13

that look out for the interests of Whites”, and “I think the Black Lives Matter movement has been very harmful to our country”.

Comparisons of alt-right adherents and non-adherents. The characteristics of the alt- right sample are brought into still starker relief when compared to the non-alt-right sample. As shown in Figure 2 and Table 2, the alt-right sample reported much higher levels of SDO, RWA, the motivations to express prejudice, harassing and offensive behavior, perceptions that the alt- right’s ingroups were at a disadvantage, concern about discrimination against the alt-right’s ingroups, concern about security issues, respect for the police, trust in alternative media, support for collective action on behalf of White people, and opposition to collective action on behalf of Black people. The alt-right sample also reported lower levels of embeddedness among ideologically similar friends, internal motivation to respond without prejudice, perceptions that the alt-right’s outgroups are at a disadvantage, concern about traditional liberal issues, and trust in mainstream media. Finally, the alt-right sample reported much higher levels of dehumanization of religious and ethnic groups, as well as the alt-right’s political opposition groups.

A Psychological Profile of the Alt-Right 14

Figure 2. Density plots of the survey variables for the alt-right and comparison samples. The heights of the density plots are scaled such that 1 is the distribution maximum.

A Psychological Profile of the Alt-Right 15

Interestingly, compared to the non-alt-right sample, the alt-right reported relatively similar levels of closeness and moral match to their friends, evaluations of the current state of the economy, concern about government corruption, and attributions of humanity to the alt-right’s in-groups (interestingly, this last finding suggests that alt-right non-adherents refrain from dehumanizing the groups the alt-right supports). Finally, the alt-right expected more improvement in the state of the economy relative to the non-alt-right sample.

Many of the differences between the alt-right and non-alt-right samples were quite large. The difference between the two samples in the humanity attributed to religious and ethnic outgroups disfavored by the alt-right (e.g., Arabs, Muslims) and political adversaries of the alt- right (23.82 and 25.93 points, respectively) was a quarter the size of the maximum difference possible on the ascent scale, representing nearly a full standard deviation in our sample (d = .91 and d = .96, respectively). If we translate the alt-right and non-alt-right ratings into their corresponding ascent silhouettes, this means that our alt-right sample saw religious, national, and political opposition groups as a full silhouette less evolved than the non-alt-right sample (see Figure 1). Similarly, the difference in support for collective action on behalf of White people was 2.41 points, about a third the size of the maximum difference possible on a 1-7 scale, and about one and a half standard deviations large (d = 1.48).

Correlational results. A full correlation matrix, presented separately for alt-right and non-alt-right participants, is shown at https://osf.io/xge8q/. We also focused more closely on the correlates of four variables we deemed to be especially important outcomes: self-reported harassment, self-reported offensive behavior, support for collective action on behalf of Whites, and opposition to the Black Lives Matter movement. As predictors of these outcomes, we chose SDO, RWA, the internal and external motivations to express prejudice, and a difference score representing the relative amounts of humanity attributed to White and Black people (White ascent – Black ascent), which we used in lieu of the separate ascent ratings of the two groups to remove individual differences in scale usage (Kteily et al., 2015). For each outcome, we fit regression models that examined all these predictors both individually and simultaneously. To assess whether these variables serve as potential explanations for the gap between alt-right and non-alt-right participants in the aggression and collective action variables, we also included a dummy-coded indicator for sample as a predictor [0=comparison; 1=alt-right]. If, for a given outcome, the coefficient for sample is reduced in the simultaneous model relative to the individual model, this tentatively suggests that the set of psychological predictors help explain why alt-right and non-alt-right participants differ on the outcome.

A Psychological Profile of the Alt-Right 16

Figure 3. Regression results. Each column represents a regression outcome, each row a predictor. In the “Individual” models, only the predictor was regressed on the outcome; in the “Simultaneous” models, all predictors were regressed on the outcome simultaneously. “Dehumanization” represents the relative ascent scores of Whites and Blacks (White ascent – Black ascent); “Sample” represents a dummy-coded indicator [0=comparison; 1=alt-right].

A Psychological Profile of the Alt-Right 17

The results are shown in Figure 3. Focusing first on self-reported harassing and offensive behavior, all five predictors were associated with these outcomes in the individual models, but the only variables to maintain robust relationships in the simultaneous models were the motivations to express prejudice. In the simultaneous models, people who reported being motivated to express prejudice for both internal, value-driven reasons and external, social reasons reported both harassing (IMP: b = .08, 95% CI = [.03, .14]; EMP: b = .28, 95% CI = [.23, .33]), and offending others (IMP: b = .09, 95% CI = [.02, .15]; EMP: b = .26, 95% CI = [.19, .32]) more frequently. The coefficients representing the comparison between the non-alt- right and alt-right samples were greatly reduced from b = .69, 95% CI = [.52, .85] and b = 1.19, 95% CI = [.99, 1.39] in the individual models to b = .17, 95% CI = [.00, .34] and b = .43, 95% CI = [.22, .65] in the simultaneous models; these reductions were both statistically significant, harassment: b = .49, 95% CI = [.31, .66], offense: b = .71, 95% CI = [.51, .92]. Keeping in mind the limitations imposed on our conclusions by our cross-sectional design, this tentatively suggests that, as a set, the psychological predictors help to account for the gap in self-reported aggression between the alt-right and non-alt-right samples.

All predictors were also associated with the collective action variables in the individual models. In the simultaneous models, the motivations to express prejudice were no longer associated with the collective action variables, but all other predictors were. Specifically, people who overtly dehumanized Blacks relative to Whites, who reported more social dominance, or who reported more right-wing authoritarianism were all more likely to support White collective action (ascent difference: b = .012, 95% CI = [.008, .016]; SDO: b = .24, 95% CI = [.16, .32]; RWA b = .37, 95% CI = [.29, .45]) and oppose Black Lives Matter (ascent difference: b = .007, 95% CI = [.002, .011]; SDO: b = .24, 95% CI = [.16, .32]; RWA: b = .37, 95% CI = [.29, .45]). The coefficients for sample were reduced from b = 2.41, 95% CI = [2.19, 2.63] and b = 1.98, 95% CI = [1.77, 2.20] in the individual models to b = 1.01, 95% CI = [.78, 1.23] and b = .90, 95% CI = [.66, 1.15] in the simultaneous models. These were statistically significant reductions (White collective action b = 1.41, 95% CI = [1.19, 1.64]; BLM opposition b = 1.09, 95% CI = [.86, 1.33]). Still, the coefficients for sample remained substantial in the simultaneous models. This suggests that our predictors may help explain part, but not all, of the differences in desire for White collective action and opposition to Black Lives Matter between the alt-right and non- alt-right samples.

Cluster analysis. Among the alt-right sample, many of the density distributions shown in Figure 1 are markedly bimodal. This is especially true for the motivations to express prejudice, the external motivation to respond without prejudice, ascent dehumanization of derogated and political opposition groups, self-reported aggressive behavior, the desire for collective action on behalf of Whites, and opposition to Black Lives Matter. These bimodal distributions hint at possible subgroups of the alt-right that differ in their psychological traits – and possibly their reasons for identifying with the alt-right.

We investigated the possibility of subgroups among our alt-right sample through model- based clustering with the mclust package (Fraley & Raftery, 2002). Model-based clustering represents each cluster as a mixture of variable distributions, each of which has a functional form that the analyst must specify. mclust handles this specification of the cluster’s distributional A Psychological Profile of the Alt-Right 18

properties by testing a variety of functional forms and choosing the one that yields the best Bayesian Information Criterion (Fraley & Raftery, 1998).

In addition to specifying a functional form, the analyst must select the number of clusters used to generate a cluster analysis solution. Methods of evaluating the plausibility of a particular number of clusters generally fall into two categories: internal criteria and stability criteria. The internal criteria evaluate the quality based on whether the clusters each have a compact structure and are separated from each other in multidimensional space, whereas the stability criteria evaluate quality based on whether the cluster solutions change when each variable is deleted from the dataset. Internal criteria thus assess fit of the cluster solution to the data, whereas stability criteria assess robustness of the cluster solution to alternative specifications.

Table 3. Validation measures for cluster analysis solutions with varying numbers of clusters. The “best” score for each validation measure is highlighted in yellow.

After standardizing all variables to ensure that none of them were overweighted during clustering, we chose the number of clusters and validated our choice using the internal and stability criteria from the clValid package (Brock, Pihur, Datta, & Datta, 2008). As shown in Table 3, the two-cluster solution performed best on two of the stability metrics, the Average Proportion of Overlap (APN) and the Average Distance Between Means (ADM). All solutions had very similar scores on remaining two stability metrics. We concluded from these results that the two-cluster solution was most robust to alternative specifications. On the internal metrics, all clusters performed similarly on the Dunn metric, and the three-cluster solution performed best on Connectivity and the

A Psychological Profile of the Alt-Right 19

Silhouette. However, the two-cluster solution did not perform poorly on these latter two metrics; indeed, it performed second best after the three-cluster solution. We selected the two-cluster solution as the one that provides the best level of robustness while still providing good fit to the data.

The two-cluster solution used a functional form that fixed the covariance structure’s volume and orientation, while allowing the shape to vary. This resulted in clusters that contained 226 and 217 people. As shown in Figure 4, the two clusters were more similar to each other than they were to the non-alt-right sample. As shown in Table 4, these similarities spanned a large number of variables, including right-wing authoritarianism, trust in mainstream and alternative media, attitudes about the current and future state of the economy, support for collective action on behalf of Whites, and opposition to the Black Lives Matter movement. These similarities suggest that membership in the alt-right better differentiates our measured psychological characteristics than does membership in the two alt-right subgroups.

Nevertheless, the two clusters did differ meaningfully. The larger cluster (N = 226), which we labeled the “populists”, was distinguished by its relatively high levels of concern about government corruption, M = 5.68, SD = 1.48. The smaller cluster (N = 217), which we labeled “supremacists”, was distinguished by their higher levels of SDO, their heavy dehumanization of ethnic/religious outgroups and political opposition groups (national ascent: M = 51.49, SD = 29.51; opposition ascent: M = 51.68, SD = 29.68), and their internal and external motivations to express prejudice toward Black people (IMP: M = 4.85, SD = 2.18; EMP: M = 4.16, SD = 2.11) — all features central to the view of certain groups of people as inferior to others. The people in this cluster also reported high levels of Dark Triad traits, M = 4.53, SD = .93 and aggressive behavior (harassment: M = 3.01, SD = 1.49; offense: M = 3.98, SD = 1.40), constructs that we interpreted as indicative of extremist tendencies. Of note, the supremacist cluster was also high relative to the populist cluster in their ideological embeddedness among other alt-right friends.

We should note that, although those belonging to the supremacist cluster prioritized the interests of White people, this was neither their exclusive concern nor their singular defining characteristic. Indeed, as we note above, populists and supremacists alike were equally high in their desire for White collective action and their opposition to the Black Lives Matter movement. Rather, supremacists were most notably distinct from populists in their willingness to derogate other groups, including but not limited to Blacks. To illustrate, in addition to perceiving Black people half-way between the ape-like human ancestor and the modern “full” human (M = 51.45, SD = 33.35), supremacists similarly dehumanized Democrats (M = 52.15, SD = 33.90) and journalists from mainstream media outlets (M = 51.48, SD = 34.29), and rated both Muslims (M = 44.78, SD = 34.68) and feminists (M = 46.88, SD = 34.19) closer to the ape-like human ancestor than the modern human. A Psychological Profile of the Alt-Right 20

Figure 4. Density plots of the survey variables for the comparison sample, the populists, and the supremacists. The heights of the density plots are scaled such that 1 is the distribution maximum.

A Psychological Profile of the Alt-Right 21

Table 4. Descriptive statistics of the populist and supremacist clusters. “Estimate” reflects the difference in means across clusters (Msupremacist – Mpopulist), “Lower” reflects the lower bound of the 95% CI, “Upper” reflects the upper bound, and “d” reflects the standardized mean difference. The means are color coded such that red represents the scale minimum, white represents the scale midpoint, and green represents the scale maximum. The mean differences are color coded such that red represents a third of the maximum possible negative difference, white represents no difference, and green represents a third of the maximum possible positive difference.

A Psychological Profile of the Alt-Right 22

Discussion

Portrayals of the alt-right vary widely, with some emphasizing the movement’s anti- globalist and anti-establishment views (Guardian style editors, 2016; Bokhari & Yiannopoulos, 2016) and others emphasizing the movement’s interest in maintaining structures of group-based privilege (Lyons, 2017; SLPC, 2017; Caldwell, 2016; NPR staff, 2016; Armstrong, 2017).

We found some evidence for the populist portrayal, as alt-right supporters expressed suspicion of mainstream media and trust in alternative media. Interestingly, we found little evidence that this populism extended to economic issues: alt-right supporters were more optimistic about the current and future states of the economy than non-supporters.

In contrast, we found abundant support for portrayals of the alt-right that emphasize their perception that certain historically advantaged groups are superior to other groups and need their interests protected. Our alt-right sample reported high levels of social dominance orientation, strong support for collective action on behalf of White people, and strong opposition to collective action on behalf of Black people. They also tended to their favored groups as less advantaged than their outgroups and adversaries, and saw discrimination against groups like Whites and men as more of a problem than that against groups like women and Blacks. Taken together, these results suggest that members of the alt-right feel that the social positions of their favored groups are under threat, consistent with theoretical accounts describing a conservative shift in response to status-anxiety (Craig & Richeson, 2014; Gest et al., 2017; Norton & Sommers, 2011). Alt-right adherents also expressed hostility that could be considered extremist: they were quite willing to blatantly dehumanize both religious/national outgroups and political opposition groups, reported high levels of the motivations to express prejudice towards Black people, and reported high levels of harassing and offensive behavior.

Although these patterns characterized our alt-right sample overall, our cluster analyses revealed heterogeneity in the form of subgroups. One of these, which we labeled “supremacists”, showed especially pronounced bias favoring certain groups over others and reported characteristics potentially reflective of extremism: they reported very high motivations to express prejudice, extreme dehumanization of religious, national, and political opposition groups, as well as very high Dark Triad scores and more frequent aggressive behavior. The other subgroup, which we labeled “populists”, reported lower extremist tendencies and greater concern about government corruption.

The exact relationship between these two subgroups is unclear. It is possible, for example, that the clusters represent two stages in a developmental trajectory of alt-right identification, with people starting in the populist cluster and then moving into the supremacist cluster as they acquire more alt-right friends — a possibility consistent our finding that those in the supremacist cluster were relatively ideologically embedded among fellow alt-righters. Becoming more embedded within alt-right social networks may further motivate people to express prejudice, both for value-based and normative reasons, causing more dehumanization and aggression. People in the two clusters may also simply have different personalities, a possibility consistent with the finding that those in the supremacist cluster reported higher Dark Triad scores.

A Psychological Profile of the Alt-Right 23

Regardless, our findings clearly suggest that blatant, explicit forms of intergroup bias deserve continued empirical attention (see Forscher et al., 2015; Kteily et al., 2015; Kteily & Bruneau, in press). Despite psychology’s rich history of examining intentional and/or blatant intergroup bias (e.g., Adorno et al., 1950; Westie, 1964; Pettigrew, 1958), contemporary intergroup research focuses on more subtle forms (e.g., Dovidio, Kawakami, & Gaertner, 2002; see Forscher & Devine, 2015). This pattern suggests an unstated — and perhaps unduly optimistic — assumption that blatant intergroup bias is a feature of a bygone past. Although our data do not permit statements about the relative prevalence of either alt-right support or blatant intergroup bias, they do support the argument that blatant intergroup bias has by no means disappeared and must be given greater attention in contemporary theorizing on intergroup issues.

Indeed, it was not just the mean levels of these variables that were notable: consistent with the work highlighting the predictive utility of explicit measures (e.g., Kteily et al., 2015; Forscher et al., 2015), we also observed that several explicit measures, including the motivations to express prejudice toward Blacks, relative dehumanization of Blacks versus Whites, social dominance, and right-wing authoritarianism all predicted self-reported aggression and support for race-based collective action. The fact that these constructs uniquely predicted separate outcomes supports the view that these variables have both discriminant and predictive validity and should be considered in combination rather than isolation. Of particular note, ascent dehumanization appears to be more than a motivation to express negativity toward a group, and the motivations to express prejudice appear to be more than blatant dehumanization.

Limitations

This study is not without its limitations. Our study is cross-sectional, so it can only speak strongly to the correlates rather than causes of alt-right membership. We also only measured a subset of potentially relevant variables, likely capturing an incomplete psychological profile of the alt-right. Of particular interest is whether alt-right adherents differ on variables like the full moral foundations questionnaire (we were only able to measure purity reliably), status anxiety, and trait-based trust.

Our participants were also recruited through convenience sampling. Thus, we cannot speak to whether the psychological profile we documented generalizes to either other alt-right members or other insurgent right-wing political movements. Lastly, our sampling depended on alt-right participants self-reporting their identification, which raises the possibility that they were lying. We do not think this is likely for at least two reasons. First, we probed for alt-right membership a second time once the survey was complete, and ensured participants had no incentive to lie, a best-practice advocated by Chandler and Paolacci (2017). Second, our participants’ free responses (available at (https://osf.io/xge8q/) match our quantitative findings and use terminology unique to the alt-right subculture. For example, several respondents used insults specific to the alt-right (e.g., “”, “”) and dismissed accusations of on grounds of racial realism. One respondent answered the question, “What are your thoughts when people claim the alt-right is racist?” with the following:

[…] If it were not for Europeans, there would be nothing but the third world. Racist really needs defined. Is it racist to not want your community flooded with 3,000 low IQ blacks from the Congo? I would suggest almost everyone would not. It is not racist to want to

A Psychological Profile of the Alt-Right 24

live among your own. […] Through media [the Jews] lie about the Holohoax, and the slave trade. Jews were the slave traders, not Europeans.. many people don't even understand these simple things. […]

We believe a person merely lying for money is unlikely to generate this kind of response.

Conclusion

Our work takes an important step toward examining the psychology underlying identification with the alt-right. Our work reveals support both for portrayals that emphasize their anti-establishment sentiments and those that highlight supremacist tendencies. The group also appears to have some extremist elements. Given the rise in hate groups since Trump’s election (Potok, 2017), and in light of recent calls for social scientists investigate the causes of extremism (Nature Human Behavior, 2017; Baez et al., 2017), understanding the psychological roots of the alt-right and other similar groups is a high priority for future research.

A Psychological Profile of the Alt-Right 25

References Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian personality. Oxford, UK: Harpers. Altemeyer, B. (1996). The authoritarian specter. Cambridge, US: Harvard University Press. Armstrong, A. (2017, February 16). Ex-Neo Nazis Explain What’s Driving the Alt-Right. Vice News. Retrieved from https://www.vice.com/en_us/article/bmpn7q/ex-neo-nazis-explain- whats-driving-the-alt-right Baez, S., Herrera, E., García, A. M., Manes, F., Young, L., & Ibáñez, A. (2017). Outcome- oriented moral evaluation in terrorists. Nature Human Behaviour, 1(6), 118. Brock, G., Pihur, V., Susmita, D., & Somnath, D. (2008). clValid, an R package for cluster validation. Journal of Statistical Software. Broderick, R. (2017, May 21). Trump Supporters Have Built A Document With The Addresses And Phone Numbers Of Thousands Of Anti-Trump Activists. Buzzfeed News. Retrieved from https://www.buzzfeed.com/ryanhatesthis/trump-supporters-have-built-a-document- with-the-addresses Caldwell, C. (2016, December 2). What the Alt-Right Really Means. . Retrieved from https://www.nytimes.com/2016/12/02/opinion/sunday/what-the-alt-right- really-means.html Chandler, J., & Paolacci, G. (2017). Lie for a Dime: When most prescreening responses are honest but most “eligible” respondents are lies. Retrieved from osf.io/preprints/psyarxiv/mcvwk Daniszewski, J. (2016, November 28). Writing about the “alt-right.” Retrieved from https://blog.ap.org/_70e Duggan, M. (2014). Online Harrassment. Pew Research Center. Retrieved from http://www.pewinternet.org/2014/10/22/online-harassment/ Forscher, P. S., Cox, W. T. L., Graetz, N., & Devine, P. G. (2015). The motivation to express prejudice. Journal of Personality and Social Psychology, 109, 791–812. Forscher, P. S., & Devine, P. G. (2016). The role of intentions in conceptions of prejudice: An historical perspective. In T. D. Nelson (Ed.), Handbook of Prejudice, Stereotyping, and Discrimination (2nd ed., pp. 241–254). New York, NY: Psychology Press. Fraley, C., & Raftery, A. E. (1998). How Many Clusters? Which Clustering Method? Answers Via Model-Based Cluster Analysis. The Computer Journal, 41, 578–588. Fraley, C., & Raftery, A. E. (2002). MCLUST: Software for Model-Based Clustering, Density Estimation and Discriminant Analysis (No. TR-415). Washington University Seattle Department of Statistics. Gaertner, S., & Dovidio, J. F. (1986). The aversive form of racism. In J. F. Dovidio & S. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 61–89). Orlando: Academic Press. Gest, J., Reny, T., & Mayer, J. (2017). Roots of the Radical Right: Nostalgic Deprivation in the United States and Britain. Comparative Political Studies, 00, 1-26. Graham, J., Nosek, B. A., Haidt, J., Iyer, R., Koleva, S., & Ditto, P. H. (2011). Mapping the moral domain. Journal of Personality and Social Psychology, 101, 366–385. Guardian style editors. (2016, November 30). Alt-right’: why decided not to ban use of the term. The Guardian. Retrieved from https://www.theguardian.com/world/mind-your-language/2016/nov/30/alt-right-why-the- guardian-decided-not-to-ban-use-of-the-term

A Psychological Profile of the Alt-Right 26

Ho, A. K., Sidanius, J., Kteily, N., Sheehy-Skeffington, J., Pratto, F., Henkel, K. E., … Stewart, A. L. (2015). The nature of social dominance orientation: Theorizing and measuring preferences for intergroup inequality using the new SDO? scale. Journal of Personality and Social Psychology, 109, 1003–1028. Jones, D. N., & Paulhus, D. L. (2011). Differentiating the Dark Triad within the interpersonal circumplex. In L. M. Horowitz & S. Strack (Eds.), Handbook of interpersonal psychology: Theory, research, assessment, and therapeutic interventions (pp. 249–269). John Wiley & Sons Inc. Jones, D. N., & Paulhus, D. L. (2014). Introducing the Short Dark Triad (SD3): A Brief Measure of Dark Personality Traits. Assessment, 21, 28–41. Kteily, N. S., Bruneau, E., Waytz, A., & Cotterill, S. (2015). The ascent of man: Theoretical and empirical evidence for blatant dehumanization. Journal of Personality and Social Psychology, 109, 901–931. Kteily, N. S. & Bruneau, E. (In press). Darker Demons of Our Nature: The Need to (Re-)Focus Attention on Blatant Forms of Dehumanization. Current Directions in Psychological Science. Lyons, M. N. (2017). Ctrl-Alt-Delete: The origins and ideology of the Alternative Right. Political Research Associates. Retrieved from http://www.politicalresearch.org/2017/01/20/ctrl- alt-delete-report-on-the-alternative-right NPR staff. (2016, August 26). What You Need To Know About The Alt-Right Movement. National Public Radio. Retrieved from http://www.npr.org/2016/08/26/491452721/the- history-of-the-alt-right Pettigrew, T. F. (1958). Personality and sociocultural factors in intergroup attitudes: a cross- national comparison. Journal of Conflict Resolution, 2, 29–42. Potok, M. (2017). The year in hate and extremism (Intelligence Report No. 162) (pp. 37–42). Southern Poverty Law Center. Retrieved from https://www.splcenter.org/sites/default/files/intelligence_report_162.pdf Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation: A personality variable predicting social and political attitudes. Journal of Personality and Social Psychology, 67, 741–763. Schönbrodt, F. D., & Perugini, M. (2013). At what sample size do correlations stabilize? Journal of Research in Personality, 47, 609–612. Schreckinger, B. (2017, January/February). The Alt-Right Comes to Washington. Politico Magazine. Retrieved from http://www.politico.com/magazine/story/2017/01/alt-right- trump-washington-dc-power-milo-214629 Southern Poverty Law Center. (n.d.). Alternative Right. Retrieved from https://www.splcenter.org/fighting-hate/extremist-files/ideology/alternative-right Understanding and countering terrorism. (2017). Nature Human Behaviour, 1, 134. Westie, F. R. (1964). Race and ethnic relations. In R. E. L. Faris (Ed.), Handbook of modern sociology (pp. 576–618). Chicago, US: Rand McNally. Wik, R., Simmons, K., Vice, M., & Bishop, C. (2016). In Key African Nations, Widespread Discontent With Economy, Corruption. Pew Research Center. Retrieved from http://www.pewglobal.org/2016/11/14/in-key-african-nations-widespread-discontent- with-economy-corruption/