The April 2021 Morning Consult Eight Nation Survey
Into the Weeds, Deeply
Bob Altemeyer, May 2023
On June 17, 2021 Cameron Easley of the nonpartisan Morning Consult polling firm asked me for comments on a report concerning the levels of authoritarianism in eight Western countries. The data had been collected some weeks earlier (on April 26-28) in the United States, Canada, the United Kingdom, Australia, Italy, Spain, Germany and France, asking (by and large) the same questions of a thousand participants in each place. There has never been so grand and sweeping a study on authoritarianism, but I instantly feared it would have used—as public opinion surveys usually do—shortened versions of various psychological measures. Which, as far as I am concerned, is like trying to observe protozoa with a dirty, low-powered microscope. To my delight, however, I discovered the survey, which had been compiled by Dr. Rachel Venaglia and Dr. Laura Maxwell of Morning Consult, was built around the survey Patrick Murray of Monmouth Polls had conducted for John Dean and me in the fall of 2019, which used full scales[1]. The new results proved nowhere near as breathtaking as the Monmouth endeavor, but clearly the 2021 Morning Consult survey constituted the best overall study ever done in this field. Nothing comes close to its range, representativeness, and opportunity for cross validation and international comparison. You can find their write-up of it at https://morningconsult.com/2021/06/28/global-right-wing-authoritarian-test/ and their report of the survey’s methodology at https://morningconsult.com/2021/06/28/right-wing-authoritarianism-international-study-methodology/ .
I reported the major finding of the survey, that Americans are far more authoritarian than any of the other populations sampled, at the end of my October 2021 essay, “Lessons of the 2020 American Election.” (This paper rests two-clicks away on this website.) Then I happily used the oodles of responses that Morning Consult had collected to develop power-packed, shortened versions of two scales—to encourage other researchers to use better microscopes. (See “A Shorter Version of the RWA Scale,” also on this website.) But no one has reported yet the basic psychometric findings regarding the personality tests used in the Morning Consult survey, which I am offering here because someday, somebody will possibly care, and I for one cannot guarantee to be around to answer any questions.
But be warned: We shall be slogging through “the weeds” on an expedition that has nothing to do with pot, thickets where bean-counters can survive on statistical Bits and Pieces for weeks, but which normal people will quickly find unfulfilling. If you’re one of them normal people, I’ll bet you’ll be doing something else in two minutes. Or less.
The Sampling Procedure Used by Morning Consult
Like many established polling firms now, Morning Consult has recruited “panels” of citizens in various countries whose composition represents the population in certain ways. In gender, age, education, and region-of-the-country in this case. Panels are then asked to provide the grist and gist of polls, and it’s not unheard of that responders get paid a farthing or two for their trouble. Olivia Petersen of Morning Consult told me panels were contacted in the eight countries involved in their authoritarianism study and invited to “opt-in” for the study. The firm stopped taking responses when N=1000 (1001 in the USA for some reason). I asked how many people they had to contact to get their samples, and I have not received an answer. (Polling firms are not eager to share such details, it seems.)
Pollsters don’t magically get a perfectly representative sample this way. 50.3% of the American results, for example, came from men, whereas life conspires in countless ways (mostly self-inflicted) to whack away at us’n with Y chromosomes, such that guys only constitute 49.5% of the adult population. Polling firms handle the discrepancies that appear when all the data are in the barn by “weighting” clumps of citizenry differently. In this case, responses from women would be slightly boosted mathematically to give them the swagger their larger part of the population deserves. But it obviously would only make a difference to someone who was trying to connect authoritarianism to, say, sales of women’s shoes. Which would be no one.
Morning Consult professionally and dutifully weighted its 8001 respondents’ answers, nation by nation, but I found you get essentially the same results with the unweighted answers as they were. So that’s what I’ll report here, as I did with the Monmouth data. But that’s one of the reasons my version of the results will be a little different from Morning Consult’s if you look them up.
(If you did NOT read this far in this paper, write me and tell me I was right about the “two-minute thing.”)
The Contents of the Surveys
The American sample took a 172-question survey, built around the 20-item RWA Scale, two 8-item Social Dominance Scales, a 24-item Prejudice Scale, and a 12-item Religious Fundamentalism Scale. The other hundred questions included a 39-item Morning Consult stab at measuring left-wing authoritarianism (which proved unrewarding, as usual), 11 questions about preferred news sources (nothing surprising found, nothing big either); 10 questions about perceived threats from global warming, terrorists, and so on (again, mainly small, predictable relationships with the personality measures), numerous single-item attitude measures about various things (ditto; single-item measures have trouble producing big connections because the answers they draw are usually pretty unreliable).
Morning Consult also included four “child-rearing” items that purport to measure authoritarianism. They produced poor results, as usual. Since no one, to my knowledge, has said that any of my criticisms of this test were wrong (see Appendix I in Dean & Altemeyer, 2020), I shall not report these particular Morning Consult results. As far as I can tell, there’s no “there,” there, folks, and we have to hack our way through enough weeds as it is.
The respondents in the other seven countries had a somewhat shorter questionnaire, as the items about news sources were omitted. They also, unlike Americans, were not asked for their racial identification, which handicapped analyses of the prejudice scores. Nor were their political party preferences solicited, which too would have been of some use.
So Heigh Ho, Heigh Ho, into the weeds we go. Our first task will be to see if the personality tests were up to snuff, psychometrically. Were they reliable yardsticks? Can we trust the numbers they generated? Then we shall compare the average scores on the various scales in the eight countries. For example, if the United States had the most authoritarian sample, who had the least? Finally, we shall examine the connections among our measures. To foreshadow, prejudice is highly correlated with authoritarianism in North America. Is it in Australia? Spain? Germany?
Performance of the Scales
Reliability of the RWA Scale.
The Right-Wing Authoritarianism Scale was developed in Manitoba, Canada[2] in the 1970s to assess the extent to which sentiments of authoritarian submission, authoritarian aggression, and conventionalism go together in people’s thinking. The 20-item version used here cover a range of topics[3]. Ten of them, called “pro-trait” items, ask about things like supporting strong leaders who want to persecute unconventional people. The ten “con-trait” items, where disagreement = authoritarianism, praise (rather than attack) unconventional people such as homosexuals, atheists, “free thinkers,” and uppity women who have confronted (rather than submited to) authorities.
Cross-validating studies in the United States since the 1970s have shown that the scale measures authoritarian follower attitudes just as well there as in Canada.
Responses to the 20 items intercorrelated .299 on average in Morning Consult’s USA sample, and .317 in Canada[4]. That produced alpha reliability coefficients[5] of .896 and .904 respectively. (See Tables 1 and 2 in the Appendix.)
These are quite respectable figures for a multi-topic scale balanced with pro-trait and con-trait statements, and they fit right in with the results obtained in decades of studies of other populations. But they pale before the astonishing findings found by Monmouth in 2019: .482/.949 in a May pilot survey of New Jersey residents and .516/.955 in the autumn USA nationwide poll. I think the striking cohesion Monmouth discovered arose primarily from a) sampling procedures (we wanted to understand the electorate, so Monmouth contacted previous voters, whereas Morning Consult sampled nonvoters as well) and b) the degree of polarization in America during Trump’s presidency. You hated him and his positions, or you loved him and all he stood for. That splitting may have increased the number of extreme answers to the RWA Scale’s items. In addition, by April 2021, January 6th had happened, and that gave some of Trump’s backers pause, if only temporarily. As well, supporting “the established authorities” and “the proper authorities in government” was more connected to drop-kicking dissenters when Trump was in the White House than they were in April 2021.
Connections in Other English-Speaking Countries. You can get into an argument lickety-split in a British or Australian pub by declaring that their country is just like America. Any North American who has had to learn to drive on the left side of the road in those places knows there are lots of differences. But virtually none showed up in the organization of attitudes solicited by the RWA Scale. The intercorrelation/alpha splits in Britain came striding in at .269/.882, and in Australia, .291/.892. Which is to say that the personality trait of authoritarianism existed in the Morning Consult poll to about the same degree in the United Kingdom and Australia as in the United States and Canada. (See Tables 3 and 4 in the Appendix.) Authoritarian followers in all of those countries had, relatively speaking, pretty similar attitudes connected together pretty similarly on a wide range of subjects. They were peas basically from the same pod.
Connections in the “European” Countries. The same was less true elsewhere, however. The Italian peas (.211/.843), the Spanish peas (.206/.840), the German peas (.181/.818), and especially the French peas (.126/.747) looked noticeably different from the ones in the English-speaking nations. (See Tables 5-8 in the Appendix.) Some of this probably occurred because of translation difficulties. But such problems can reflect underlying cultural differences. The trait might well be organized somewhat differently in those four countries. But another factor, present in all the samples, made the answers to the RWA Scale noticeably less cohesive in Germany and especially France: Evil, evil, evil response sets.
A response set is a tendency to give the same answer to statements more or less regardless of the issue raised. The best known is “yea-saying,” which increases “agree” and “True” and “Yes” responses to a survey. There is also “nay-saying” which does the opposite. And there are others, including that provided by a student in one of my studies who blackened spaces on the bubble sheet to make a frameable rendition of Mickey Mouse.
Response sets arise for many reasons. The more ambiguous the statements, the more people will yea-say or nay-say. People who are poorly motivated to answer the questionnaire will also do this more. So will respondents with low reading comprehension. So will folks who have gotten tired of their task. (In that way, they would be like readers trudging their way through a long, complicated, boring, weedy essay who start skipping to the next paragraph and miss the golden bits that come after the first few sentences.)
Whatever the origin, response sets have a pernicious effect on surveys. They artificially tighten up the responses among pro-trait items, and similarly tighten up the responses among con-trait items. But they also torpedo the connection between the pro-traits and the con-traits. The upshot is, you can safely bet the farm that a balanced scale will have stronger correlations within the pro-trait set, and within the con-trait set, than between those sets. But all of these bugger up the data. They make noise when you need quiet to determine if there’s a signal there. And the stronger the response sets, the less trustworthy the data. That is what happened to the RWA Scale in France. The correlation between the pro-trait items and the con-trait items was so unpositive it was negative (-.219). The more authoritarian this sample looked on the pro-trait items, the less authoritarian they looked on the con-traits. This family had two sets of kids at war with one another[6].
The Social Dominance Scales
Felicia Pratto, Jim Sidanius[7], Lise Stallworth and Bert Malle published the Social Dominance Orientation Scale in 1994 to measure whether one preferred hierarchical or equal relationships between groups. It had items like “Some groups of people are simply not the equals of others” and “If people were treated more equally we would have fewer problems in this country.” Responses to the test had good internal consistency and correlated well with various measures of prejudice—better usually than the RWA Scale did. Moreover, the two tests worked hand in glove, each explaining what the other could not. Together they accounted for most of the variance in prejudice scores. I suggested that the SDO Scale measured the dominating “leader” personality in authoritarian systems, and the RWA Scale corralled the “followers.” Most highly prejudiced persons were either high SDOs or high RWAs, and that seemed an important development.
In 2012 Sidanius, Pratto, and five others led by Arnold Ho published a reconceptualization of social dominance that split the “dominating” aspect from the “inequality” aspect. Each aspect was measured by a balanced 8-item scale. When I was developing the Monmouth Survey I asked Felicia Pratto and Jim Sidanius how they wanted their items presented. They had different wishes. One wanted the various items intermixed. But I followed the wishes of the person who is perhaps better known in this field, which was to package the 16 “Social Dominance”[10] items as 4 pro-trait Dominance statements followed by 4 con-trait Dominance statements, followed by 4 pro-trait Antiequality items and 4 con-trait Antiequality items. Thus this part of the Monmouth survey began with the following eight Dominance items.
“Some groups of people must be kept in their place.”
“It’s probably a good thing that certain groups are at the top and other groups are at the bottom.”
“An ideal society requires some groups to be on top and others to be on the bottom.”
“Some groups of people are simply inferior to other groups.”
“Groups at the bottom are just as deserving as groups at the top.”
“No one group should dominate in society.”
“Groups at the bottom should not have to stay in their place.”
“Group dominance is a poor principle.”
I think I made the wrong choice. This format probably invites another response set, “stereotyped answering,” to muck up the search for truth because some respondents will see they’re being asked the same basic question over and over in slightly different ways. This could even lead to bass-ackward answering on the con-trait items by respondents who, like our bored reader four paragraphs and two end-notes ago, stop reading the statements with sufficient care (or at all).
Something hurt the Social Dominance scales noticeably in the 2019 Monmouth survey, as the two scores correlated substantially less with prejudice than the RWA Scale did, rather than more. If there’s one thing social dominance ought to predict like crazy, it’s inter-group prejudice, so this was a real disappointment. Combining the two Social Dominance scores into one—which set aside the apparently convincing reasons for creating two different measures of social dominance—improved things a little, but not much.
If I had been asked when the Morning Consult survey was being prepared, I would have suggested including the original version of the SDO Scale. As it was, just the “Monmouth” version was used again. Consulting your old friends, Tables 1-8 in the Appendix will reveal (unless I have copied down some numbers wrong) that, for the SDO-Dominance Scale, the intercorrelation/alpha splits were: USA: .237/.717; Canada: .300/.775; United Kingdom: .280/.758; Australia: .274/.753; Italy: .278/.755; Spain: .192/.660; Germany: .151/.584; and France: .158/.606. The results were better for the SDO-Antiequality Scale: USA: .356/.813; Canada: .432/.853; United Kingdom: .361/.814; Australia: .388/.830; Italy: .375/.821; Spain: .339/.800; Germany: .249;.722; and France: .272/.746. Rolling your tired eyes over the numbers in this virtually unreadable paragraph may give you the occasional thrill. But overall, the levels of intercorrelation raise a question. After all, the Dominance Scale is just about one thing: how do you feel about some groups dominating other groups? It asks this eight times. And the Antiequality Scale seeks your opinion about (anti-) equality eight times. Shouldn’t the interitem correlations on each test rise to majestic heights? But they did not.
I suspect some stereotyped answering triggered by my decision on how to arrange the SDO items piled onto ordinary yay-saying and nay-saying in all the samples, especially in France and Germany. Answers to the pro-trait items on the Dominance Scale barely correlated (.073) with the con-trait items in the French sample, and in Germany they formed even less of a crowd (.009). Things shook out a little better on the Antiequality items (.218 and .175 respectively). But the pro-trait responses on the Dominance Scale correlated much stronger with the pro-trait answers on the Antiequality Scale than they did with the con-trait items about Dominance. And so did the two con-trait sets (r’s in the .50s and .60s). So response sets appear to have “wagged the dog” on these measures in France and Germany.
Racial and Ethnic Prejudice
The Prejudice Scale used in the Monmouth poll (Exhibit 1) was based on “proper name” measures I began developing in the 1970s to assess negative beliefs about various racial and ethnic minorities. This particular measure was tailored, as you can see, for an American audience. Responses to the scale proved an excellent predictor of support for Donald Trump. In fact, save for political party preference, Trump supporters were most distinguished from the rest of the sample by their high levels of prejudice—a remarkable finding.
Exhibit 1. The Twenty-four Prejudice Items Used in the 2019 Monmouth Survey
- There are entirely too many people from the wrong places getting into the United States now.
- Muslims bring a valuable new element to American society and should be welcomed.*
- White people are the major victims of discrimination in the United States. The government is on everybody else’s side but theirs.
- The more diverse America becomes, with different people with different religions and heritages from everywhere else in the world, the stronger it will be.*
- Black Americans continue to get less than their fair share of our country’s wealth because of discrimination.*
- Latin Americans are naturally lazy, promiscuous, and irresponsible.
- We should tear down the walls that keep people from different cultures away from us, rather than build new ones. *
- Two really bad things about the Jews is that they killed Jesus, and they secretly control the world’s banks and economy.
- Racial minorities have had it good for years in the United States because of all the government programs that help them get ahead of white people.
- It will be great if someday America has become such a mixture of different people that white persons are in a minority like everybody else.*
- Instead of complaining and protesting all the time, African-Americans should be grateful for how good they have it here compared to where they came from.
- Religions like Islam, Hinduism, and Buddhism are just as true as Christianity, and produce as much good behavior.*
- Every person we let into the country as a “refugee” means another American won’t be able to find a job, or another foreigner will go on welfare here.
- Overall the white race has mainly brought exploitation and suffering to the other peoples of the world.*
- America should stop giving foreign aid to backward countries. They made their mess by themselves, so let them clean it up.
- Black people are just naturally more violent than white people.
- Christianity has very unjustly persecuted the Jews over the centuries.*
- Most minorities on welfare would rather work, but they can’t get jobs that pay a living wage.*
- There’s no way a religion like Islam that produces so many terrorists is as good a religion as Christianity is.
- Latin-Americans are as hard-working, law-abiding, and responsible as any other group of Americans.*
- Certain races of people clearly do NOT have the natural intelligence and “get up and go” of the white race.
- Americans are NOT exceptionally noble compared to the rest of the world. “American Exceptionalism” is just another ugly “master race” theory.*
- The immigrants from the Caribbean and Africa who have come to America have mainly brought disease, ignorance, and violence with them.
- It is good to live in a country where there are so many minority groups present, such as African-Americans, Hispanics, and Asians.*
- Indicates a con-trait statement, for which the scoring key is reversed.
These 24 statements were included in the 2021 Morning Consult survey administered in the United States and, as well, in the other seven countries they studied. The items obviously have less relevance elsewhere, and probably do not tap the major prejudices corroding life in some nations. For example, the group most discriminated against in Canada, First Nations citizens (“Indians”) was not mentioned in the measure Morning Consult used. We shall try to work around this limitation.
As for the internal consistency of responses to the 24 items, the usual interitem correlation/ alpha coefficient splits played out as follows: USA: .303/.913; Canada: .396/.940; United Kingdom: .361/.932; Australia: .350/.929; Italy: .352/.929; Spain: .283/.905; Germany: .357/.931; and France: .250/.890. You can see that French respondents showed the lowest cohesion in their answers.
Given that the items involved evaluations of whites, African Americans, Hispanics, Jews, Muslims, multiculturalism, immigration, persons on welfare, and foreign aid, the level of intercorrelation is striking. But this is one of the seminal findings in social psychology. Early explanations of prejudice held that negative attitudes toward groups were based on unfortunate personal contacts with minorities. But it became clear that prejudiced people usually had negative stereotypes of groups they had never met. Prejudice was something you mainly got from prejudiced people, and it usually came in “All of the above” packages. “You’ve got to be taught.”
We’ll consider the correlations between our other tests and prejudice later in this paper.
Religious Fundamentalism
The last scale we shall consider measures religious fundamentalism and was developed with my dear, late friend Bruce Hunsberger. It does not solicit beliefs on particular subjects (e.g. “Jesus was the Son of God”) so much as the attitude that one’s religious beliefs form the Truest Truth, the fundamentals, the foundation of our relations with the Deity. I included the scale (Exhibit 2) in the Monmouth study because so much of Donald Trump’s support comes from white Evangelicals, who score very highly on this measure. And they did.
Exhibit 2. The Religious Fundamentalism Scale
- God has given humanity a complete, unfailing guide to happiness and salvation, which must be totally followed.
- No single book of religious teachings contains all the intrinsic, fundamental truths about life.*
- The basic cause of evil in this world is Satan, who is still constantly and ferociously fighting against God.
- It is more important to be a good person than to believe in God and the right religion.*
- There is a particular set of religious teachings in this world that are so true, you can’t go any “deeper” because they are the basic, bedrock message that God has given humanity.
- When you get right down to it, there are basically only two kinds of people in the world: the Righteous, who will be rewarded by God, and the rest, who will not.
- Scriptures may contain general truths, but they should NOT be considered completely, literally true from beginning to end.*
- To lead the best, most meaningful life, one must belong to the one, fundamentally true religion.
- “Satan” is just the name people give to their own bad impulses. There really is no such thing as a diabolical “Prince of Darkness” who tempts us.*
- Whenever science and sacred scripture conflict, science is probably right.*
- The fundamentals of God’s religion should never be tampered with, or compromised with others’ beliefs.
- All of the religions in the world have flaws and wrong teachings. There is no perfectly true, right religion.*
- Indicates a con-trait statement, for which the scoring key is reversed.
The splits by nations in the Morning Consult survey came in as follows: USA: .446/.906; Canada: .524/.930; United Kingdom: .475/.916; Australia: .500/.923; Italy: .335/.858; Spain: .332/.850; Germany: .367/.874; and France: .292/.831. One notices that once again the items “hung together” better in the four English-speaking countries than in the four “European” ones. And as before, the French sample showed notably weaker intercorrelation. But the twelve items on the scale seem to be measuring the construct rather well in all the places.
We Have Cleared the First Patch of Statistical Weeds. What have we learned? The five scales used in the Morning Consult survey performed better in the “Anglo” samples than in the others, undoubtedly because of translation problems. (See Table 9). But as noted earlier, underlying cultural differences can affect things too. The issues raised in some statements on a scale may not Velcro onto others in, say, Spain, as well as they do in Canada. But for my money, all of the tests performed better outside North America—where they were developed — than one might have suspected. except for France. Where it seems pronounced response sets undermined all the scales quite a bit.[8]
Table 9. Internal Consistency of the Scales used in the Morning Consult Surveys, by Nation
Scale | USA | Canada | U.K. | Australia | Italy | Spain | Germany | France |
RWA | .299/.896* | 317/904 | 269/882 | 291/842 | 211/843 | 206/840 | 181/818 | 126/747 |
SD/Dom | .237/717 | 300/775 | 280/758 | 274/753 | 278/755 | 192/660 | 151/584 | 158/606 |
SD/Anti | .356/813 | 432/853 | 361/814 | 388/830 | 375/821 | 339/800 | 249/722 | 272/746 |
Prejudice | .303/.913 | 396/940 | 361/932 | 350/929 | 352/929 | 283/905 | 357/931 | 250/890 |
Rel. Fun. | .446/.906 | 524/930 | 475/916 | 500/923 | 335/858 | 332/850 | 367/874 | 292/831 |
*xxx/yyy = mean inter-item correlation/alpha.
If you require an alpha coefficient of at least .80 (which gives a signal-to-noise ratio of 4:1) to make a test worth administering, then all of the measures save SDO-Dominance rang the bell. One would naturally prefer alphas of .90 or better (which have signal-to-noise ratios of at least 9:1), and one finds instances of them in Table 9. I was gob-smacked by the stunning performance of these scales in the Monmouth poll, but the transportability of these North American measures to the other nations involved in the Morning Consult effort amazes me almost as much. In the kingdom of research institutions, most of these tests were developed in a “corner store” in the middle of Canada. North even of North Dakota. Who knew?
Weed Patch No. 2: Differences in Mean Scores on the Scales, by Nation
(Be of good cheer. The last two patches take much less time than the first one.)
RWA Scale. The average scores on the RWA Scale in the countries sampled by Morning Consult were: USA: 94.3; Canada: 84.0; United Kingdom: 87.1; Australia: 85.9; Italy: 84.8; Spain: 85.2; Germany: 83.0; and France: 89.2. The American score sticks out like a sore thumb. The biggest difference between the “Yanks” and the rest of the countries predictably occurred on an item that mentions abortion. But the American sample’s response to, “What our country really needs is a strong, determined leader who will crush evil and take us back to our true path” also stood out, compared to the others. As did, “Our country will be great if we honor the ways of our forefathers, do what the authorities tell us to do, and get rid of the ’rotten apples’ who are ruining everything.” And, “Our country will be destroyed someday if we do not smash the perversions eating away at our moral fiber and traditional beliefs.” But beyond these heavy hitters, differences popped up throughout the test. Americans scored higher than the rest of the respondents on 19 of the 20 RWA Scale items.
One notes the German sample scored lowest in authoritarianism. It particularly disagreed with the items mentioned above. Many Germans have apparently leaned, to their great sorrow, what comes from submitting to a strong leader intent on stomping out various groups.
The Social Dominance Scales. The mean scores on the 8-item Dominance Scale were: USA: 31.3; Canada: 29.4; U.K.: 32.6; Australia: 31.5; Italy: 31.2; Spain: 33.7; Germany: 34.5; and France: 34.7. The averages on the Antiequality Scale posted as 32.1, 29.0, 32.1, 30.5, 28.8, 30.6, 33.4 and 32.5 respectively. The German and French samples polled highest, and the Canadians the lowest overall.
Prejudice.[9] The average scores over the 24 prejudice items came in as follows: USA: 102.6; Canada: 89.2; U.K.: 100.3; Australia: 96.2; Italy: 98.7; Spain: 98.8; Germany: 99.8; and France: 105.2. But as noted earlier, the 24 statements were written to tap prejudice in the United States, and some of them would have little relevance elsewhere. As well, prejudices prominent in other countries—such as those against Asians—were hardly assessed at all.
We can’t gauge prejudices against groups that were not mentioned in the survey, but we can look past statements that have little relevance outside the United States to put the eight countries on a more level playing field. I decided to set aside Items 6, 15, 17, 19, 20 and 22 in Exhibit 1 on grounds of being “too American” or having weak connections with the rest of the test, leaving a balanced 18-item scale that features statements about White supremacy, Black lives, Muslims, Jews, immigration policies, and multiculturalism—which you might agree are relevant in all of the eight countries involved in the Morning Consult survey.
This more transferable measure of prejudice provoked the following means: USA: 77.1; Canada: 66.9; U.K.: 75.5; Australia: 72.8; Italy: 75.4; Spain: 74.6; Germany: 76.7; and France: 79.9. The Canadian score catches your eye. Whatever the results would have been if aboriginal groups had been mentioned, Canadians appear to be less prejudiced in favor of whites, less prejudiced against blacks, more open to multiculturalism and immigration, and so on than the other groups sampled. Australians also scored low (and they too likely benefitted by the omission of indigenous targets). France and the United States had the highest means, but not by much.
“Double Highs” and Prejudice. The level of hostility toward a minority group can be powerfully affected by the social norms of one’s environment. But personality variables play a role as well. Who are the most prejudiced people in a country? Over the years the best personality-based predictors of racial and ethnic prejudice have been scores on the Social Dominance and RWA Scales. In particular, people who score highly (i.e. in the top quartile) on both of those tests—whom I call “Double Highs”—have displayed more prejudice than anyone else. Most of them are thought to have strong inclinations toward dominance and would want to hold positions of power in authoritarian (and most other kinds of) organizations. They score highly on the RWA Scale, which asks about submitting to authorities, not because they are inclined to submit themselves, but because they like the idea of other people submitting to them. They comprise the pool of wannabe dictators that eventually produce the leader of totalitarian states and his minions.
In the Morning Consult study, the High RWA-High “Social Dominance” subjects (N=739 over all eight samples) once again took home the gold medal in the Prejudice Olympics with a mean score on our 18-item prejudice scale of 97.0. The (1279) High-SDO-But-Not-High-RWA subjects came in a distant second (88.5), and the (1197) High-RWA-But-Not-High-SDO’s had to settle for bronze with 86.4. Furthermore, the Double Highs had the highest prejudice scores in each of the eight nations, while the other groupings took turns sharing Second and Third place. The United States had more than its fair share of the Double Highs (19.5%) and Germany (9.3%) had the fewest.
Religious Fundamentalism
Despite rumors of its demise, religious fundamentalism is still alive and living in the United States more than anywhere else in this study. The American mean on the Religious Fundamentalism Scale equaled 56.8. In neighboring Canada it amen’d but 45.6, in the United Kingdom 45.0, in Australia 45.6, in Italy 49.8, in Spain 46.7, in Germany 45.8, and in France 48.8. Organized religion has taken a hit in all of these countries over recent generations, and various polls indicate even the number of Protestant evangelicals in the United States is slowly declining. (But non-evangelical Protestant ranks are thinning faster.) See https://www.pewresearch.org/religion/2019/10/17/in-u-s-decline-of-christianity-continues-at-rapid-pace/ .
Summary of Things Found in Weed Patch No. 2. Table 10 presents the average scores obtained of our five personality measures in the eight countries sampled. Overall, the Americans’ data leap out at one because they scored higher on our measures than anyone else. That’s ruling the roost. Unfortunately, it’s the roost in a totalitarian shed. We can also shed a tear for France. Things seem quite a bit better for democracy in Canada, which scored lowest or second-lowest on all five measures.
Table 10. The Mean Scores on the Five Personality Scales, by Country
Country | RWA | Dominance | Antiequality | Prejudice-18 | Rel. Fundamen. |
USA | 94.3 | 31.3 | 32.1 | 77.1 | 56.8 |
CANADA | 84.0 | 29.4 | 29.0 | 66.9 | 45.6 |
UNITED KINGDOM | 87.1 | 32.6 | 32.1 | 75.5 | 45.0 |
AUSTRALIA | 85.9 | 31.5 | 30.5 | 72.8 | 45.6 |
ITALY | 84.8 | 31.2 | 28.8 | 75.4 | 49.8 |
SPAIN | 85.2 | 33.7 | 30.6 | 74.6 | 46.7 |
GERMANY | 83.0 | 34.5 | 33.4 | 76.7 | 45.8 |
FRANCE | 89.2 | 34.7 | 32.5 | 79.9 | 48.8 |
One may not be surprised that so many religious fundamentalists filled the American ranks. But shouldn’t we be amazed that the USA sample scored so highly in RWA and prejudice, and had more of the particularly dangerous Double Highs than anyone else? After all, the United States is, economically and militarily, the redoubt of democracy on this planet. Surely things would be better if its “national personality” led the pack in being UNauthoritarian and UNprejudiced.
Every nation in the Morning Consult study has issues and is experiencing discord to some degree. But America is visibly writhing as it wrestles with its inner demons. We might live to see the day that the United States collapses into an angry, revenge-driven, authoritarian savagery. How unthinkable is that? For decades, research on authoritarianism seemed to me a worthy enough academic pursuit, but not obviously relevant to current affairs. Yes, I found lots of highly authoritarian people in my studies, but they were not organized and marching down the street at night with torches ablaze. And they didn’t have a leader. But now the wolf is at the door. We came very close to losing it all in January 2021. Certain individuals (I’m thinking of Vice-President Pence, General Milley, various judges, and the main-stream media in particular) saved the country then. I think it likely our commitment to democracy will be tested again soon.
Correlations Found Among Our Scales in the Nations Surveyed
Time to forge ahead into our last field of weeds. Table 11 shows the inter-correlations among our personality scales obtained in the four English-speaking countries. Scrutinizing the American results, I immediately noted the connections between the RWA Scale and the two measures of Social Dominance Orientation (.461 and .442 respectively). As noted earlier, once upon a time these two constructs had loosey-goosey correlations, leading to the belief that SDO tapped the authoritarian leader personality while the RWA Scale identified authoritarian followers. The goose is not so loose in these data. When one combines the two SDO scores, “SDO-16” correlates .492 with RWA, meaning they shared 24% of their variance. They used to share about 5%-10%. Something has changed.[11]
The connections with prejudice reinforce this suspicion. SDO scores used to correlate stronger than RWA with prejudice measures. That made all the sense in the world to me, given the content of the scales. Attitudes toward whites, blacks, immigrants, and so on might connect to attitudes about authoritarian submission, authoritarian aggression, and conventionalism. But beliefs in group dominance and equality ought to slam dunk. However now, as also happened in the Monmouth survey, the SDO correlations with prejudice have noticeably slipped. RWA scores correlated .709 with prejudice in the United States, accounting for most (by the hair on its chinny-chin-chin at 50.3%) of the variance in prejudice scores by itself. SDO-Dominance pegged at .614, and SDO-Antiequality came in at .645. Their combined score correlated .687 with prejudice.
Table 11. Inter-correlations Among the Personality Scales in English-Speaking Countries USA
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. |
RWA | – | .461 | .442 | .709 | .751 |
SDO-Dominance | – | .681 | .614 | .329 | |
SDO-Antiequality | – | .645 | .256 | ||
Prejudice (18) |
|
.493 |
CANADA
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. |
RWA | – | .550 | .454 | .673 | .711 |
SDO-Dominance | – | .746 | .658 | .371 | |
SDO-Antiequality | – | .633 | .331 | ||
Prejudice (18) | – | .478 |
UNITED KINGDOM
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. | |
RWA | – | .590 | .517 | .708 | .620 | |
SDO-Dominance | – | .743 | .605 | .456 | ||
SDO-Antiequality | – | .595 | .340 | |||
Prejudice (18) | – | .431 |
AUSTRALIA
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. |
RWA | – | .509 | .492 | .667 | .670 |
SDO-Dominance | – | .728 | .588 | .454 | |
SDO-Antieq1uality | – | .626 | .408 | ||
Prejudice (18) | – | .471 |
Combining the two SDO scores with RWA in a multiple regression analysis churned out a multiple-R of .810, boosting the combined predictive power of these personality scales to nearly 66%. So the chest-thumping, vine-swinging conclusion that authoritarianism can explain most prejudice among Americans was replicated. That’s not a shabby accomplishment in the behavioral sciences. But in olden times it would have been even more, I think. RWA and SDO are not explaining the different facets of prejudice in these data as much as they used to.
Religious fundamentalism had a well-precedented and sizeable relationship with prejudice, .493. But this was all but wiped out by the shared connections with RWA. As has been found in many earlier studies, evangelicals are not prejudiced so much because of their religious beliefs (which can turn them in quite the opposite direction), but because they are so often authoritarian followers.
The Canadian Findings
The connections between SDO and RWA thumped in even stronger in Canada (.550 for Dominance; .454 for Antiequality), and equaled .534 for the sum of all 16 SDO items. And the correlations with prejudice reminded me a bit of days past. While the RWA Scale again had the strongest relationship with prejudice (.673), it fell short of its performance in the American sample, and the summed score on the SDO items correlated better (.690) with prejudice than RWA did. The multiple-R equaled .779, so once again SDO and RWA could explain most (61%) of the variance in the prejudice responses. Time for another bellow and ride on a vine.
While in the United Kingdom
Crossing the Atlantic, we find that the unhelpful dominance and antiequality correlations with RWA proved substantially stronger (i.e., worse for our purposes) in Great Britain than they had been in North America, equaling .591for the combined 16 social dominance items. RWA correlated with prejudice almost exactly as well (.708) as it had in America, but the SDO scales again fell short of their former glory at explaining these destructive sentiments. Combining the latter’s 16 items produced a correlation of .642 with prejudice, and together the RWA and Social Dominance responses produced a multiple regression correlation of .761 with prejudice, explaining 58% of the latter. Calling Lord Greystoke.
And in Australia
Rounding out our probe of the English-speaking samples, the Australian numbers tucked in with those obtained in our first three countries. We again find unhappily strong connections between the social dominance scores and RWA, equaling .548 when the dominance and antiequality items were combined. RWA again pulled down the best connection with prejudice (.667), with the combined social dominance responses right behind at .654. The multiple R equaled .754, explaining 57% of the prejudice variance. Here come the elephants.
The Results in Italy
Table 12 shows the correlations among our scales obtained in the four “European” countries in the Morning Consult survey. They are smaller than those in Table 11, undoubtedly so because of translation problems and greater cultural differences. But the figures from Italy show a familiar pattern, down to the elephants. The combined SDO-RWA correlation still rode high (.536), and RWA correlated a bit higher with prejudice (.651) than either of the social dominance measures—but only the tiniest bit higher when the 16 social dominance responses were combined (.647). The multiple regression prediction of prejudice equaled .740, accounting for 55% of the latter.
Table 12. Intercorrelations Among the Personality Scales in the “European” Countries ITALY
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. |
RWA | – | .516 | .489 | .651 | .683 |
SDO-Dominance | – | .751 | .611 | .381 | |
SDO-Antiequality | – | .600 | .384 | ||
Prejudice (18) | – | .454 |
SPAIN
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen |
RWA | – | .537 | .518 | .656 | .660 |
SDO-Dominance | – | .690 | .609 | .470 | |
SDO-Antiequality | – | .644 | .496 | ||
Prejudice (18) | – | .598 |
GERMANY
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. |
RWA | – | .465 | .463 | .545 | .636 |
SDO-Dominance | – | .632 | .400 | .398 | |
SDO-Antiequality | – | .478 | .418 | ||
Prejudice (18) | – | .379 |
FRANCE
Scale | RWA | SDO-Dominance | SDO-Antiequality | Prejudice (18) | Religious Fundamen. |
RWA | – | .443 | .415 | .579 | .481 |
SDO-Dominance | – | .699 | .518 | .355 | |
SDO-Antieq1uality | – | .507 | .414 | ||
Prejudice (18) | – | .310 |
Spain
Welcome to Canada, at least as far as the connections among our scales goes. RWA and summed scores on the social dominance items correlated .572, and while RWA predicted prejudice slightly better (.656) than SDO-Dominance or SDO-Antiequality, it did not predict better than SDO-Dominance and SDO-Antiequality (.683). Antiequality, Dominance, and RWA produced a multiple-R of .758 with prejudice, explaining 57% of its variance.
Germany
The numbers tumbled here, but they tell much the same story. RWA and combined social dominance scores correlated .513, and yet again RWA scores provided stronger connections with prejudice (.545) than either of the two social dominance scales, and their 16 items taken together (.488). The multiple-R predictor of prejudice only equaled .603, which can account for 36% of the latter’s results. (Quite a good show in most studies, but we are spoiled.)
France
The correlations also floundered in France, but still formed the familiar pattern. RWA and total Social Dominance correlated .464, and the former connected with prejudice better (.579) than Dominance, Antiequality, or the Combo (.555). The multiple regression analysis produced an r = .663 with prejudice, accounting for 44% of its variance.
Summary. In all eight countries, the RWA and the social dominance scales correlated much higher than expected (and desired). In all of the samples the Dominance and Antiequality scales correlated less with prejudice than the RWA Scale did, despite the strong connect-ability one would expect between wishes for group dominance and inequality on the one hand and prejudice on the other. Combining the two 8-item scales produced a stronger connection with prejudice, but even so it only did better than RWA in two of the eight countries. It used to be stronger. I suspect it would have been if the apparent effect of response sets had played less a role.
But let’s put this in perspective. Back in the 1960s (you may recall, but probably do not) a researcher was pretty thrilled to get a correlation in the .40s between a personality measure and something else. I’m now quibbling about how green the grass is in the Personality Test Promised Land. In most of the countries sampled, despite everything, these measures of what I think is follower and leader authoritarianism explained most of the differences in prejudice found in large, representative samples of their populations. Together they kicked butt.
I can’t imagine this will prove true everywhere. After all, 193 countries belong to the United Nations, and whatever their differences, the samples in the Morning Consult study shared an awful lot of cultural background that differs quite a bit from, say, China’s and India’s and Zaire’s and Margaret Mead-dom’s. But let’s enjoy the moment. Thanks to the scientific method, our scales have scaled at least some mountains.
Appendix
Table 1 . Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: USA (N=1001)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-178 | 94.3 |
773 |
.299 | .896 | |
SDO-DOM | 8 | 8-72 | 8-65 | 31.3 | 110 | .237 |
. 717 |
|
SDO-ANTI | 8 | 8-72 | 8-70 | 32.1 | 145 | .356 | 813 | |
PREJUDICE | 24 | 24-216 | 24-203 | 102.6 | 1108 | .303 | .913 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-108 | 56.8 | 478 | .446 | .906 |
Table 2 . Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: CANADA (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-170 | 84.0 |
656 |
.317 | .904 | |
SDO-DOM | 8 | 8-72 | 8-59 | 29.4 | 101 | .300 |
. 775 |
|
SDO-ANTI | 8 | 8-72 | 8-72 | 29.0 | 125 | .432 | .853 | |
PREJUDICE | 24 | 24-216 | 24-208 | 89.2 | 1130 | .396 | .940 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-108 | 45.6 | 449 | .524 | .930 |
Table 3 . Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: UNITED KINGDOM (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-156 | 87.1 |
522 |
.269 | .882 | |
SDO-DOM | 8 | 8-72 | 8-71 | 32.6 | 91 | .280 |
. 758 |
|
SDO-ANTI | 8 | 8-72 | 8-64 | 32.1 | 105 | .361 | .814 | |
PREJUDICE | 24 | 24-216 | 24-216 | 100.3 | 1063 | .361 | .932 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-107 | 45.0 | 368 | .475 | .916 |
Table 4. Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: AUSTRALIA (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-176 | 85.9 |
607 |
.291 | .892 | |
SDO-DOM | 8 | 8-72 | 8-68 | 31.5 | 98 | .274 |
. 753 |
|
SDO-ANTI | 8 | 8-72 | 8-64 | 30.5 | 116 | .388 | .830 | |
PREJUDICE | 24 | 24-216 | 24-191 | 96.2 | 1076 | .350 | .929 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-108 | 45.6 | 424 | .500 | .923 |
Table 5. Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: ITALY (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-173 | 84.8 |
523 |
.211 | .843 | |
SDO-DOM | 8 | 8-72 | 8-67 | 31.2 | 108 | .278 |
. 755 |
|
SDO-ANTI | 8 | 8-72 | 8-69 | 28.8 | 130 | .375 | .821 | |
PREJUDICE | 24 | 24-216 | 24-200 | 98.7 | 1111 | .352 | .929 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-108 | 49.8 | 308 | .335 | .858 |
Table 6. Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: SPAIN (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-172 | 85.2 |
504 |
.206 | .840 | |
SDO-DOM | 8 | 8-72 | 8-72 | 33.7 | 78 | .192 |
. 660 |
|
SDO-ANTI | 8 | 8-72 | 8-72 | 30.6 | 119 | .339 | .800 | |
PREJUDICE | 24 | 24-216 | 24-192 | 98.8 | 888 | .283 | .905 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-105 | 46.7 | 298 | .332 | .850 |
Table 7. Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: GERMANY (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 20-172 | 83.0 |
433 |
.181 | .818 | |
SDO-DOM | 8 | 8-72 | 8-57 | 34.5 | 63 | .151 |
. 584 |
|
SDO-ANTI | 8 | 8-72 | 8-72 | 33.4 | 79 | .249 | .722 | |
PREJUDICE | 24 | 24-216 | 24-207 | 99.8 | 1111 | .357 | .931 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-107 | 45.8 | 310 | .367 | .874 |
Table 8. Statistical Properties of Five Personality Tests Used in the Morning Consult April 2021 Survey
Sample: FRANCE (N=1000)
Scale | Number
Of Items |
Possible
Range |
Actual
Range |
Mean |
Variance |
Mean
Inter-Item Correlation |
Alpha
Reliability Coefficient |
|
RWA | 20 | 20-180 | 23-159 | 89.2 |
356 |
.126 | .747 | |
SDO-DOM | 8 | 8-72 | 9-67 | 34.7 | 74 | .158 |
. 606 |
|
SDO-ANTI | 8 | 8-72 | 8-69 | 32.5 | 101 | .272 | .746 | |
PREJUDICE | 24 | 24-216 | 24-205 | 105.2 | 825 | .250 | .890 | |
RELIGIOUS
FUNDAMENTALISM |
12 | 12-108 | 12-104 | 48.8 | 269 | .292 | .831 |
Footnotes
[1] John W. Dean and Bob Altemeyer, Authoritarian Nightmare, 2019, Brooklyn, NY Melville House, Chapter Ten.
[2] Which is north of North Dakota, if you can imagine such a thing.
[3] You can find them in Chapter One of the free book available on this website.
[4] What? Imagine the Smith family has 20 children. (Most of them were adopted.) How well do they all get along with each other? Aggie has a relationship with 19 siblings, and Bill has connections with 18, plus the one with Aggie. If you work it out (19+18+17+16+…+1) there are 190 relationships percolating at a Smith Family Progeny Picnic. Some of them may be pleasant and positive, and others may be positively negative because some Smith offsprung don’t like each other. Whether positive or negative, the connections will also vary in intensity. Aggie likes Bill and especially likes Cathy, but wishes Donald would drop dead!
We can get an idea of how harmonious the Smith kids are by combining all the feelings swirling around and dividing by 190—the number of relationships we’ve added up. This average (statisticians call it a “mean”) lets us compare the Smiths to families of different sizes, the way batting averages let us compare hitters who have different numbers of at-bats.
If you’re with me so far, we’re rounding third and heading home. On a psychological test, the “feeling” two items share is called a correlation, and in the American sample virtually all of the feelings on the RWA Scale were positive (i.e., there were only four, tiny, negative correlations out of 190). Their average, which could have gone from 0 to 1.00, was .299. That may seem pretty piddling, but the statements on the test cover a wide range of topics. They’re like a family with lots of adopted kids who spent their early years in different cultures. The real story in the family, and the scale, is, “Why is there as much positive inter-connecting as there is, despite all the obvious differences? Why do they go together as much as they do?” The answer tells you what they are about, deep down inside.
[5] What’s “alpha”? It’s a number that again can go from 0 to 1.00, which tells you how stable, how reliable, how trustworthy the score on a test is. It is determined by A) your new friend, the average correlation among the items on the test, and by B) how many items there are on the test. Longer tests have more stable results than shorter ones, if the items interconnect as well. (And that’s the trick.) Each item gives you a hint about what you’re trying to find, and twenty hints are a lot better than one. The neat thing about the alpha coefficient is it tells you how much of the information you get from a test is “signal” and how much is “noise.” The alpha coefficient of the RWA Scale in the American sample was .896. That means the signal-to-noise ratio in the data was about 9 to 1. You’d be very unhappy with a TV whose picture had 10% “snow,” but it’s quite good for a psychological test. The best IQ tests have alphas about .90. In personality testing, .90 is an “A” or even an “A+.” It’s an alpha alpha.
[6] For whatever reason the Canadian data were least affected by response sets. Answers to the pro-trait RWA Scale items intercorrelated .474 on the average, and those among the con-traits .400. The summed pro-traits correlated .430 with the summed con-traits. If you’re thinking we can avoid this problem by wording all the statements in the pro-trait (or con-trait) direction, don’t go there. The response sets will still occur, increasing the connection among the items, thus creating a false impression of how firmly the trait exists in people’s thinking. And it will artificially increase the connection between that scale and any other test so unidirectionally worded. This happened with an early measure of authoritarianism, the F Scale, with disastrous consequences for research in the area for over 20 years (Bob Altemeyer, Right-Wing Authoritarianism, 1981, Winnipeg, University of Manitoba Press, Chapters 1 and 2). It turned out the F Scale was mainly triggering response sets, mostly nay-saying in this case because the F Scale’s items were so quirky. Balancing a test with pro-trait and con-trait items cannot stop response sets from affecting answers. (But it can keep most of them from influencing the summed score on the test, a worthy outcome.)
[7] Social psychology lost one of its brightest stars when Jim Sidanius left us in June 2021. See https://www.google.com/search?client=firefox-b-d&q=jim+sidanius+obituary .
[8] Factor analysis, for those who know their way around communalities and eigenvalues, also demonstrates how powerfully response sets affected the data. A confirmatory principle axes analysis (not the much less informative exploratory principle components solution) found two factors underlay the responses of the American sample to the RWA Scale. All of the pro-trait items had their higher loading on Factor I, and all of the con-trait items loaded higher on Factor II. This was also true for both Social Dominance scales, the Prejudice measure, and the Religious Fundamentalism Scale. And it was also true in all the other samples for all the scales used. And I have found two direction-of-wording factors in virtually every factor analysis of balanced scales I have conducted for the past 50 years. If you do not control for their effect, response sets will pull your pants down in test-based research, and you won’t even know it.
[9] You will recall that the Morning Consult polls only asked for racial identification in the United States. So all of the analyses presented here are based upon the total samples in each country, which will include some “nonwhite” respondents as well as a preponderance of whites. In the USA, 81.4% of the respondents said they were white, and they were a touch more prejudiced overall than the non-white Americans. (But their Likes and Dislikes turned up on different items, as you might expect.)
[10] To enable comparisons with previous results, I have combined the Dominance and the Anti-equality scores as “Social Dominance.”
[11] The energetic and careful Jake Womick at the University of North Carolina gathered American data on the (complete!) RWA and SDO scales in 2016. Unfortunately, Dr. Womick was not lucky enough to have met someone famous (as I did) who could get a major polling firm to do his survey. So he went with Amazon Turk, which collects data from people who take lots of surveys for a pittance. The respondents left something to be desired, collectively, having many more well-educated persons, many more women, and many more agnostics or atheists than a representative sample would contain. The SDO-Dominance items were presented on one page, and the Anti-equality items on another. Two different samples produced RWA alphas of .95 and .96, Dominance alphas of .94 and .94. and Anti-equality alphas of .94 and .94. RWA correlated .43-.46 with the Social Dominance scores. RWA made appreciably stronger predictions of preference for Donald Trump than the SDO scales. (See Studies 1 and 2 in Jake Womick, Tobias Rothmund, Flavio Azevedo, Laura A. King, and John T. Jost, “Group-Based Dominance and Authoritarian Aggression Predict Support for Donald Trump in the 2016 U.S. Presidential Election,” Social Psychological and Personality Science, 2018, 1-10.)