How racial groups rate each other

The plot below reports the mean rating from Whites, Blacks, Hispanics, and Asians of Whites, Blacks, Hispanics, and Asians, using data from the preliminary release of the 2020 ANES Time Series Study.

---

NOTES

1. Data source: American National Election Studies. 2021. ANES 2020 Time Series Study Preliminary Release: Combined Pre-Election and Post-Election Data [dataset and documentation]. March 24, 2021 version. www.electionstudies.org.

2. Stata code. Stata output. R code for the plots. Dataset for the R plot.

Tagged with: ,

64 Comments on “How racial groups rate each other

  1. Has anyone offered any interpretations of this pattern?
    Isn't there a narrative that Whites think of themselves more favorably than other groups. This pattern seems to contradict that. What's going on?

    • Hi Robert,

      ANES data indicate that White ingroup bias relative to ratings of Blacks has been declining over time (https://www.ljzigerell.com/?p=8168). Work by Zach Goldberg indicates that White liberals now have a racial outgroup preference ("America's White Saviors", at the link), and this at least somewhat offsets a racial ingroup preference among White conservatives (https://www.tabletmag.com/contributors/zach-goldberg). Zach's "How the Media Led the Great Racial Awakening" article at the link discusses potential media influence on over-time change in racial attitudes (https://www.tabletmag.com/contributors/zach-goldberg).

      Patterns from items directly asking participants to rate racial groups can be interpreted only so much. The relative lack of net ingroup bias among Whites is consistent with other survey experiment work (https://journals.sagepub.com/doi/full/10.1177/2053168017753862), but I wouldn't interpret any of these results to cover real-world discrimination, especially discrimination detected in field experiments (https://www.aeaweb.org/articles?id=10.1257/0002828042002561).

    • Hatred, dismissed as fighting white privilege, has been normalized and expected in a higher-education environment. Any minority in disagreement with it will affect their social life and grades while attending college.

      • All the criteria for white privilege can be said about Asian Americans. Asians make more money, get higher education, are more often self employed, have higher IQ, commit less crime, less divorce rates, more likely to have two parents and so much more but nobody says Asian privilege. It seems like hard work in their culture privilege, just like it is in a lot of white households. The whole White privilege thing is only for lazy idiots to spout.

        • It isn't so much lazy idiots, per se, as simply anti-White.

          And did you notice the survey when it was asking about family values? It labeled that under "authoritarianism". Values that promote a strong and stable White family are "authoritarian", and so negative.

          Now, do you know who were critical in creating that concept in psychology? Like, "The Authoritarian Personality", if you want to phrase it like that ...

      • I don't know if anyone has brought this up yet, but age is also very important in these results as well. Was there anything in your studies that showed a particular age the respondents were?

        • Good idea, George. Average ages in the dataset were 50 among Whites, 46 among Blacks, 41 among Hispanics, and 41 among Asians.

          I calculated a measure of ingroup bias as a respondent's rating about their own racial group minus the respondent's average rating about the other three included racial groups. So, for example, for a White respondent, the number for ingroup bias is the White respondent's rating about Whites minus the White respondent's average rating about Blacks, Hispanics, and Asians.

          For respondents aged 18 to 30, the ingroup bias is -4 among Whites (N=571), +26 among Blacks (N=91), +11 among Hispanics (N=164), and +14 among Asians (N=46). The negative ingroup bias among Whites means that the rating about Whites was lower than the average rating about Blacks, Hispanics, and Asians.

          For respondents aged 31 to 50, the ingroup bias is -1 among Whites (N=1,424), +17 among Blacks (N=213), +9 among Hispanics (N=245), and +10 among Asians (N=97).

          For respondents aged 51 and older, the ingroup bias is +4 among Whites (N=2,695), +17 among Blacks (N=235), +10 among Hispanics (N=175), and +13 among Asians (N= 80).

          Confidence intervals and other output are at: https://www.ljzigerell.com/wp-content/uploads/2022/07/FT-age.txt

      • I’m curious what percentage of the people who took the survey decline to answer the “how would rate Race”questions? was there any form of lie detectors to verify if their rating was how they truly felt? What was the political demographic of people and what was there rating? Is there a direct correlation between other questions answered and their rating of race X?

        • Hi Andrea. The response rate for the items rating racial groups is pretty high. See, for example, around page 4 here: https://www.ljzigerell.com/wp-content/uploads/2021/03/ANES-2020-TS-How-racial-groups-rate-each-other.pdf.

          The overall sample size for the pre-election survey and the post-election survey was 8,280 respondents, but 750 respondents who took the pre-election survey did not take the post-election survey that the items rating racial groups were on, and there were another 77 and 87 respondents whose data were deleted or who broke of the post-election survey before getting to the items rating racial groups. So, of the 7,366 respondents who received the item about rating Blacks, only 263 refused to provide a numeric rating (so that's a 3.6% refusal).

          No lie detectors, so inferences are better thought of as representing what respondents report instead of their true feelings, although presumably social desirability is lower for unequal ratings of racial groups.

          And the ratings of the racial groups tend to associate with outcomes as expected, such as colder ratings of Blacks than Whites predicting opposition to national anthem protests and colder ratings of Whites than Blacks predicting support for national anthem protests (https://www.ljzigerell.com/?p=12537).

  2. Im not sure I would interpret the results of the AEA study as having detected real-world discrimination.

    To indicate discrimination, we must assume that employers view black and white applicants with the same resume qualification(s) as equally productive employees. But that’s far from obvious given widely known confounders such affirmative action.

    • Hi D.B. If, for example, employers discount individual applications with stereotypically Black names because of real or perceived group-level characteristics such as an affirmative action boost to Black applicants in selection to elite academic institutions. I think that it's reasonable to refer to that as "discrimination". I realize that statistical discrimination conducted correctly can produce more fair outcomes at the group level and can lead to better outcomes: for example, if a higher percentage of men than women are carrying contraband, then suspicionless searching a higher percentage of men than women can lead to discovery of more contraband, compared to if the suspicionless searching were conducted randomly by suspect sex. But I think that it's reasonable to consider statistical discrimination to be unfair at the individual level: if an employer who didn't call me for a job interview would have called me but for my race, I think that it's reasonable for me to consider that to be unfair at the individual level.

    • Not hard to believe. This is why it's only non-whites immigrating to white majority countries and nobody has an issue with it. It's why whites are statistically killed the most by non-whites when it comes to interacial murder and don't cause an uproar, it's why you don't see whites collectivize for white interests. It's why whites are cool with their declining birthrates and migration replacement. It's why you see more whites at BLM rallies even though more of them are killed by police once factoring in crime rate disparities. Whites are the least tribal group, and that is quite clearly being taken advantage of.

      • Uh sweetie, that has nothing to do with anything, Eastern Europeans are last time I checked, white and they are mass migrating all over the world. And before them Europeans mass migrated all over the world. It's how Europeans even got to the Americas which was Native land. When whites are killed by police it's usually because the incompotence of the American police force, and not due to discrimination, racial profiling is REAL. You I take it are white, so you don't have a single clue the racial bias non-whites deal with from authorities on a daily basis. You are not opressed, and it's time to stop pretending you are. Whites don't cause uproars because you are not a victim, never have been a victim. You nor your people suffered from systematic oppression that still effects black communities to this day. What you deal with on a daily basis is not racial discrimination. I know you want so bad for your atrocities of the past to go away, but sorry honey, those effects will linger for hundreds of years and cannot be washed away overnight. Black people are "tribal" because you FORCED them to be for decades, you even had systems in place to keep the educated blacks in ghettoes and now you're upset about a culture you fostered when you could have simply treated the people better, and then maybe they would have felt better about you today or trusted you. Hun, giving blacks a few pats on the back is not going to make up for what you did to them, old wounds aren't erased like that, it's time you realize that. They don't owe you anything for trying to make up for your ancestors wrong doing, You want a pat on the back for that, you want to be seen as the hero after being the villain for centuries, it's not happening.

        • Comical to hear people talk about their own race and how other races don’t understand them yet feel entitled and egotistical to think they know what it’s like to be everyone else’s race. You’re part of the problem! Until you’ve walked one step in another persons shoes you have zero idea of who they are, what they’ve been through and the struggles they’ve had to overcome. Life is rough regardless of skin color and people like you with your narrow point of view just make it harder

        • Blacks aren't oppressed. They're not victims. They're privileged.

          They're given special advantages through affirmative action, preferential hiring treatment, race quotas, race-exclusive scholarships, exclusive career opportunities, etc.

          And no one here has been alive for centuries. Get your brain checked.

        • Uh sweetie yourself. You don’t know us as well as you think. No one group is more ostracized than White Americans. We have to be on our guard 24/7. How can we trust anyone, when you don’t trust us? This isn’t 1960. Times have changed and today’s European Americans are under constant scrutiny and segregation more than others. Quit trying to sound sophisticated and like you know something we don’t. Your data and references show your age. Have you ever been around Appalachian Americans? Do you realize that White Americans account for more people living under the poverty belt than any other group? Your goals aren’t equality but displacement and power.

        • Hi ‘sweetie’,

          Firstly, stop dehumanizing white people by claiming ‘their’ atrocities. No one alive today was responsible for the slave trade. Secondly, African rulers and governments benefitted massively from the slave trade and initiated it in most instances. Thirdly, if you have issues with the systemic racism in America that has blacks sequestered in ghetto environments then you need to be looking at who has benefitted from that societal segregation. It’s the Democrats. The same Democrats who used redlining and the FHA to segregate America and deny blacks access to the means to create intergenerational wealth through ownership of real estate. The same Democrats the black community has been voting in as their representatives for over 100 years. The same Democrats that are mayors in the cities that blacks are kept in and governors of the states with the largest urban population centers of blacks. The same Democrats who drop pithy handouts and ask for our votes with one outstretched hand while imprisoning us with the other. ‘Honey’, you should spend less time being condescending to people on the internet and more time learning your history. You are a tool for the oppressors, a tool for division, being used just as they made you - and you willingly allow it with your willful ignorance. You are happy to be used, and in return you get the blessings of superiority over those you have been turned against. You are a fool. Do better.

        • Damn, sweaty ur retarded and full of hate. Ur not blind to the truth. Ur just too dumb to get it. Cope and seethe jealous minority.

  3. What kinds of studies would persuade you of discrimination?
    Audit studies where the same work products are attributed to people of different groups (e.g., Black, White), and the products are judged differently depending the producer's supposed group?

    • Hi. I saw your graph used on FB today to claim that the biggest racists in the US are groups other than white people. Could you possibly comment on that?

      • Hi David. The plot of mean ratings about racial groups by race depicts group-level ratings, so individual-level analyses would be more appropriate for making inferences about individuals. I have a plot of individual-level analyses indicating that a higher percentage of Whites than of Blacks, Hispanics, and Asians rated all four included racial groups equally (https://www.ljzigerell.com/wp-content/uploads/2021/03/EQ-by-race.png).

        I think it's fine to describe unequal ratings about racial groups descriptively as "racial favoring" or "racial bias", but "racism" is such a disputed term that involves value judgments that I wouldn't want to classify anything as racist without a definition of "racism".

        Moreover, if the intent is to measure a general tendency of individuals or groups, I'd rely on more than a particular measure of racial attitudes from a particular survey. But, given that intent, I think it's a good idea to incorporate results from racial feeling thermometers.

        • So, let me put the question another way. Are you happy with people using your graph on social media to claim that this proves whites are less racist than other groups? What would you say to those people, and can I Quote you?

      • Hi David. For your March 31 comment, I think that it's fair to quote in context anything that I publicly post. My prior comment in this thread discussed the claim that you asked about. I'm not interested in adding a description of my feelings about the claim. Have a nice day!

        • Thanks. Very helpful. My questions were prompted by a genuine desire to boost my understanding. I certainly wasn’t trying to tie you down to a personal opinion. You’re right it is a very touchy area, and somewhat ill defined term. Keeping things objective is important. Cheers.

  4. That’s a good question. As I understand you, yes, your example strikes me as the more efficacious of the two models. I mean surely taste-based discrimination occurs.

    My original point was simply that we too often conflate what are otherwise rational decision processes with a more invidious type of behavior. Assuming or expecting people to willfully commit base rate neglect strikes me as an untenable position. In any case, yes I think yours would be an interesting study and I wouldn’t be surprised if TB discrimination was found.

  5. Have you tried asking respondents how much of the legacy media do they watch ?
    I feel like that might be a big factor in why so many have a relatively negative view of whites.

  6. Hi, Really interesting chart! Why did you use error bars at the 83% level? Are the differences in scores not statistically significant at the 95% or even 90% level?
    Best,
    DC

    • Great question, Dave!

      95% confidence intervals are useful for assessing whether there is sufficient evidence at a p-value threshold of p=0.05 to claim that an estimate differs from a particular number. For example, the 95% confidence interval for Black respondents' rating of Whites is [59.4, 65.0], which means that there is sufficient evidence that that estimate differs from 59.3 at the p=0.05 threshold but there is not sufficient evidence that that estimate differs from 59.5 at the p=0.05 threshold: if the particular number is inside the 95% confidence interval, then there is sufficient evidence at p=0.05 that the estimate differs from that number; but if the number is not inside the 95% confidence interval, then there is not sufficient evidence at p=0.05 that the estimate differs from that number.

      However, assessing differences between two estimates at p=0.05 doesn't work as well with 95% confidence intervals, because two 95% confidence intervals can overlap a bit and the difference still have a p-value less than p=0.05. 83.4% confidence intervals are better for helping readers use confidence interval overlap/non-overlap to assess p=0.05 differences between estimates. Here is a related blog post: https://chris-said.io/2014/12/01/independent-t-tests-and-the-83-confidence-interval-a-useful-trick-for-eyeballing-your-data/.

      So I tend to report 95% confidence intervals when the comparison of interest is to a particular number, such as whether a particular association differs from zero. But I tend to report 83.4% confidence intervals when the comparison of interest is between estimates.

      I put some statistical output here to illustrate this: https://www.ljzigerell.com/wp-content/uploads/2021/03/83.4.pdf. For this illustration, I'm using unweighted data from the ANES 2020 Time Series Study. Among Black respondents, the 95% confidence intervals are [63.0, 67.0] for the rating about Whites and [66.0, 69.8] for the rating about Asians; these 95% confidence intervals overlap a bit, but the p-value is p=0.04 for a test of the null hypothesis that these estimates equal each other. The 83.4% confidence intervals don't overlap: [63.6, 66.4] and [66.5, 69.3], so the non-overlap of the 83.4% confidence intervals will send the correct signal that these estimates differ from each other at p=0.05.

      • Thank you for the detailed explanation! You really went above and beyond what I usually see on the internet and that means a lot- the statistical output was especially useful. Thank you!

  7. I wonder what happens in the long run to the only team in the world that doesn't put its own members first.

  8. Sir, I had a question regarding the hispanic category. I understand “hispanic” has become a rather racialized category in the US, but they still come in every color and appearance. Is there any data in your study regarding how color or race categories in the hispanic respondents affected the study? For example, we’re white hispanics more likely than other Hispanics to view whites positively, but themselves (Hispanics) or another minority race negatively? Or were Hispanics fairly united across the board?

      • Hi Philip. Interesting question!

        The ANES 2020 Time Series dataset that I used currently provides information for a "summary" race variable that coded participants into only one of a set of categories such as non-Hispanic White, non-Hispanic Black, Hispanic, etc. The current release of that dataset lists as restricted the individual race/ethnicity variables (such as White/non-White, Black/non-Black, etc), so that I'm not aware of a way to construct a variable such as White Hispanic with the current release.

        The ANES *2016* Time Series data have data for variables that I can use to make such a variable, so I created two variables in these data from four years ago: a "White Hispanic" variable (coded 1 for participants who self-identified as White and as Hispanic and did not self-identify as Black, American Indian/Alaska Native, Asian, or Native Hawaiian/Other Pacific Islander), and a "nonWhite Hispanic" variable (coded 1 for participants who self-identified as Hispanic and did not self-identify as White).

        I also coded three "ingroup favoring" variables for each participant: "FTha" (rating about Hispanics minus rating about Asians), "FThb" (rating about Hispanics minus participant rating about Blacks), and "FThw" (rating about Hispanics minus rating about Whites). So zero indicates an equal rating of the two groups, positive numbers indicate a higher rating about Hispanics relative to the other group, and negative numbers indicate a lower rating about Hispanics relative to the other group

        Results in the output provide enough evidence to conclude that White Hispanics on average favored Hispanics in each of the three comparisons (compared to Asians, to Blacks, and to Whites) and that nonWhite Hispanics also favored Hispanics in each of the three comparisons (compared to Asians, to Blacks, and to Whites). Sample sizes were fairly small for the analyses, so these data aren't enough to conclude with much confidence that White Hispanics had stronger or less strong ingroup favoring than nonWhite Hispanics did, but, for each of the three comparisons, the point estimate for ingroup favoring was stronger for nonWhite Hispanics than for White Hispanics.

        Output is here: https://www.ljzigerell.com/wp-content/uploads/2021/03/Hispanic-FT-2016-ANES.pdf

  9. How many people participated in this study? Where do the participants come from, all over the US? I can't find this information anywhere on their website.

    • Hi Abigail. The description of the survey sample is under "Sample Design and Respondent Recruitment" in the User Guide and Codebook: https://electionstudies.org/data-center/2020-time-series-study/. Sample sizes for a particular analysis depend on the racial group rated and the racial group that made the rating. Samples can be found in the output: https://www.ljzigerell.com/wp-content/uploads/2021/03/ANES-2020-TS-How-racial-groups-rate-each-other.pdf. For example, "Subpop. no. obs = 5,157" in Line 40 indicates that the sample for the estimate of the mean rating of Whites among White respondents was 5,157. Line 41 indicates that the sample for the estimate of the mean rating of Blacks among White respondents was 5,178. Some participants did not respond to certain questions, so the sample size varies a bit. In the code, "FTwhite" means feeling thermometer rating about Whites on a 0-to-100 scale, and "Rwhite" indicates White respondents. FTblack, Rblack, etc have similar meanings (e.g., rating about Blacks, Black respondents).

  10. So this shows that as a group Blacks are overwhelmingly Tribalistic and Racist against Whites especially and Asians to a slighltly lesser degree.

  11. So if I'm reading the results correctly, Asians apparently rate Whites lower than any other ethnic group (which is surprising) -- how many of them are either 1) immigrants, or 2) offspring of recent immigrants, i.e. people who specifically chose to emigrate to a white country?

    • Hi eah. For Asian respondents as a whole in these data, the 83.4% confidence intervals overlap for Asian respondents' mean ratings about Whites, Blacks, and Hispanics, which means that we don't have enough evidence at the 95% confidence level that these three mean ratings differ from each other.

      For point estimates, Asian respondents who indicated that they were born in another country (i.e., not the U.S., D.C., or a U.S. territory) had these mean ratings about Whites, Blacks, Hispanics, and Asians: 67, 67, 67, and 78 (sample sizes about 148).

  12. It looks like the whites are most balanced group in their opinions of others (and themselves) and at the same time are lowest rated group among other races

    • Yes, sdfsf: In the aggregate, ratings about the racial groups were most balanced among White respondents. Plus, at the individual level, White respondents were more likely than other respondents to rate each racial group equally (https://twitter.com/LJZigerell/status/1377489267866402819). And these data provide enough evidence to conclude at p<0.05 that Whites were the lowest rated of the four groups among Black and Hispanic respondents, but there isn't enough evidence to conclude that at p<0.05 for Asian respondents, in these data.

  13. Huh, the results of this poll are perplexing to me. The fact that Asians and Hispanics rate blacks higher than whites is the really strange thing. Particularly considering how bad relations with blacks and hispanics are, like the amount of gang wars and hate crimes committed on each other for example, you'd think the placement of blacks and whites would be switched! Is there a way to see the amount of people polled per race and political biases of the surveyors?

  14. Pure speculation but given the social trends in the last 10 to 15 years I suspect this is more indicative of how acceptable it is for the respondent's race to express an opinion of other races. Would be interesting to see an accompanying metric for truthfulness for each rating as well as a follow-up question for each answer deemed untruthful as to why they weren't honest. I know the study is from 2021 and it's too late, just wondering. Good insight all the same.

      • Hi L.J.
        Im confused at the scale of the X-axis, because I don’t see a label for the numbers on the scale. Do the number correspond to the thermometer rating in the questionnaire 0-50, 59, 50-100 on lake 121-124?
        If that is the case dud you change the scale to 60-100 on the graph to make it more readable your graph and does that implies that most study participants rated out-groups favorably? Am I reading this correctly?

        • Hi Brian. The x-axis for each of the four plots ranges from 60 through 90. The top row of plots doesn't repeat the x-axis, but it would have been clearer if each of the four plots labeled the range of the x-axis. So, yes, each estimate in on the warmer / favorable side of the neutral rating of 50. The intent of the 60-to-90 x-axis was to zoom in on the variation, but there is a version of the plot here (https://www.ljzigerell.com/wp-content/uploads/R-gallery.html, under the section "Estimates plot: Facet shading") that has an x-axis that ranges from 0 to 100.

  15. Was this study conducted in Illinois and if so did you ask how they identified politically? I wonder if their political orientation had anything to do with how whites in particular answered the survey and had this study been conducted in a different region of the country like Alabama or Tennessee would the whites respondents had answer with the same mean scores.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.