White liberals view other races more warmly than they do Whites. Why?

The roles of age, ideology, interest, and linked fate

Peter Licari, PhD
13 min readOct 9, 2021


Crowd of people, predominantly White, participating in a Black Lives Matter Protest
Photo by Life Matters from Pexels

White liberals have spent the past few years marching steadily to the left.

This sentence is a rarity in American politics today: it garners pretty broad agreement across the political spectrum. (Although, predictably, with drastically different underlying sentiments depending on what part of said spectrum you ask). And this sentence is especially true when it comes to issues of race. Over the past few years, a variety of polls and research projects have shown that White liberals have shifted their positions leftwards in response to cues from in-party elites, in reaction to (and against) then-President Trump, and in response to (and probably to some extent propelling) the increased importance of race in the broader American zeitgeist.

In fact, on an appreciable number of issues, the views of White liberals are more to the left than their non-White co-partisans.

In 2018, PhD Candidate Zach Goldberg penned an essay on this last point for Tablet sardonically titled ‘America’s White Saviors’. Using data from the 2018 American National Elections Study (ANES) Pilot Survey, he finds that:

Remarkably, white liberals were the only subgroup exhibiting a pro-outgroup bias — meaning white liberals were more favorable toward nonwhites and are the only group to show this preference for groups other than their own. Indeed, on average, white liberals rated ethnic and racial minority groups 13 points (or half a standard deviation) warmer than whites.

As he noted “the emergence and growth of a pro-outgroup bias is actually…[an] unprecedented phenomenon.”

There are a million ways to frame this finding, many of which aren’t necessarily exclusive: White liberals are coming more to terms with their privilege; they’re becoming more aware of racist systems and their roles within them; they’re witnessing increased signaling among White liberal political and cultural elites on the topic; and they’re feeling increasing amounts of White guilt (which, for what it’s worth, seems one of the sources Goldberg is predominantly considering at the moment). Regardless of how you decide to approach it though, it’s certainly a unique phenomenon among the intersections of America’s political and racial cleavages for a group to prefer others over themselves.

But Goldberg made his observations in 2018. And a lot, an impossible amount really, has happened since then. I was curious to see if this pattern has continued. After all, one of the things separating the social sciences from the “hard” STEM fields is the near constant potential for change. Sometimes unprecedented things return back to their precedents.

Not here, though. What I’ve found, using the 2020 ANES, is that the trend continues. Non-Hispanic White liberals (what I’ll refer to simply as ‘White liberals’ hereon out for convenience’s sake) continue to rank other ethnoracial groups (e.g., Blacks, Asians, and Hispanics) higher than they rank Whites. On average, they rank Blacks and Hispanics around 6 points higher than they do Whites and they rank Asians 4 points more. But statistical analyses (my way of saying “I ran a regression”, but fancier) shows that these singular numbers mask a lot of heterogeneity. The average size of the gap is affected by things such as gender, age, ideology, and perceptions of “linked fate” with other Whites.

Before discussing these associations, let’s get acquainted with the data and how I analyzed it.

The Data: What we’re measuring and how

The evidence that I’m using is the same that Goldberg used back in 2018: a survey instrument called a “feeling thermometer.” Surveyors asked respondents to identify how they felt about a host of different politicians, entities, social movements, and groups on a scale from 0–100, with lower values being colder feelings towards the target, higher values being warmer feelings, and 50 indicating feeling neither “particularly warm or cold.”

ANES respondents are asked dozens of such questions but four are especially pertinent here: “Asians”, “Hispanics”, “blacks” and “whites.” I looked only at the responses to these questions given by non-Hispanic White liberals, and only counted responses that came with an answer rather than saying something like “don’t know” or “haven’t thought much about it.” I then subtracted respondents’ rankings of Whites from their rankings of all the other racial groups. If White liberals really rate members of the racial out-group higher, we’d see a whole lot of positive values. (A big positive number minus a small positive number is a positive, albeit smaller, number.) And as we can see below, that’s exactly the case; there are way more positive values than we see negative values. Meaning more White liberals are ranking Whites lower than they’re ranking Blacks, Asians, and Hispanics.

Title: White liberals gave more positive rankings to non-White groups than to Whites. Displaying a bar chart visualizing the net difference in White liberal sentiment towards Asians, Blacks, and Hispanics versus their sentiments towards Whites. In all 3 cases, a noticeable plurality of respondents say they feel the same towards the groups being compared but, on average, more of the distribution falls towards the side indicating greater preferences for the non-White groups.

Well, actually, the most conspicuous thing in these charts are those giant spikes around 0. Indeed, over 18 percent of liberal Whites ranked Asians and Whites identically, 35 percent gave the same rankings to Hispanics and Whites — about roughly the same as ranked Whites and Blacks identically (36 percent).

(Actually, this brings up an excellent opportunity to discuss one of the very real weaknesses of feeling thermometers as a measure. They’re definitely useful, but most modern Americans have had it hammered home to them that people of all races are fundamentally equal. So when they’re asked these kinds of questions, that’s how a fair chunk of them are going to respond. Feeling thermometer gaps are certainly correlated with things like racial in/out-group favoritism, but it is not a perfect proxy.)

I investigated the factors associated with greater out-group favoritism using a statistical framework that allowed me to account for the fact that individuals may be more-or-less predisposed towards giving higher/lower thermometer ratings across all three ethnoracial groups, and thus be predisposed towards having larger/smaller gaps. (For those familiar with statistics: I used a multilevel regression model with individual-level random slopes. Those curious about the technical details behind the data transformation and modeling can check out the accompanying blog post I’ve written on my personal website and find the code on my GitHub.)

My analysis showed at least six factors that are significantly associated with the size of this gap. Two of these relationships (Race and Gender) were statistically significant but weren’t all that substantial. Liberal Whites tended have a larger pro-Black gap and pro-Hispanic gap than they do a pro-Asian gap (though there isn’t a significant difference between the first two), but these only amounted to about a 1 point increase for both. White liberal women tended to have a larger gap than White liberal men, but only by about 2 points.

The other four, though, were more substantial and are worth looking at in more depth. These were age, political ideology, political interest, and linked Fate.

What enlarged the gap?


Title: Pro-outgroup bias is strongest among younger White liberals. Displaying three line charts stacked into one column, each having age as the x axis and the size of the pro-outgroup bias on the y axis. The top relates to the difference in sentiments between Asians and Whites, the middle between Black and Whites, and the bottom between Hispanic and Whites. All three line charts show a slope with lower pro-outgroup biases among older White liberals than among younger White liberals.

As the figure above shows, younger White liberals tended to have substantially larger pro-outgroup gaps than older White liberals. Indeed the regression model suggests that, all else being equal, being 10 years younger tended to inflate the pro-outgroup bias by an additional 3 points. But this shouldn’t be taken to mean that a pro-outgroup bias is exclusively a phenomenon among the youth. The model estimates that the average gap isn’t within 5 points of zero until respondents are about 58 — and the gap is estimated to be significantly distinct from zero until age 68. So while the average gap is larger among younger White liberals, it is present until we get to some of the oldest respondents in the data.


Although all of the respondents generally identified as “liberal” in the ANES, there are obviously ideological gradations within that single label — something that I couldn’t help but articulate via the universal medium of memes.

The ANES distinguishes between those who are “Extremely liberal”, “Liberal”, and “Slightly liberal.” Back in 2018, Goldberg noted that there were differences between the different levels of ideological buy-in (though the survey he used only had “Very liberal” and “Liberal” rather than the three-point scale used here). Those who were the most liberal (e.g., “Very liberal”) had, on average, pro-outgroup gaps of around 20 points — twice as much as those who simply identified themselves as “Liberal.” By and large, this relationship persisted in 2020 — though to a lesser degree.

Graph Title: More liberal respondents had larger expected pro-outgroup preferences over Whites. On the x axis is the average estimated pro-outgroup gap; on the y axis is “slightly liberal”, “liberal”, and “extremely liberal.” Solid points are accompanied by fainter ones, the latter showing bootstrapped confidence intervals. The pro-outgroup gap is higher for extremely liberal than it is for liberal — which, in turn, is higher than slightly liberal.

The chart above shows the predicted effects that the three levels of liberalism have on the size of the pro-outgroup gap. (The solid dots show the estimated effect, the faded dots show the 95% confidence intervals generated via bootstrapping). While the maximum effects is estimated to be smaller here (“Extremely liberal” respondents had a gap of about 13 points rather than 20), the estimate for “Liberal” is pretty consistent with his previous findings. But that gap is still about two-and-a-half times as large as it is for someone who identified as only “Slightly liberal.”


While many of those self-selecting into reading an article about political science probably take for granted that most people are interested in politics, the fact is that most people actually aren’t all that tuned-in to every minutia of the day-to-day drama; those vicissitudes and controversies careening on Twitter and decaying faster than a man-made atom. Those who are more politically interested tend to be more cued-in to their in-parties’, well, cues — as well as be more strongly affected by ideologically-consistent messaging. So we may expect the forces and effects driving White liberals to exhibit stronger pro-outgroup gaps to be concentrated in those with greater interest.

Chart title: Among White liberals, greater political interest was associated with larger preferences for other ethnoracial groups over Whites. On the x axis is the average estimated pro-outgroup gap, on the y is “not at all interested”, “not very interested”, “somewhat interested”, and “very interested.” Solid dots surrounded by faded ones, the latter being the bootstrap confidence interval, shows higher gap values for more interested respondents.

That’s exactly what we see in the data. The chart above shows the estimated effects of the various levels of interest on the size of the gap. All things being equal, a White liberal who is “Not at all interested” in politics will have a gap of about 3 points. Someone who is “Very interested” in politics, however, is expected to have a nearly 10 point gap; a change of roughly 7 points.

Linked Fate

In political science, the concept of “linked fate” has been thoroughly researched as a mechanism for increased political engagement. Broadly, linked fate can be thought of as the predisposition of respondents to feel that their personal experiences and trajectories are correlated (or “linked”) to that of a broader, salient group. Those with stronger feelings of linked fate tend to be more likely to be politically engaged and active. Although research into the topic has historically focused on non-White groups, recent work has shown that Whites, broadly, also experience perceptions of linked fate — and these feelings can be associated with increased political participation and stronger preferences for Whites over others in positions of political power. The 2020 ANES measured linked fate among Whites by asking respondents to what extent they felt that what happens to white people will affect their lives.

So if stronger linked fate among Whites would be expected to drive more positive feelings towards Whites, we should expect that weaker feelings of linked fate might lead to greater outgroup preferences.

Chart Title: Among White liberals, greater linked-fate with Whites generally was associated with smaller amounts of out-group preference over Whites. A chart with the x axis being average estimated pro-outgroup gap, the y axis labeled “How much of what happens to White people will affect your life?” with four levels: Not at all, not very much, some, and a lot. Lower levels of perceived linkage is associated with larger pro-outgroup gaps.

Though this is the weakest of the 4 relationships I’ve looked at here, it is statistically significant. Those who believe that what happens to White people will have “a lot” of impact on their lives personally are expected to have an average pro-outgroup gap of about 5 points. This gap increases as the level of affect decreases; by the time that respondents feel that what happens to White people will matter “not at all”, the pro-outgroup gap is estimated to be about 11 points. Higher degrees of linked-fate, then, tends to be associated with smaller preferences for other ethnoracial groups over Whites.

What didn’t enlarge the gap?

There’s one factor that seems conspicuously absent here — especially considering all of the constant (overblown) worry of “liberal indoctrination”: Education. It’s not that I didn’t include it in the model; I expected that the differences would be pronounced among those who were younger and educated, so I definitely had it in there. It’s just that the effect of having a college education wasn’t statistically significant here. And in fact, if there were any meaningful differences, it appears to be most concentrated among older White liberals rather than their younger counterparts. Even here though, the effects are modest at best.

Final thoughts

On the whole, it probably shouldn’t be all that surprising that this phenomenon persisted into 2020. Arguably, it had a better chance of happening here than it did in 2018 given the prevalence of the Black Lives Matter protests during that year in response to events like the death of George Floyd. (After all, while Whites overall were initially more supportive of BLM earlier in the year, that support eroded quite a bit by the end of 2020 — though it eroded least among, you guessed it, liberal Whites.)

These results harmonize with work that shows that much of the leftward push among Liberals can be attributed to age gaps between the youngest and oldest members, though the party as a whole has been trending leftward on racial issues regardless of age. As I mentioned above, the finding on ideological intensity coheres with Goldberg’s analysis back from 2018 which found that the gap was greater among more avid liberals and non-existent among moderate liberals — although that latter bit appears to have changed here. The finding with linked fate harmonizes with more recent work which argues that some individuals update their identities — including their ethnoracial identity — to be more in line with their political identity. That is to say, for some people, our political identities can be central enough to affect how we think about and outwardly present our other ones. In that context it makes sense that we see that those Whites who feel the least linked to the White racial identity exhibit larger pro-outgroup tendencies. Indeed, additional analyses reveal that White liberals with lower feelings of White linked fate ranked Whites significantly lower than those with higher feelings of White linked fate — but White linked fate had no significant association on the rankings of Blacks, Asians, or Hispanics.

But while it harmonizes well, it’s difficult to disentangle all the strands in the causal web. This, after all, is cross-sectional data. We shouldn’t say that, for example, “being more liberal makes people like non-White ethnoracial groups more than they like Whites.” After all, with race’s centrality in American politics right now, it’s likely that there’s a certain amount of sorting going on; White liberals who feel more warmly towards Blacks, Asians, and Hispanics as groups (or feel more antipathy towards Whites) are probably going to be more likely to identify as “Extremely” liberal simply because they recognize this gap as being an amplification of the position held by the group overall. That’s not to say that there isn’t a causal effect from ideology-to-opinion, but that the influence between the two is most likely bi-directional (and, possibly reciprocal: with ideology informing positions which, in turn, informs sorting).

Indeed one could make a similar argument for political interest and linked fate; I doubt that there isn’t a causal arrow stemming from these two to a pro-outgroup bias, but I would also be surprised if it turned out to only cut one way. Other relationships have stronger arguments for causality — such as age and gender — but we’re missing the underlying mechanism (especially since the socializing effect of college seems to carry very little weight here). But causality isn’t always everything. Sometimes it’s interesting and important to understand the state of the world as it is, and appreciate the fact that concepts do relate, without necessarily needing to have the causal flow fully fleshed out.

It’s important to emphasize that this isn’t a universal finding among White liberals. After all, much of the sample ranked all of the groups identically or otherwise had stronger feelings towards Whites. And, as I’ve shown, there are factors that cause this relationship to vary. However, for the majority of White liberals, the truism remains true: here is yet more evidence that, as a group, they are pushing leftwards on racial issues — so much so that they tend to rank other races more warmly than they do Whites overall.

This is version 1.0.0 of this analysis.

I believe all work benefits from readers’ constructive feedback as well as the writer’s own revisiting and reflection. But not all work is fit for academic publication. To that end, for the sake of transparency, I have decided to visibly index what point my non-academic projects are at once they go live in a format where others will (hopefully) come into contact with it. Minor revisions (such as grammar or minor image formatting issues) lead to an increase in the third digit. Major revisions inspired by my own revisiting and minor revisions inspired by the suggestion of readers leads to an increase in the second digit. Major revisions driven by the suggestion of readers or by future reflection and revisiting the project leads to the an increase in the first digit. After 6 months of no updates, a version should be considered “final.” The current version was published 10/9/21. All versions of this post are maintained on Github.

Peter Licari is a social data scientist specializing in American political behavior. He received his PhD in American Politics and Political Methodology from the University of Florida and is currently a Director of Commercial Data Science for Morning Consult. The views expressed here are his own. He can also be found on YouTube and on Twitter(@PRLPoliSci). What little spare time remains is dedicated to long-distance running, binging games and media with his wonderful wife, Stephanie, playing with his daughter, Rosalina, walking his dog, Dude, and holding oddly productive one-sided conversations with his cat, Asia.



Peter Licari, PhD

I’m a data scientist and social scientist specializing in political behavior. I’m also a runner, writer, gamer, YouTuber, and dinosaur enthusiast.