Racial Realignment, or Diploma Divide?

Some reflections on the Klein-Shor interview.

Photo by RUT MIIT on Unsplash

Ezra Klein interviewed David Shor this week for his podcast The Ezra Klein Show, and one particular aspect of their conversation caught my attention first. Klein posed a question to Shor about the Democratic Party’s losses in 2024 among Black, Hispanic, and Asian voters, and asked Shor’s thoughts on what’s causing that trend. Here’s Shor’s reply:

“There’s this story in American politics: If you want to understand 20th-century American politics, then the big story is that there was this giant Southern realignment in 1964…driven by the Civil Rights Act. And that carried forward. It took a really long time for that to work its way down the ballot. So it was tempting for American political scientists and many more detail-oriented Americans, political consultants and pundits to see everything…through this transformation. The most important political trend of the last 30 to 40 years, both here and in every other country in the world — at least in Western countries with elections — has been this story of education polarization. Basically, we’ve seen highly educated people move to the left, while working-class people have moved to the right.”

A quick refresher on racial realignment, cribbed from one of my own prior pieces:

Until the 1960s, the two political parties were virtually indistinguishable on matters of race. That changed mid-decade when the Democratic Johnson Administration passed the Civil Rights Act and the Voting Rights Act in 1964 and 1965, respectively. Suddenly, voters could distinguish the two parties’ stances on racism and segregation, and this set off a racial realignment of the party coalitions. White segregationists in the South began to migrate away from the Democratic Party, while Black Americans and racially liberal white voters migrated toward it.

Post-Civil Rights era, the two parties found themselves embroiled in what scholars call the “electoral temptation of race.” In order to win national elections, each needed need to appeal to uncommitted white voters — the ones who were not racial liberals but were not segregationists, either — without incurring political costs. For Republicans, the challenge was to court these voters with messages that subtly appealed to their racial biases without being so overtly racist as to violate new social norms about equality. For Democrats, the challenge was to balance a core component of their new voting base—Black voters—against the need for a broader coalition that included white voters who would be alienated by outright racial liberalism. Their mandate was to do just enough for Black voters to keep them in the coalition, but not enough to threaten the white voters needed to amass a majority.

Essentially, Shor’s argument is that racial realignment is not the all-important explanation for America’s political landscape that we make it out to be, and that education polarization is the more meaningful trend. One important thing to note, before I go further, is that the racial demographics of the two party coalitions are not necessarily the best reflection of the continued impact of racial realignment. I’d argue racial attitudes are far more influential. Racial depolarization in the 2024 election outcome does not mean that racial attitudes are becoming less influential in our politics.

Fundamentally, I disagree with Shor’s argument, for a fairly simple reason: education polarization is largely a proxy for those racial attitudes and other symbolic values.

Racial Attitudes Are Doing The Heavy Lifting

I’m not the first, or the only, public opinion researcher to make this argument. Ezra Klein interviewed political scientists John Sides and Lynn Vavreck in 2022 and posed similar questions to them about the significance of the diploma divide. Sides restated the argument he, Vavreck, and Michael Tesler made in their 2016 book Identity Crisis: among white voters, education level is “proxying values and beliefs related to minority racial and ethnic groups.” To his point, in virtually every survey where I ask questions about racial attitudes, I observe a divide in the answers by education level among white voters. In a Change Research poll of likely voters conducted in October of last year, I asked voters to indicate how much they agree or disagree with this statement: “I am tired of hearing so much about racial issues in the United States.” 63% of white voters without college degrees agreed with that statement; 59% of white voters with college degrees disagreed.

Even when we widen the aperture and focus on all voters, not simply white voters, Sides notes that yes, educational divides may absolutely be linked to policy preferences, but he rightly notes that “policy issues sometimes are not policy issues. Policy issues are statements about groups and who you like or dislike.” The bottom line: political preferences differ meaningfully based on voters’ cultural/racial attitudes, and the education divide reflects those differences, it doesn’t refute them.

Here are three off-the-cuff analyses I ran to illustrate that argument, looking at relationships between education, cultural/racial attitudes, and vote choice.

First, some simple correlation analyses using that same October 2024 polling data conducted by Change Research.

• Education Level & Vote Choice: Among white voters, the correlation between education level and voting for Trump in 2024, as of early October of last year, was -0.24, which is what we’d consider a medium strength correlation. Higher educational attainment correlates with less likelihood of voting for Trump.

• Education Level & Racial Resentment: The correlation between education level and racial resentment is -0.23, a comparably strong correlation. Higher educational attainment correlates with lower racial resentment.

• Racial Resentment & Vote Choice: The correlation between racial resentment and voting for Trump, among the same voters, is 0.77. Higher racial resentment correlates extremely strongly with higher likelihood of voting for Trump, and the relationship between these two variables is much stronger than the relationship between education and vote choice.

This graph below shows the predicted probability of voting for Trump as a function of education level, controlling for age, gender, region, partisanship, and racial resentment. As you can see, the line is almost flat. There’s no meaningful change in likelihood of choosing Trump over Harris as a function of education, when we control for racial attitudes and partisanship.

The same is plainly not true for racial attitudes. The likelihood of voting for Trump, controlling for demographics (including education level) and partisanship, differs drastically as a function of racial resentment.

Lastly, let’s look at all voters, not just white voters, and consider immigration-related attitudes. I used Change Research poll data from February of 2024 to create this graph, which shows the predicted probability of voting for Trump as a function of education level, controlling for age, gender, region, race, partisanship, and anti-immigrant sentiment, which I measured by asking people to rate their agreement/disagreement with these three statements:

• We will lose valuable parts of American culture and heritage if we let in too many immigrants from foreign countries.

• Immigration is a good thing for the country these days.

• Immigrants take job opportunities from people who were born here.

Voters are modestly less likely to choose Trump over Harris as a function of higher educational attainment, controlling for anti-immigrant attitudes and partisanship.

When I replicate the same analysis, but hold educational attainment constant and look at the relationship between anti-immigrant sentiment and vote choice, the effect is far more profound. Simply put, racial/cultural attitudes exert much more statistical influence over vote choice than education level does, when we account for both in the analysis.

What does this mean?

Political attitudes are heavily shaped by voters’ cultural/racial attitudes, and education exerts very little influence over vote choice once we account for those attitudes in the model. I’d argue this is evidence of racial realignment’s lasting legacy, not proof of its diminishing impact. That being said, the diploma divide is a meaningful trend in public opinion these days, in and of itself. We absolutely should conduct more research on the topic in order to understand these trends better, because while we know that education level relates to sociopolitical attitudes, we don’t actually know what it is about having a college degree that influences people’s beliefs. Klein himself mused about the opacity of education as an explanatory variable for vote choice in that 2022 interview with Sides and Vavreck:

“I always think it’s very unclear what education is standing in for here, what it is getting at as a variable you’re tracking, because I don’t think what’s happening is simply stuff you learned in school. I don’t think that’s what’s driving the divide.”

I share Klein’s skepticism. That argument sounds like the conservative theory of the case, that radical leftist professors are indoctrinating impressionable young minds the moment they step on campus. We know that’s false, but we don’t have a clear answer as to what the mechanism actually is. What is it about a college degree that produces, or correlates with, more liberal political beliefs throughout the lifespan? Stay tuned; I hope to explore it further with some upcoming data.