The Educational Realignment

The lead chart: Democratic margin by education, fifty years

The cleanest single chart for this chapter is the trajectory of Democratic party identification within each college group across fifty years.

Three observations on Figure 1.

The first is the 1972 starting point. The standard contemporary framing of the educational realignment — "college-educated voters used to lean Republican; now they lean Democratic" — describes the 1988–1992 period accurately. It does not describe 1972, when both groups leaned Democratic and the non-college group leaned slightly more so. The "college equals Republican" era was real but brief — roughly 1988 through 2000. The contemporary partisan order between college and non-college is partly a return to the 1972 ordering (Democrats more common among college) and partly a movement past it (non-college now Republican on net).

The second is that both lines fell over the period. The country's overall Democratic identification fell from 56% in 1972 to 36% in 2024 — a 20-point drop in the share of Americans calling themselves Democrats. Both college and non-college groups participated in that decline. The educational gap opened not because one group rose while the other fell, but because both fell at different speeds. The non-college line fell 27 points (57% → 30%); the college line fell 2 points (50% → 48%). The contemporary diploma divide is the arithmetic of asymmetric decline from a Democratic baseline.

The third is that the Republican line (Figure 2) tells a similar story in mirror image.

The pattern is consistent with the Democratic chart but tells the story from the Republican side: college-educated Americans were more Republican than non-college Americans in 1972 (the "country club" Republican era of the postwar period), and the order has now reversed. The non-college reading of 37% Republican in 2024 is the highest non-college Republican reading in the series.

Ideology by college: even sharper

The party-identification chart shows partisan realignment. The ideological chart shows something more pointed: the contemporary college-vs-non-college divide on liberal-conservative ideology has opened almost entirely through non-college voters becoming more conservative, not through college voters becoming more liberal.

Together, Figures 3 and 4 show the contemporary college-vs-non-college Liberal–Conservative margins.

The Table 1 finding is the chapter's most pointed: the contemporary educational divide on ideological identification is largely a story about the non-college population. The college-educated share of self-identified Liberals has held in a 5-point band for fifty years. The non-college share of self-identified Conservatives has grown sharply. The "college-educated have shifted Left" framing common in political journalism is partially right on partisan identification but is not supported on the underlying liberal-conservative measure. The shift is on the other side.

Income vs education: the cross-cutting story

The chapter's other central question is whether the educational realignment is really an educational realignment, or whether it is a proxy for income. The proposal hypothesizes that education and income push in opposite political directions: education tends to push voters toward the Democratic coalition (via cultural liberalism) while income tends to push voters toward the Republican coalition (via economic conservatism). The GSS data lets us test this with the income-by-party chart.

The income story is more complex than the standard model predicts. In 1972, both bottom-quintile (61% Democrat) and top-quintile (53% Democrat) voters were majority-Democratic, with the bottom quintile more so. By 2024, both have fallen — bottom to 38%, top to 49% — and the top quintile is now more Democratic than the bottom. The contemporary Democratic Party is, on the income dimension, the more upper-income coalition.

This is the cross-cutting finding the proposal flagged: education has been pulling Democrats toward college-educated voters; income has been pulling them toward higher-income voters. Both vectors push in the same direction at the top of the distribution (high-income college graduates are the most Democratic-identified cell in the entire matrix), but they pull in opposite directions among low-income college graduates (Democratic on education, Republican on income) and high-income non-college voters (Republican on education, Republican on income — these are now the most Republican cell).

Where the diploma divide bites

A point this book has documented repeatedly: the diploma divide is the dominant axis on some kinds of questions and a secondary axis on others. Where it bites hardest:

• Confidence in the scientific community (§2.3): the 2024 college-vs-non-college gap is 15 points, persisting throughout the series even as both groups have fallen.
• Confidence in major companies and banks (§2.5): college-educated voters are more confident in business institutions than non-college voters by 5–10 points throughout the series.
• Confidence in education (§2.3): this is the exception — non-college voters are more confident than college voters in "the people running education," and have been for fifty years.
• Confidence in the press (§2.4): the college gap is small in the early series and has actually narrowed over time as confidence has collapsed across both groups.
• Generalized trust (§6): the college gap is large (about 18–26 points within each income tier) and persistent.

Where the diploma divide is not the dominant axis:

• Sexual morality (§4.3): religion (specifically church attendance) dominates education by a large margin.
• Abortion (§4.1): church attendance again dominates education.
• Police use of force (§5): race dominates education.
• Discrimination explanations (§4.4): race of respondent matters at least as much as education.

The pattern is roughly that information-and-expertise questions show the cleanest diploma divides, while moral and lived-experience questions show that other axes — religion, race, age — matter more.

What this chapter changes

Three observations to carry into the rest of Part IV.

First, the educational realignment is real but is best described as "non-college voters became more Republican," not "college voters became more Democratic." The college-Democrat share has barely changed since 1972 (50% → 48%). The non-college-Democrat share fell 27 points (57% → 30%) over the same window. The realignment is a story about which coalition lost the working-class vote, not about which coalition gained the educated vote.

Second, education and income push in opposite directions at the bottom of the distribution. Bottom-income, college-educated voters are pulled toward Democrats by education and toward Democrats by their income position; they are roughly the most Democratic-leaning cell in the GSS. Bottom-income, non-college voters are pulled toward Democrats by income (historically) but toward Republicans by their non-college status; they are now an increasingly Republican-leaning cell. The two vectors of class composition that Democrats have historically relied on — college and lower-income — have separated. The Democratic Party can no longer count on both at once.

Third, the contemporary diploma divide is the dominant axis on information and expertise questions but is secondary to religion, race, and age on moral and lived-experience questions. Any single-variable explanation of American polarization — "it's all education" or "it's all race" or "it's all religion" — is incomplete. The contemporary partisan coalitions sort along multiple axes simultaneously, and which axis dominates depends on what kind of attitudinal question is being asked.

The next chapter, on gender and political identity, turns to the second-largest axis in the contemporary GSS — the partisan gender gap, which has emerged and widened over roughly the same period the diploma divide has, with overlapping but distinct mechanisms.