The Great Sorting · Values
The college degree is the great fault line of American politics. But the gap between graduates and everyone else was never really about money — and what little economic gap there was has now closed. What remains, and holds, is a gulf over culture.
A college degree now predicts how an American votes better than almost anything else in the pollster's toolkit. Diploma-holders have swung sharply toward the Democrats; voters without a four-year degree have moved the other way, and the realignment shows no sign of reversing. The fact is so well established that it has a nickname — the diploma divide — and a cottage industry of explanations. Most of them reach, sooner or later, for economics: the educated are the winners of the knowledge economy, the argument goes, and they vote their class interest.
Run the attitudes themselves through the data and that story falls apart. The education gap in what Americans believe — not how they vote, but where they stand on the issues — turns out to have almost nothing to do with the economic questions a class-interest account would predict, and almost everything to do with culture. Thomas Piketty gave the pattern its memorable name: the “Brahmin Left,” a left party of the highly educated that has detached from the redistributive concerns of the less credentialed.1 The American attitude record shows the Brahmin Left being born in real time.
The measure is deliberately simple. For each survey item in each year, take the weighted average answer among college graduates and subtract the weighted average among everyone else, then divide by how spread out the answers are that year — a standardized gap, in standard-deviation units, that lets a question about abortion sit on the same ruler as a question about government spending. Sort the questions into two families: economic ones about redistribution — the government's role in guaranteeing jobs, funding services, providing health insurance — and cultural ones about race, immigration, gay rights, and abortion. Then average the size of the education gap within each family, decade by decade.
The economic education gap collapsed; the cultural one held
Mean absolute education gap (college minus non-college, in standard-deviation units) on economic vs cultural attitudes, by year. Heavy dots mark the three-era averages.
Read it as: the gold line — the education gap on economics — rises, peaks, then dives to near zero. The blue line — culture — barely moves. By the most recent era the cultural gap is about three times the economic one.
Three eras, three numbers per family. On economics the mean education gap ran 0.18 standard deviations in 1972–84, edged up to 0.24 across 1986–2004 — the era of welfare reform and the Clinton boom, when degree-holders were modestly more pro-market — and then collapsed to 0.09 in 2008–24. On culture the gap held nearly flat the whole way: 0.30, then 0.30, then 0.26. The two families started a standard-deviation-tenth apart and ended a country apart. By 2008–24 the cultural education gap is 2.9 times the economic one.
Averages can hide a multitude, so break the latest era into its seven separate questions. The dumbbell below plots each item's signed education gap: how far, and in which direction, graduates sit from non-graduates. A gap near the center line means a degree barely moves the needle. A gap far from center means the degree is doing real work.
In 2008–24, the degree barely touches economics and dominates culture
Signed education gap by item (standard-deviation units), pooled 2008–2024. Positive and negative reflect the item's own scale; the label notes which side college graduates lean toward.
Read it as: the three economic dots huddle against the zero line — a degree predicts almost nothing about redistribution attitudes. The four cultural dots fan out: graduates are markedly more for legal abortion and gay adoption, and markedly more against cutting immigration.
The collapse on economics is the news, so it deserves to be stated plainly. On whether the government should guarantee jobs, graduates and non-graduates now differ by seven-hundredths of a standard deviation; on government services and spending, nine-hundredths; on a public role in health insurance, three-hundredths — statistically a rounding error. Whatever separates the diploma haves from the have-nots, it is not a disagreement about the size of the welfare state.
The college degree no longer marks a position on redistribution. It marks a position on race, sex, the border, and the family.
Culture is another country. Graduates are 0.31 standard deviations more likely to want abortion always legal, 0.25 more likely to favor gay adoption, and 0.34 more opposed to cutting immigration. The one cultural item where the gap is modest — government aid to Black Americans, at 0.15 — is also the one most entangled with redistribution, which is the exception that draws the boundary. The further a question moves from the dollar and toward identity, the more a diploma predicts the answer.
An honest accounting starts with what is not here. The education gap on economics was never large; at its mid-1990s peak it reached only about a quarter of a standard deviation. The claim is not that a chasm closed but that a modest gap shrank to nothing while a larger cultural gap stayed put — so that the diploma divide, whatever its size in any given year, is now composed almost entirely of culture. That is a claim about the shape of the divide, not a verdict on which dollars graduates earn or how they spend them.
It also is not a claim about the vote. Other work in this series traces how income and geography sort Americans into the parties; this piece stays on the narrower question of stated attitudes, where the economic-versus-cultural contrast is cleanest. Gethin, Martínez-Toledano, and Piketty document the same realignment across twenty-one Western democracies: the educated did not abandon the left so much as the axis of left-right shifted under them, from who-gets-what to questions of nation, identity, and values.2 Kitschelt and Rehm have long argued that the modern political space is two-dimensional — an economic axis and a cultural one that need not point the same way — and education loads heavily on the second.3 The American attitude data put a number on it: by the 2010s, the second dimension is where the diploma lives, and the first is where it has all but vanished.
The political consequence is the part worth sitting with. A party coalition built on the cultural confidence of the credentialed is not, by the evidence here, a coalition organized around redistributing income — its college base simply does not divide from the rest of the country on that question anymore. The diploma divide is real, durable, and growing. It is just not, and perhaps never was, a fight about money.
Data. ANES Cumulative File, 1972–2024 (22 surveyed years with at least one fielded item), weighted by the cumulative full-sample weight. Education is the cleaned college_grad indicator (four-year degree or higher vs not). Economic family: government job guarantee, services/spending, and health-insurance 7-point scales. Cultural family: government aid to Black Americans (7-point), decrease-immigration (indicator), favor-gay-adoption (indicator), and abortion-legality (4-point).
Measure. For each item in each year, the standardized education gap is (weighted mean among college graduates − weighted mean among non-graduates) divided by the weighted standard deviation of the item that year (variance from w2.wvar, frequency-weight convention). A year-item gap is computed only where at least 150 respondents are valid and each education group has at least 30. The family figure averages the absolute gap across that family's items; the per-item dumbbell keeps the sign and averages over the years in the 2008–24 era.
Honest limits. Standardizing by each year's spread makes items comparable but means the gap is measured in within-year units, not raw scale points; a flat standardized line can coexist with a changing raw distribution. The economic gap was modest even at its peak (~0.24 SD), so the story is its collapse toward zero, not the closing of a chasm. The immigration and gay-adoption items begin in 1986/1988, thinning the cultural family's earliest years; the era averages weight all available item-years equally. Aid to Black Americans straddles the economic and cultural families conceptually and is classed as cultural here, following its loading on racial rather than redistributive attitudes; moving it to the economic side would only widen the gap between families.