Affective Polarization · ANES 1978–2024

Americans Didn't Fall in Love With Their Party. They Learned to Despise the Other One.

Forty-six years of feeling thermometers show warmth toward one's own party has not moved an inch. The entire rise of partisan animosity is collapsing regard for the other side — and it happened inside every kind of partisan, not because partisans got stronger.

Ask a partisan how they feel about their own party today, and the answer is almost exactly what it was when Jimmy Carter was president. In 1978, Democrats and Republicans rated their own party at 71.1 on the American National Election Studies' 0-to-100 feeling thermometer. In 2024: 70.5. Across nearly half a century of cable news, talk radio, social media, two impeachments, and a riot at the Capitol, in-party affection is statistically frozen.

The other line is the story. Over the same span, partisans' rating of the opposing party collapsed from 48.0 — a few degrees shy of neutral — to 20.6. The gap between love and loathing widened from 23.1 points to 49.9, after peaking at 52.2 in 2020. Everything we call affective polarization lives in one line. The jaws opened from the bottom.

The jaws of polarization

Weighted mean feeling thermometer (0–100) among partisans, leaners included · ANES 1978–2024 · hover or tap for values

How to read it: 50 is the thermometer's neutral point. The olive line (own party) never strays far from the low seventies; the sienna line (opposing party) falls below freezing. The shaded wedge is the in/out gap.

The bottom fell out of the distribution

Averages flatten the drama. In 1978 the typical partisan put the other party near the thermometer's midpoint — cool, not hostile. A third of partisans, 31.2%, actually rated the other side warmly, above 50. By 2024 that share had collapsed to 6.5% — roughly a fifth of what it was.

And the mass didn't drift down gently. It migrated to the most extreme answer the survey allows. The share of partisans rating the opposing party at literal zero — the coldest possible response — went from 5.4% in 1978 to 40.0% in 2020, easing to 33.2% in 2024. A response that was once a statistical curiosity is now the single most common answer.

Where partisans put the other party, 1978 → 2024

Weighted share of partisans in each out-party thermometer bin · the leftmost bar is literal zero · hover for shares

The spike at zero. Each row is one survey year; bars show the weighted share of partisans in each thermometer bin. The distribution slides left for decades, then piles up against the wall.

It's not that partisans got stronger

There is an obvious objection: maybe nobody's feelings changed — maybe America simply has more strong partisans now, and strong partisans have always disliked the other side. The composition story is half-true and almost entirely beside the point. Strong identifiers did grow as a share of partisans, from 30.9% in 1980 to 46.1% in 2024. But every kind of partisan — strong, weak, leaning — grew dramatically colder.

A shift-share decomposition makes it precise. Hold the 1980 mix of strong, weak, and leaning partisans fixed, and apply each group's 2024 feelings: that counterfactual reproduces 23.7 of the 26.0-point decline between 1980 and 2024 — 91%. Run it the other way — keep 1980 feelings, swap in the 2024 mix — and composition alone moves the average just 0.9 points, about 4%. The hate is within people, not between types of people.

Strong partisans illustrate both halves of the headline at once. Their warmth toward their own party actually slipped a little, from 84.5 in 1980 to 80.5 in 2024. Their rating of the other party fell from 42.0 to 13.0. Even the people most in love with their team didn't fall further in love. They just learned contempt.

Decomposing the collapse, 1980 → 2024

Shift-share: how much of the out-party thermometer decline is feelings changing within strength groups vs. the mix of partisans changing · hover the segments

Within, not between. Applying 2024 within-group means to the 1980 composition explains 23.7 of the 26.0-point drop; changing only the composition explains 0.9 points. Strength groups: strong / weak / leaning, ANES 7-point party ID.

Act two: 2024's selective thaw

Then 2024 did something the series hadn't done in a generation. The out-party thermometer rose. From 19.3 in 2020 to 20.6 in 2024 — a tick of +1.4, the first between-cycle increase since 2000. The path down had been relentless: 41.6 in 2000, 39.6 in 2004, 34.7 in 2008, 27.9 in 2012, 26.1 in 2016, then the cliff to 2020. The in/out gap narrowed from 52.2 to 49.9, its first narrowing since the 2012-to-2016 stretch.

A small uptick after a decades-long fall is a bend, not a reversal — and it is strikingly selective about who it touched. Republicans warmed toward Democrats by +2.3 points; Democrats warmed toward Republicans by just +0.5. Weak partisans and leaners thawed by +1.7; people who say they're not much interested in campaigns thawed by +2.3. And strong partisans? -0.1. From 13.1 to 13.0. The most committed partisans did not thaw at all.

Who thawed, 2020 → 2024

Change in weighted out-party thermometer by subgroup · dots show 2020 and 2024 levels · hover for detail

The edges softened; the core didn't. The thaw came from Republicans, weaker identifiers, and the politically disengaged. Strong partisans were flat to the decimal.

Except for the people on the ballot

The thaw also stopped cold at the candidates themselves. While feelings toward the opposing party warmed slightly, feelings toward the opposing presidential candidate set another record low: 13.4 in 2024, against 54.3 in 1968, when the candidate series begins. The share of partisans rating the opposing candidate at literal zero hit 59.7% in 2024 — up from 57.6% in 2020 and 5.3% in 1968. Democrats rated Donald Trump at 10.7; Republicans rated Kamala Harris at 16.3.

And here, too, the love side of the ledger is intact. Partisans rated their own candidate at 75.2 in 2024 — essentially the series' high-water mark of 75.9, set in 1984 for Reagan and Mondale. Whatever 2024's bend in party feeling means, person-targeted animosity has not bent with it.

Parties vs. people: the 2024 divergence

Weighted thermometer toward the opposing party (1978–2024) vs. the opposing presidential candidate (1968–2024) · toggle to the share rating the out-candidate at zero · hover for values

Institutional hatred plateaued; personal hatred didn't. After 2020 the out-party line ticks up while the out-candidate line keeps falling and the zero-share keeps climbing.
The story is not mutual radicalization of love and hate. Love stood still. Hate did all the moving.

What this shows — and what it doesn't

Feeling thermometers measure expressed affect, not behavior. Telling an interviewer the other party rates a zero is not the same as treating your Republican neighbor like one. The 2020 and 2024 ANES were mixed-mode and heavily web-based, and self-administered surveys elicit more extreme and more zero responses, so part of the 2016-to-2020 cliff may be a mode artifact — though published validation work finds the thermometer trend robust, and the 2020-to-2024 comparison is cleaner precisely because both waves shared a mode. Pure independents are excluded by construction, since they have no "own" party. The 2024 uptick is small enough that sample composition could account for some of it; "plateau" is the responsible read, not "healing." And candidate thermometers are contaminated by personality — Trump appears in three consecutive cycles, which blurs the line between disliking a nominee and disliking a recurring figure.

None of those caveats touch the central fact, which is older, larger, and remarkably clean: across forty-six years, the warm line never moved. Americans did not become more devoted to their parties. Every point of the polarization that now defines the country's politics came from one place — a deep collapse in the simple willingness to feel anything but cold toward the other side.