Americans didn’t fall in love with their party. They learned to despise the other one.

There are two ways a country can polarize. Its citizens can fall harder for their own side, or they can turn on the other one. For nearly half a century, the American National Election Studies has asked partisans to do something clarifying: rate both major parties on a “feeling thermometer” that runs from zero (cold, hostile) to 100 (warm, favorable). The answers settle the question.

In 1978, partisans — counting independents who lean toward a party — rated their own party at 71.1 degrees. In 2024, after twelve presidential elections, cable news, talk radio, the internet, and everything else, they rated it at 70.5. The line is flat. Whatever happened to American politics over those decades, it did not make people love their party more.

The other line did all the work. In 1978, partisans rated the opposing party at 48.0 — a hair under the scale’s neutral midpoint, roughly the temperature of a neighbor you disagree with but still wave to. By 2024 it had fallen to 20.6. The gap between affection and animosity more than doubled, from 23.1 points to 49.9, after peaking at 52.2 in 2020. Plotted together, the two series open like a set of jaws: a stable upper lip of in-party warmth and a lower jaw of out-party regard dropping away beneath it.

The crowd migrated to zero

Averages flatten what is, up close, a much stranger picture. The thermometer is a 0-to-100 scale, and people can park an answer anywhere on it. In 1978 they mostly parked near the middle: the most common rating of the opposing party fell in the band just under 50, and only 5.4% of partisans handed the other side a literal zero — the coldest answer the instrument allows.

By 2020, zero was the single most common answer. Four partisans in ten — 40.0% — went all the way to the floor. The share eased to 33.2% in 2024, which still leaves one partisan in three reaching for the bottom of the scale when asked about the other party. Warmth, meanwhile, has nearly vanished: 31.2% of partisans rated the opposing party above 50 in 1978; in 2024, 6.5% did. A five-fold collapse in the simple willingness to say something kind about the other side.

It is not that partisans got stronger

An obvious objection: maybe the people changed, not the feeling. Americans today are likelier to call themselves strong partisans — 30.9% of all partisans in 1980, 46.1% in 2024 — and strong partisans have always been the coldest toward the other side. Perhaps the collapse is just arithmetic, the average dragged down by a hotter mix of identifiers.

It isn’t, and a shift-share decomposition shows why. Freeze the 1980 mix of strong, weak, and leaning partisans, and let each group feel about the other party the way its members did in 2024: that counterfactual alone reproduces 23.7 of the 26.0-point decline — 91% of it. Run the experiment the other way, giving 2024’s mix of identifiers 1980’s feelings, and composition moves the needle by 0.9 points, about 4%. The change happened within every kind of partisan, not between them.

Strong identifiers rated the opposing party at 42.0 in 1980 and 13.0 in 2024. Weak partisans and leaners traced the same slope from a slightly higher perch. And even among the most committed, love of one’s own side did not grow — it slipped, from 84.5 to 80.5. Nobody fell harder for their team. Everyone learned contempt for the other one.

2024: the first thaw in a quarter century

Then the series did something it had not done since the Clinton administration. The out-party thermometer fell every presidential cycle from 41.6 in 2000 down to 19.3 in 2020 — and in 2024 it ticked up, to 20.6. A rise of +1.4 points: tiny, but the first between-cycle increase since 2000. The gap between in-party and out-party feeling narrowed from 52.2 to 49.9, its first narrowing since 2012–2016.

The thaw was selective, and the selection is telling. Republicans warmed toward Democrats by +2.3 points; Democrats warmed toward Republicans by just +0.5. Weak partisans and leaners gained +1.7; people who say they are not much interested in campaigns gained +2.3. And strong partisans? They went from 13.1 to 13.0 — a change of -0.1, which is to say none at all. The ice melted at the edges of the coalitions and stayed solid at the core.

…except for the people on the ballot

If hatred of the opposing party has plateaued, hatred of the opposing person has not. Since 1968 the ANES has also asked partisans to rate the presidential candidates. In 1968, the opposing nominee scored 54.3 — warm side of neutral. In 2024 he or she scored 13.4, the coldest reading in the history of the series, lower even than the rating of the opposing party in the same year.

The zeros tell it more brutally. The share of partisans rating the opposing presidential candidate at literal zero went from 5.3% in 1968 to 57.6% in 2020 to 59.7% in 2024 — nearly six partisans in ten assigning the other side’s nominee the lowest score a survey will record. Democrats put Trump at 10.7 degrees. Republicans put Harris at 16.3.

And here the asymmetry that defines the whole era reappears at the level of persons: love for one’s own candidate never wavered. Partisans rated their own nominee at 75.2 in 2024 — a whisker from the warmest reading the series has ever produced, 75.9 in 1984. Same pattern, fifty-six years running: the warm line holds, the cold line falls. In 2024 the two cold lines finally split from each other — the party line bent upward while the candidate line set another record low.

What this does and doesn’t show

Three honest cautions. First, mode: the 2020 and 2024 ANES waves were mixed-mode and heavily online, and self-administered surveys elicit more extreme answers — more zeros — than interviews do. Some portion of the cliff between 2016 and 2020 is probably mode, not mood. That same fact cuts the other way for the thaw: 2020 and 2024 were fielded similarly, so the +1.4 uptick is at least an apples-to-apples comparison. But it is small, and “plateau” is a safer reading than “reversal.”

Second, thermometers measure expressed feeling, not behavior. A zero is a statement, not a fist. Third, the universe here is partisans including leaners; pure independents are excluded by construction, and Trump appears on three consecutive ballots, which blurs how much of the candidate series is about candidates in general versus one candidate in particular.

What the record shows without qualification is the shape of the thing. Affective polarization in America was never a love story. The warmth people feel for their own party is exactly where it stood when Jimmy Carter was president. Everything that changed — all of it, in every kind of partisan — is the disappearance of ordinary, lukewarm, wave-across-the-fence regard for the other side. The jaws opened from the bottom.