The Great Sorting · Wave 2
For most of the twentieth century, abortion cut across the parties as often as it divided them. Then it became the line itself. Four decades of survey data trace a single issue traveling from the margins of partisanship to its core — and the voters who once straddled it nearly vanishing.
In 1980 a Democrat and a Republican were about equally likely to say abortion should always be a matter of personal choice. Among Democrats the share was 37 percent; among Republicans, 34. The three-point gap between them was the kind of difference that vanishes inside a margin of error. Abortion belonged to a different map of American politics, one drawn by religion and region: pro-life Catholic Democrats in the industrial Northeast, pro-choice country-club Republicans in the suburbs. It was, in the language of political science, a cross-cutting issue, and cross-cutting issues are supposed to dampen conflict by scrambling the lines along which it runs.
By 2024 the same question split the parties by more than fifty points. Among Democrats, 74 percent said abortion should always be allowed; among Republicans, 21. What had been a rounding error became one of the widest partisan chasms in the entire American National Election Studies. The issue did not merely polarize. It changed jobs — from a force that blurred the party divide into the clearest single marker of which side a voter is on.
A three-point gap became a fifty-point gap
Weighted share of each party (identifiers including leaners) saying abortion should “always be allowed,” 1980–2024. Shaded bands mark presidential terms.
Takeaway: the two lines start almost touching in the 1980s and fan apart into a gulf, with the Democratic share climbing and the Republican share drifting down.
The widening was not symmetric, and that asymmetry is the mechanism. Republicans barely moved: their “always legal” share drifted from 34 percent in 1980 to 21 in 2024. The motion was on the Democratic side, where support for unrestricted access climbed from the high thirties into the low seventies. Carmines and Stimson called this kind of trajectory issue evolution — the slow capture of a new issue by the party system as activists, then candidates, then ordinary voters fall into line.1 They built the theory on race. Greg Adams showed two decades later that abortion was following the same script, with party elites in Congress diverging on roll-call votes well before the mass public did.2 What the long survey series now records is that script played to its end: the elite cue, repeated election after election, finally rewrote the electorate.
The clearest casualty of the sort is the voter who used to sit athwart it. The pro-life Democrat — the union Catholic, the Southern churchgoer who pulled the Democratic lever for reasons that had nothing to do with abortion — was once a real and countable bloc. Across the 1980s, roughly 13 percent of Democrats said abortion should never be permitted. By 2024 that share had fallen to 1.4 percent. The pro-choice Republican has thinned in parallel, from better than a third of the party to about a fifth. The space in the middle, where the two coalitions used to overlap, has been squeezed nearly flat.
That space had been wide for a reason, and the reason was that the parties did not yet own the issue. In the years just after Roe v. Wade, both coalitions contained prominent voices on each side: Republican governors who defended legal abortion, Democratic stalwarts who opposed it on religious grounds. A voter could be cross-pressured because the cue from her party was itself mixed. Adams's roll-call evidence pinpoints when that ambiguity ended — party elites in Congress pulled apart on abortion across the 1980s, handing voters an unmistakable signal about which position belonged to which side. The mass figures here trace the lag: the elite divergence Adams measured arrives first, and the electorate's catches up over the two decades that follow, until by the 2010s the cue and the public are moving together.
The cross-pressured were squeezed toward zero
Weighted share of Democrats who say abortion should never be permitted (“pro-life Democrats,” filled) and of Republicans who say it should always be allowed (“pro-choice Republicans,” line), 1980–2024.
Takeaway: the filled band of pro-life Democrats melts from one-in-eight toward one-in-seventy, while pro-choice Republicans thin more gently — the overlap that once linked the parties is closing.
A cross-cutting issue is supposed to soften the party divide. This one hardened into the divide itself.
As the positions sorted, abortion also became a better and better predictor of how people vote. Correlate a respondent's abortion view with whether they cast a Democratic presidential ballot, and in 1980 the relationship is essentially nothing — a weighted coefficient of −0.05, indistinguishable from zero and, if anything, faintly pointing the wrong way. By 2000 it had risen to +0.24; by 2016, +0.42; by 2024, +0.59. An attitude that carried almost no information about a person's vote at the start of the series had become one of the most reliable single tells in it.
Abortion went from telling you nothing about a vote to telling you almost everything
Weighted correlation between abortion position and a Democratic presidential vote, presidential-election years, 1980–2024.
Takeaway: the correlation climbs steadily off the zero line, from no relationship in 1980 to a strong one by 2024.
The timing of the steepest climb is its own argument. The lines pull apart fastest after 2012, through the years when abortion access moved from a settled background condition to an active front — state-level restrictions, the reshaping of the Supreme Court, and in June 2022 the Dobbs decision returning the question to the states. A large panel study spanning that decision found the causal arrow running in both directions at once: when abortion suddenly became policy-relevant, Americans whose abortion views were out of step with their party tended to bring their party into line with their abortion views, rather than the reverse.3 Cassese, Ondercin, and Randall, surveying the post-Dobbs electorate, reach a parallel conclusion: partisans are now more divided on abortion than at any point on record.4 The first reading the ANES took after Dobbs, in 2024, is the widest gap in the forty-four-year series.
Sorting on position is only half of it; the other half is which side cares enough to vote on it. For most of the period that energy belonged to the pro-life movement, and the single-issue abortion voter was a Republican fixture. Dobbs appears to have inverted that, too, by converting an abstract right into a concrete loss for the side that had taken it for granted — the kind of shock that turns a held opinion into a voting priority. The figures here measure where Americans stand, not how hard they will fight for it, but the two have moved together before, and a position this neatly aligned with party rarely stays quiet for long.
There is a tidy reading of all this and a darker one. The tidy reading is representation working as designed: people care about abortion, the parties offer a clear choice, and voters sort to the side that matches their conviction. The darker reading is that the sorting has run past the point of choice. When a single issue maps this cleanly onto party — when knowing someone's view on abortion lets you guess their vote nearly six times in ten — it stops behaving like an issue voters weigh and starts behaving like a badge they wear. The pro-life Democrat and the pro-choice Republican were inconvenient people. They were also the proof that Americans could disagree about abortion and still share a coalition. Their near-disappearance is what a thoroughly sorted politics looks like from the inside.
Data. American National Election Studies Cumulative File, 1980–2024 (17 surveyed years in which the four-category abortion item was fielded), weighted by the ANES design weight. The item asks which of four positions comes closest to the respondent's view, coded 1 = abortion never permitted, 2 = permitted only for rape, incest, or to save the woman's life, 3 = permitted for other reasons once need is established, 4 = always allowed as a matter of personal choice. “Always legal” is category 4; the “pro-life” position is category 1.
Party. Respondents are grouped by seven-point party identification with leaners folded in: Democrats are 1–3, Republicans 5–7. A “pro-life Democrat” is a Democratic identifier choosing category 1; a “pro-choice Republican” is a Republican choosing category 4. The vote correlation is the weighted Pearson correlation of the four-category scale with a Democratic presidential vote, computed only in presidential-election years.
Denominator and the 2008 outlier. Every share is computed among respondents who answered the abortion item, not among all respondents — the natural denominator for an attitude, and a deliberate choice that matters most in 2008, when roughly fifty-five percent of the sample was not asked or did not answer the item. Counting those non-responses as if they were “not always legal” would have halved the 2008 figures and manufactured a salience dip that is really a coverage gap. On the opinion-holder denominator used here, 2008 still reads below its neighbors — a genuinely lower-salience, narrower-gap year — and is flagged as an outlier rather than smoothed away. (An earlier internal diagnostic used the all-respondents denominator and so reported lower 2008 levels and a slightly smaller endpoint gap; the opinion-holder figures are the ones reported here.)
Honest limits. The 2020 and 2024 waves are mixed-mode; the partisan gap is enormous in both, so the trend is not a mode artifact, but the exact 2024 level sits at the end of a changing instrument. Item non-response is also elevated in 2004. The correlation is descriptive, not causal; the post-Dobbs panel cited below is the appropriate source for the direction of effect.