Abortion Opinion, 1980-2024
The most divisive issue in American politics, and
whether opinion has actually moved
Democracy’s Data
Last updated September 15, 2026 at 4:42p EDT
Abortion is the most divisive issue in American politics.
You have heard the sentence, and it smuggles in an assumption: a country
this torn must be a country whose views are moving. Forty-four years of
the same survey question can check that. This brief asks the question
the sentence skips: has abortion opinion actually moved — and if so,
whose?
The short answer, visible in the figures below, is that the country’s
answers moved far less than its politics did. What moved was who gives
which answer. That is a different kind of change, and the headline
number cannot show it.
The test
The measure is a four-option abortion question — an item, in
survey language — numbered VCF0838 in the ANES Time Series
Cumulative Data File, 1948–2024. That file is collected by the
University of Michigan and Stanford University for the American National
Election Studies (ANES), the academic survey that has interviewed a
national sample of Americans — a few thousand people chosen at random to
stand in for the country — in every presidential election since 1948. It
is free, but must be requested through a browser with an account at electionstudies.org.
The item has been read from the same printed card, handed to the
respondent by the interviewer, in 17 studies from 1980 to 2024. The ANES chapter introduces the
cumulative file and its codebook; two companion briefs, party identification and ideology, built the
identities this brief splits by.
Only a survey can hold this question, and no election ever grades the
answer. A referendum counts votes on one state’s one proposal; it never
measures what the country thinks about abortion in general, so there is
no result to score these answers against. The method has to be the
credential, and the credential here is constancy: the same four options,
in the same words, for forty-four years. A claim about change
is only as good as the thing that stayed the same.
The test runs in three parts. First, read the most recent study on
its own, because a trend line is unreadable until you know what a single
row means. Second, split that study by party and by ideology and measure
both gradients — how steadily an answer climbs or falls from one end of
each scale to the other. Third, follow all four options across every
year, and follow both gradients with them.
The data
Here is the question as the respondent meets it, verbatim, with five
real rows of the file underneath.
The respondent is handed a booklet and asked which opinion best
agrees with their view. The four options, verbatim:
1 By law, abortion should never be permitted.
2 The law should permit abortion only in case of rape, incest,
or when the woman's life is in danger.
3 The law should permit abortion for reasons other than rape,
incest, or danger to the woman's life, but only after the need
for the abortion has been clearly established.
4 By law, a woman should always be able to obtain an abortion as
a matter of personal choice.
9 DK; other
Five random rows from the 2024 study, with the two identity
columns this chapter uses alongside:
year VCF0838 VCF0301 VCF0803 abortion_view
2024 3 3 5 Other reasons, but only after the
2024 3 3 4 Other reasons, but only after the
2024 4 1 2 Always, as a matter of personal ch
2024 4 1 1 Always, as a matter of personal ch
2024 4 2 4 Always, as a matter of personal ch
Option 3 is the one to watch. It is not a middle position between
1 and 4 -- it is a permission with a gatekeeper, and it is where
the movement of the last four years went.
Option 3 deserves a second read, because it is not a midpoint between
never and always. It is a permission with a gatekeeper —
abortion allowed, but only once someone has established that the need is
real. A person choosing it has not split the difference; they have
chosen a procedural answer to a question the other three options answer
in principle. Keep it in view. It is where the last four years went.
Three decisions about how to sort the answers were made when these
tables were built, and each is argued for elsewhere rather than repeated
here. Party identification is kept at all seven points, because the party identification brief
shows that collapsing the scale is exactly where the leaners disappear.
Ideological self-placement — where people put themselves on a scale from
extremely liberal to extremely conservative — keeps the people who said
haven’t thought much about it as their own group, because the
ideology brief shows the
codebook files that answer under Valid. And code 9 on
the abortion item itself, don’t know, is dropped, along with
respondents not asked.
Five tables come out. now.csv is the most recent study’s
four shares. by_party.csv and by_ideology.csv
split that study by the two identities, one row per group, with each
group’s full distribution. overtime.csv runs the four
options across every study year, with a pct_missing column
— 2008 is 55.3% missing on this item because, as the codebook records, a
randomly chosen half of that year’s respondents was asked a different
set of abortion questions instead. gaps.csv subtracts:
Democrats minus Republicans, and liberals minus conservatives, in the
share saying always, for every year.
Before you read on, commit to a view. By the end of
this brief you will have seen a party gradient and an ideology gradient
on this question, both strong. Which of the two do you expect
was already strong in 1980, and which one had to be built?
Write down your answer.
What it says about democracy
Start with what Americans said in 2024.
| Never permitted |
293 |
5.8% |
| Only for rape, incest, or danger to the woman’s
life |
1,166 |
23.2% |
| Other reasons, but only after the need is
established |
1,059 |
21.1% |
| Always, as a matter of personal choice |
2,503 |
49.9% |
Half of the people asked, 49.9%, take the fourth option: abortion as
a matter of personal choice. 5.8% take the first, never permitted. The
two absolute answers hold half the country between them, and the two
conditional ones hold the rest.
Now the same study, split by the seven-point party scale.
| Strong Democrat |
1,207 |
1.4% |
6.7% |
12.8% |
79.0% |
| Weak Democrat |
582 |
1.9% |
8.8% |
18.7% |
70.6% |
| Independent-Democrat |
670 |
1.5% |
8.1% |
20.1% |
70.3% |
| Independent-Independent |
330 |
5.5% |
19.1% |
23.3% |
52.1% |
| Independent-Republican |
624 |
8.5% |
38.5% |
27.2% |
25.8% |
| Weak Republican |
524 |
7.6% |
34.5% |
27.7% |
30.2% |
| Strong Republican |
1,059 |
13.2% |
46.3% |
24.8% |
15.7% |
Figure 1. The seven-point party scale against the abortion
item, 2024. Bars across the party scale, because the claim is
about a gradient — how steadily an answer climbs or falls from strong
Democrat to strong Republican. The buttons switch which of the four
options is drawn, and hovering a bar gives that group’s whole
distribution. Option 4, the one that gets reported, runs from 15.7% to
79.0%, a spread of 63.3 points; option 3, the gatekeeper answer, runs
only from 12.8% to 27.7%. Roughly a quarter as steep, which is how it
can absorb movement without registering as a partisan story.
The gradient runs the way you would expect, with one thing worth
stopping on. Leaning Republicans are further from the pro-choice
position than weak Republicans, at 25.8% against 30.2%. The
people who decline the party label hold the more conservative view on
this issue than the people who accept it. That is the same pattern the
party identification brief found in vote choice, showing up in a second
place. Fold the leaners into “independent,” as the file’s own
three-category column invites, and you would report an independent group
near 52.1% while the actual unaffiliated middle is far smaller.
Ideology sorts the same study even more sharply.
| Extremely liberal |
235 |
3.8% |
4.7% |
88.1% |
| Liberal |
758 |
0.8% |
9.2% |
86.3% |
| Slightly liberal |
515 |
1.4% |
20.2% |
72.4% |
| Moderate, middle of the road |
1,085 |
2.6% |
25.8% |
54.0% |
| Slightly conservative |
532 |
4.9% |
31.4% |
34.2% |
| Conservative |
928 |
9.9% |
26.2% |
15.1% |
| Extremely conservative |
278 |
27.7% |
13.3% |
12.6% |
| Haven’t thought much about it |
679 |
7.1% |
21.5% |
47.1% |
The always-permitted share runs from 88.1% among the extremely
liberal to 12.6% among the extremely conservative. And one row exists
only because the ideology brief refused to throw it away. People who
said they haven’t thought much about liberal and
conservative are 47.1% in favour of the always option. That sits between
the moderates (54.0%) and the slightly conservative (34.2%), and nowhere
near the midpoint of the scale. They have a definite position; what they
decline is the vocabulary. Dropping them as missing would have removed
679 people with opinions from a table about opinions.
Now the whole series.
Figure 2. All four options, every study year the item was
asked. Scrub across the plot to read all four shares at any
year; the buttons mute a series so the others can be compared without
it. Hollow points mark 2008, where the split half-sample leaves the item
55.3% missing. The line is drawn through that year rather than bridged
silently, because a gap that has been filled in looks exactly like
data.
The top line is the one that gets reported, and it is the one that
has moved least. Always permitted was 36.8% in 1980 and 49.9%
in 2024 — a rise of 13.1 points across forty-four years, most of it
arriving by the early 1990s. Underneath, between 2020 and 2024:
| Never permitted |
11.0% |
5.8% |
-5.2 |
| Only rape, incest, or life |
24.3% |
23.2% |
-1.1 |
| Only after the need is established |
13.9% |
21.1% |
+7.2 |
| Always, personal choice |
50.8% |
49.9% |
-0.9 |
The absolutist position roughly halved, and the mass
did not go to the opposite pole — always permitted fell
slightly. It went into option 3, the permission with a gatekeeper, which
gained 7.2 points. A reader watching only the headline number would
conclude that nothing much happened between 2020 and 2024. Something
quite large happened, and it happened in the categories nobody
reports.
The last figure is the one the prediction prompt was about.
Figure 3. The party gap and the ideology gap in the share
saying always permitted. Democrats minus Republicans, leaners
included on both sides; liberals minus conservatives. Higher means
further apart. Hovering reports the four shares behind the two
differences, which is where the interesting part is.
The gap is a subtraction. In 1980, 37.3% of Democrats and 33.8% of
Republicans took the always option, and 37.3 − 33.8 = 3.5: the
party gap on abortion was 3.5 points. Democrats and
Republicans held almost the same distribution of views; abortion cut
across the parties rather than between them. By 2024 the gap is
52.7 points, and it widened mostly because one side
moved: the Democratic share saying always rose 37.4 points
between 1980 and 2024 while the Republican share fell 11.8. The ideology
gap started at 23.0 points — already wide in 1980. Liberal and
conservative meant something about abortion before the party labels did,
and that gap widened too, to 61.3.
This is the finding the three-part test was built to deliver. The
aggregate looks like a country whose mind changed slowly. Split it, and
you see a country whose mind changed much less than its sorting
did: roughly the same distribution of opinions, redistributed onto
different teams. That is what “most divisive issue” turns out to mean in
this file — not a country changing its answer, but a country
reorganizing its parties around an answer it already had.
What this brief cannot tell you
2008 is 55.3% missing on this item, and that is
documented rather than broken: a random half sample received a branching
series of questions instead of the four-option card. Read the codebook
note and your sample is halved for a known reason; miss it and 2008
looks like a collapse in response.
Four options are not a scale. The distance from
never to rape, incest or life is not the same quantity
as the distance from need established to always, so
this brief reports shares and differences of shares, and never a mean. A
mean of these four numbers would be arithmetic on labels.
Nothing here is weighted — no answer is counted more
than another to make the sample’s mix match the country’s — so every
percentage describes the people ANES interviewed rather than the
country. The file’s weight column, VCF0009z, is there for a
reader who wants to apply it.
The 2020-to-2024 movement invites a Dobbs story, and
this brief does not tell one. Dobbs v. Jackson Women’s
Health Organization is the June 2022 decision in which the Supreme
Court overturned Roe v. Wade and returned abortion law to the
states, and it falls squarely between the two studies. Two studies four years
apart cannot separate a court decision’s effect from everything else
that happened, and the codebook records no wording change that would
explain it either. What can be said is what the table says: the movement
was out of the absolute prohibition and into the conditional permission,
and not toward personal choice.
And the two gradients are not independent of each
other. Liberals are mostly Democrats and conservatives mostly
Republicans, and this brief never holds one identity constant while
varying the other. Which of the two is doing the work is a real question
and a different brief.
What you should have learned
- The headline barely moved: always permitted went from 36.8%
in 1980 to 49.9% in 2024, a rise of 13.1 points in forty-four
years.
- The party gap did move: from 3.5 points in 1980 to 52.7 in 2024,
mostly because one side changed its answer.
- The ideology gap was already 23.0 points in 1980 — the ideological
sorting of abortion predates the partisan sorting.
- Between 2020 and 2024 the never option roughly halved, and
the movement went into the gatekeeper option, which gained 7.2 points —
a real change the headline number cannot see.
- People who decline the ideological vocabulary still have a position:
47.1% of them take the always option, between the moderates and the
slightly conservative.
Extensions
None of these is answered above, and every one can be answered from
the tables the Sources section hands over.
gaps.csv carries ind_always and
notplaced_always, two middle groups no figure above draws.
Follow both across the years. Do they move with the country, or with one
party?by_party.csv has the full distribution for every group,
including pct_2, the rape-incest-life option. Which groups
choose it most, and does it run with the gradient or against it?overtime.csv has all four options by year. Find the
year each option peaks. Does any of the four ever hold a majority?overtime.csv has pct_missing. Set 2008
aside and find the next-highest years. Is item-missingness drifting up
as the politics heat up, or holding steady?
Two stretch ideas point past the spreadsheet. The General Social
Survey asks its own battery of abortion questions, in a file anyone can
download from NORC at the University of Chicago at https://gss.norc.org/us/en/gss/get-the-data.html —
rebuild Figure 2’s flat top line from abany and check
whether it replicates. And after 2022, several states put abortion
directly on the ballot; find two of those referendum results and ask
which of the four ANES options the winning side’s share sits closest
to.
Sources
Data
- American National Election Studies. ANES Time Series Cumulative
Data File (1948–2024), version of 5 February 2026. Ann Arbor, MI
and Palo Alto, CA: University of Michigan and Stanford University. https://electionstudies.org/data-center/anes-time-series-cumulative-data-file/
The file must be requested through a browser with a free account; the ANES chapter describes what it costs
to get.
- ANES. Time Series Cumulative Data File Codebook, same
release. The four options and the 2008 split-sample note are quoted from
it directly.
Academic literature
Court decisions
The data itself
Every figure above is drawn from these tables. They are plain CSV,
they open in a spreadsheet, and they are yours to take further.
by_ideology.csv, by_party.csv, gaps.csv, now.csv, overtime.csv
