Chapter 4.1

The abortion sort

In the American National Election Studies the gap between Democrats and Republicans in support for always-legal abortion grew from 3.5 points in 1980 to 52.3 in 2024, mostly on the Democratic side, and the General Social Survey shows the same widening.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 30 September 2026

In the American National Election Studies (ANES), 73.7% of Democrats and 21.4% of Republicans said in 2024 that abortion should always be allowed as a matter of personal choice, a gap of 52.3 points. In 1980 the two shares were 37.3% and 33.8%, a gap of 3.5 points that is too small to distinguish from zero, so the gap between the parties widened over these years. The General Social Survey (GSS) asks a different question, whether a legal abortion should be possible if the woman wants it for any reason, and shows the same widening, with a gap of 2.5 points in 1980, also too small to distinguish from zero, and 47.5 points in 2024.

Both surveys ask for a position on legal abortion, and this chapter counts the most permissive answer in each: "always allowed" in the ANES and "yes" in the GSS.The ANES question asks which of four statements comes closest to the respondent's view, from "abortion should never be permitted by law" to "a woman should always be able to obtain an abortion as a matter of personal choice." The GSS question is one of seven about circumstances in which a legal abortion should be possible, and "any reason" is the most permissive. A gap is the Democratic share minus the Republican share, in points, so a positive gap means that Democrats were more likely than Republicans to give the permissive answer. Party includes leaners, and every interval is a 95% confidence interval. The change is one case of party sorting, in which an opinion that once cut across the parties comes to line up with them.Shares use the survey weights and cover respondents who answered the question: 45,017 in 17 ANES studies from 1980 to 2024 and 41,398 in 29 GSS surveys from 1977 to 2024. The ANES cumulative file gives every respondent the same weight from 1980 to 1990, so those years are unweighted. Intervals use the Kish effective sample size [1] and treat the weighted sample as if respondents had been drawn at random one by one. Both surveys draw respondents in stages, area by area, and the intervals ignore that, which can make them narrower or wider.

Political scientists call the path by which a new issue becomes part of the party system issue evolution [2]. Adams applied the account to abortion with congressional roll calls and General Social Survey data from 1972 to 1994 [3]. O'Brian finds that by 1970 opposition to abortion among ordinary voters correlated with conservative views on other issues, and by 1972 with the Republican presidential vote [4]. The two surveys used here begin in 1977 and 1980 and follow the question through 2024.

The gap opened in the early 1990s

Figure 1 plots the gap in every survey year. In the ANES studies from 1980 to 1990 the interval includes zero in 5 of 6; the exception is 1984, at 6.6 (95% CI 2.3 to 10.9) points. From 1992 on, every ANES interval lies above zero. The GSS interval includes zero in 9 of the 12 surveys from 1977 to 1991. In 1977 and 1983 it lies below zero, so Republicans were more likely than Democrats to say yes, and in 1988 it lies above zero.The 1977 GSS interval treats respondents as drawn one by one. With the stages of the sample design taken into account, the 1977 gap is −6.5 (95% CI −13.6 to 0.7) points and its interval includes zero, so whether 1977 lies below zero depends on the interval method. From 1993 on the GSS interval stays above zero. The two surveys agree that there was no consistent party gap in the 1980s and that one appeared in the early 1990s.

Figure 1

Democrats and Republicans grew apart on legal abortion in both surveys

Democratic minus Republican share saying abortion should always be allowed (ANES) or legal for any reason (GSS), party identifiers including leaners, 1977 to 2024; 95% intervals

Show data

Data for Democrats and Republicans grew apart on legal abortion in both surveys
Survey	Year	Gap (points)	Lower	Upper	Respondents
ANES	1980	3.5	−1.8	8.8	1,313
ANES	1982	−0.1	−5.8	5.6	1,173
ANES	1984	6.6	2.3	10.9	1,888
ANES	1986	2.7	−1.8	7.3	1,818
ANES	1988	3.5	−0.9	8.0	1,751
ANES	1990	4.7	−0.1	9.4	1,692
ANES	1992	14.1	9.8	18.3	2,097
ANES	1994	12.4	7.1	17.8	1,537
ANES	1996	12.3	6.8	17.9	1,521
ANES	1998	15.5	9.2	21.7	1,106
ANES	2000	16.2	10.8	21.5	1,532
ANES	2004	18.9	12.2	25.5	931
ANES	2008	21.7	13.8	29.7	893
ANES	2012	30.0	26.4	33.5	5,068
ANES	2016	32.9	29.2	36.6	3,630
ANES	2020	48.1	45.3	50.9	7,002
ANES	2024	52.3	48.8	55.9	4,666
GSS	1977	−6.5	−12.7	−0.3	1,299
GSS	1978	−3.8	−10.6	3.0	1,263
GSS	1980	2.5	−4.6	9.5	1,165
GSS	1982	−0.9	−7.7	5.9	1,532
GSS	1983	−7.4	−14.0	−0.8	1,322
GSS	1984	−4.7	−11.7	2.2	1,245
GSS	1985	1.9	−4.6	8.5	1,322
GSS	1987	−2.4	−8.7	3.8	1,527
GSS	1988	9.2	1.8	16.5	817
GSS	1989	4.5	−2.9	11.9	861
GSS	1990	6.4	−1.6	14.3	766
GSS	1991	2.7	−4.9	10.3	816
GSS	1993	15.3	7.9	22.7	872
GSS	1994	8.6	3.2	14.1	1,647
GSS	1996	11.3	5.8	16.8	1,513
GSS	1998	14.5	8.9	20.1	1,449
GSS	2000	12.2	6.5	17.9	1,378
GSS	2002	19.0	10.8	27.1	719
GSS	2004	17.4	9.2	25.7	700
GSS	2006	19.9	13.9	25.8	1,486
GSS	2008	11.9	4.8	19.0	1,051
GSS	2010	19.2	11.6	26.8	968
GSS	2012	23.9	16.3	31.5	966
GSS	2014	23.5	17.2	29.7	1,290
GSS	2016	21.5	15.3	27.7	1,465
GSS	2018	31.0	23.9	38.0	1,189
GSS	2021	36.8	29.6	44.0	1,044
GSS	2022	53.5	44.7	62.3	831
GSS	2024	47.5	41.4	53.5	1,534

Both lines rise from a gap near zero in 1980 to their highest values in the 2020s, but neither rises at every step. A value above zero means Democrats were more likely than Republicans to say yes. Each line joins survey years, each band is a 95% confidence interval, and the shaded band marks the change in ANES survey mode between 2016 and 2020.Note: The 2012 and 2016 ANES studies combined in-person and web samples, and the 2020 study had no in-person interviews.Note: GSS interviews were mostly in person through 2018, mostly on the web in 2021, and a mix of in-person and web in 2022 and 2024.Note: The GSS question was first asked in 1977. The 2008 ANES asked the question of fewer respondents than any other study, so its interval is wider; its estimate lies between those of 2004 and 2012, and this chapter treats 2008 as an ordinary year.Source: ANES Cumulative Data File, 1980-2024, and General Social Survey, 1977-2024; weighted (VCF0009z, which equals 1 for every respondent from 1980 to 1990; weight_use)

From 1992 to 2000 the ANES gap was between 12.3 and 16.2 points in every study, and no step between adjacent studies in those years has an interval that excludes zero. From 2000 to 2016 the estimate rises in each study, by 16.7 (95% CI 10.2 to 23.2) points in all, but the change between adjacent studies has an interval that includes zero in 5 of the 6 steps from 2000 to 2024. The exception is the step from 2016 to 2020, taken up below.

In the GSS the gap was lower in 2024 than in 2022 by 6.0 (95% CI −4.7 to 16.7) points, a difference whose interval includes zero. The GSS estimates from 1993 to 2016 fall as well as rise between adjacent surveys, and no step between them has an interval that excludes zero.

The 2016 to 2020 rise coincides with a mode change

The largest change between adjacent ANES studies is the one from 2016 to 2020: the gap rose by 15.2 (95% CI 10.6 to 19.8) points, from 32.9 to 48.1. The step from 1990 to 1992, 9.4 (95% CI 3.0 to 15.8) points, is the only other one whose interval excludes zero, and the two intervals overlap, so the data do not rank the two steps. The 2020 study also changed how people were interviewed. The ANES interviewed respondents mainly in person through 2008 and added a web sample next to the in-person sample in 2012, 2016 and 2024, with separate weights for each mode. The 2020 study conducted no in-person interviews, and answers were collected online, by telephone or by video [5]. A change of survey mode can change answers, and the 2020 study has no separate weights by mode, so the ANES cannot separate the mode change from a change in opinion on its own.

Four checks bear on the step (Figure 2). Three compare respondents interviewed in the same mode: ANES web respondents in 2016 with the 2020 study, which was collected mostly online, and GSS in-person and web respondents before 2022 with the 2022 and 2024 surveys combined, when the GSS kept in-person interviewing next to a web sample. The GSS web comparison starts in 2021, the first GSS survey conducted mostly on the web (89.2% of respondents who answered the question, counted without weights). The gap rose in each of the three, by between 14.1 and 16.9 points, and each interval lies above zero. The fourth check is the 2000 ANES, which interviewed part of its sample by telephone [6]; its gap of 16.2 (95% CI 10.8 to 21.5) points lies between the 1998 gap of 15.5 points and the 2004 gap of 18.9 points.

Figure 2

The party gap rose by similar amounts within each mode

Democratic minus Republican gap in saying yes: change within one interview mode (top) and web minus in-person difference in the same study (bottom); leaners included; 95% intervals

Change in the party gap within one mode

Web minus in-person party gap, same study

Change or difference (percentage points)

Show data

Data for The party gap rose by similar amounts within each mode
Panel	Comparison	Gaps compared	First gap (points)	Lower	Upper	Second gap (points)	Lower	Upper	Second minus first (points)	Lower	Upper	Respondents
Change in the party gap within one mode	ANES 2016 to 2020	2016 (all respondents), then 2020	32.9	29.2	36.6	48.1	45.3	50.9	15.2	10.6	19.8	10,632
Change in the party gap within one mode	ANES web only	2016 (web respondents), then 2020 (all respondents)	31.2	26.9	35.5	48.1	45.3	50.9	16.9	11.7	22.0	9,602
Change in the party gap within one mode	GSS in person	2018 (in person), then 2022 and 2024 combined (in person)	31.6	24.2	39.0	46.4	38.1	54.8	14.9	3.7	26.0	1,954
Change in the party gap within one mode	GSS web	2021 (web), then 2022 and 2024 combined (web)	38.5	30.9	46.1	52.6	46.0	59.3	14.1	4.1	24.2	2,148
Web minus in-person party gap, same study	ANES 2012	In person, then web	26.0	19.8	32.1	32.2	27.9	36.5	6.3	−1.3	13.8	5,068
Web minus in-person party gap, same study	ANES 2016	In person, then web	37.3	30.4	44.3	31.2	26.9	35.5	−6.1	−14.3	2.1	3,630
Web minus in-person party gap, same study	ANES 2024	In person, then web	52.3	44.8	59.7	53.8	48.6	59.1	1.6	−7.5	10.7	2,714
Web minus in-person party gap, same study	GSS 2022 and 2024	In person, then web	46.4	38.1	54.8	52.6	46.0	59.3	6.2	−4.5	16.8	2,102

Every rise in the top panel is above zero and the intervals share a range. In the bottom panel the web minus in-person differences change sign and every interval includes zero, but the intervals are wide. Dots are estimates and lines are 95% confidence intervals. The GSS rows compare 2018 (in person) or 2021 (web) with the 2022 and 2024 surveys combined, the ANES web row compares web respondents in 2016 with all respondents in 2020, and the 2024 ANES row uses only the newly recruited in-person and web samples.Note: The 2020 ANES study has no separate weights by mode, so it has no row in the bottom panel.Note: The GSS rows combine 2022 and 2024 because one year alone leaves too few in-person or web respondents in each party.Source: ANES Cumulative Data File, 2012, 2016, 2020 and 2024, and General Social Survey, 2018 to 2024; weighted (weight_ftf and weight_web, the ANES weights for each mode; weight_use)

The direct comparison of web and in-person respondents in the same study, in the lower panel, changes sign across the ANES years and has an interval that includes zero in every row. The intervals are wide: their ends run from −14.3 to +16.8 points, so they do not rule out a mode difference of the size of the step. These checks do not rule out a mode effect that appears only in 2020. They show that the gap also rose by similar amounts among respondents interviewed in the same mode.

The GSS rise in 2022 and the Dobbs decision

In June 2022 the Supreme Court decided Dobbs v. Jackson Women's Health Organization, which overturned Roe v. Wade [7]. The two surveys place the last increase differently around it. The ANES gap was already 48.1 (95% CI 45.3 to 50.9) points in 2020 and rose by a further 4.2 (95% CI −0.3 to 8.7) points to 2024, an interval that includes zero. The GSS gap rose by 16.7 (95% CI 5.3 to 28.1) points between 2021 and 2022 (Figure 1). Weighted, 89.2% (95% CI 86.4% to 91.9%) of the 2022 respondents who answered the question were interviewed on or after the day of the decision (82.6% without weights). That rise mixes a change from a mostly online survey to one that combined in-person and web interviews with any change in opinion, so the same-mode comparisons above are the firmer evidence. Two studies of the period look at other outcomes. Lockhart, Gerber and Huber report, from a large panel that spans the decision, that partisans whose abortion preferences did not match their party were more likely to change their party identification afterward [8]. Cassese, Ondercin and Randall report a shift in the link between partisanship and the intensity of abortion preferences [9]. The data used here cannot observe either change, because they have no panel and no measure of intensity.

Democratic support rose by more than Republican support fell

A gap can widen because one party moves, the other moves or both do. Figure 3 shows the shares behind it. In the ANES the Democratic share rose by 36.4 (95% CI 32.2 to 40.6) points between 1980 and 2024 and the Republican share fell by 12.4 (95% CI 7.6 to 17.2) points. The GSS series starts in 1977, before the ANES series, when its gap was −6.5 (95% CI −12.7 to −0.3) points. In the GSS the Democratic share rose by 45.3 (95% CI 40.0 to 50.7) points between 1977 and 2024, from 35.2% to 80.6%, and the Republican share fell by 8.6 (95% CI 1.8 to 15.4) points, from 41.7% to 33.1%. Most of that GSS Republican fall came before 1998: the share was 31.1% in 1998 and has been between 28.7% and 35.0% in every GSS survey since. The questions differ, so the GSS levels are not the ANES levels, and the comparison is of direction and timing.Among all respondents who answered, including independents who lean toward neither party, the share rose by 12.1 (95% CI 9.0 to 15.3) points in the ANES and 20.6 (95% CI 16.5 to 24.6) points in the GSS, less than the Democratic share rose in each survey.

Figure 3

Democratic support for legal abortion rose while Republican support fell

Share saying abortion should always be allowed (ANES) or legal for any reason (GSS), by party identification including leaners; 95% intervals

Show data

Data for Democratic support for legal abortion rose while Republican support fell
Survey and item	Group	Year	Share (%)	Lower	Upper	Respondents
ANES: always allowed	Democrats	1980	37.3	34.0	40.7	801
ANES: always allowed	Democrats	1982	36.0	32.6	39.5	741
ANES: always allowed	Democrats	1984	39.4	36.5	42.4	1,032
ANES: always allowed	Democrats	1986	40.9	37.9	43.9	1,061
ANES: always allowed	Democrats	1988	37.0	33.9	40.1	938
ANES: always allowed	Democrats	1990	43.3	40.2	46.4	993
ANES: always allowed	Democrats	1992	52.8	49.9	55.6	1,201
ANES: always allowed	Democrats	1994	46.4	42.5	50.2	810
ANES: always allowed	Democrats	1996	46.1	42.4	49.8	879
ANES: always allowed	Democrats	1998	47.2	43.0	51.4	647
ANES: always allowed	Democrats	2000	47.2	43.5	50.9	868
ANES: always allowed	Democrats	2004	44.6	39.7	49.4	498
ANES: always allowed	Democrats	2008	51.0	45.9	56.2	600
ANES: always allowed	Democrats	2012	59.0	56.5	61.4	3,081
ANES: always allowed	Democrats	2016	60.4	57.7	63.0	1,915
ANES: always allowed	Democrats	2020	72.1	70.1	74.1	3,713
ANES: always allowed	Democrats	2024	73.7	71.2	76.2	2,459
ANES: always allowed	Republicans	1980	33.8	29.7	37.9	512
ANES: always allowed	Republicans	1982	36.1	31.6	40.6	432
ANES: always allowed	Republicans	1984	32.8	29.7	36.0	856
ANES: always allowed	Republicans	1986	38.2	34.7	41.6	757
ANES: always allowed	Republicans	1988	33.5	30.2	36.7	813
ANES: always allowed	Republicans	1990	38.6	35.0	42.2	699
ANES: always allowed	Republicans	1992	38.7	35.5	41.9	896
ANES: always allowed	Republicans	1994	33.9	30.1	37.7	727
ANES: always allowed	Republicans	1996	33.7	29.6	37.9	642
ANES: always allowed	Republicans	1998	31.7	27.1	36.4	459
ANES: always allowed	Republicans	2000	31.0	27.2	34.9	664
ANES: always allowed	Republicans	2004	25.7	21.3	30.2	433
ANES: always allowed	Republicans	2008	29.3	23.3	35.3	293
ANES: always allowed	Republicans	2012	29.0	26.4	31.6	1,987
ANES: always allowed	Republicans	2016	27.5	24.9	30.0	1,715
ANES: always allowed	Republicans	2020	24.0	22.1	25.9	3,289
ANES: always allowed	Republicans	2024	21.4	18.9	23.8	2,207
GSS: any reason	Democrats	1977	35.2	31.6	38.9	852
GSS: any reason	Democrats	1978	32.1	28.0	36.2	781
GSS: any reason	Democrats	1980	40.7	36.3	45.2	725
GSS: any reason	Democrats	1982	40.4	36.5	44.4	1,052
GSS: any reason	Democrats	1983	30.5	26.3	34.6	810
GSS: any reason	Democrats	1984	36.2	31.9	40.6	731
GSS: any reason	Democrats	1985	39.2	34.6	43.7	725
GSS: any reason	Democrats	1987	40.2	36.3	44.1	990
GSS: any reason	Democrats	1988	41.6	36.6	46.6	464
GSS: any reason	Democrats	1989	41.3	36.2	46.5	459
GSS: any reason	Democrats	1990	46.3	40.5	52.1	364
GSS: any reason	Democrats	1991	43.9	38.4	49.4	407
GSS: any reason	Democrats	1993	52.6	47.5	57.7	473
GSS: any reason	Democrats	1994	50.3	46.7	54.0	925
GSS: any reason	Democrats	1996	49.3	45.4	53.2	830
GSS: any reason	Democrats	1998	45.6	41.8	49.4	846
GSS: any reason	Democrats	2000	43.4	39.4	47.3	785
GSS: any reason	Democrats	2002	48.8	43.0	54.6	397
GSS: any reason	Democrats	2004	48.9	42.9	54.8	376
GSS: any reason	Democrats	2006	49.8	45.6	54.0	834
GSS: any reason	Democrats	2008	47.0	42.1	51.8	614
GSS: any reason	Democrats	2010	51.1	45.9	56.2	560
GSS: any reason	Democrats	2012	52.6	47.7	57.6	580
GSS: any reason	Democrats	2014	56.2	52.0	60.3	775
GSS: any reason	Democrats	2016	56.3	52.1	60.4	877
GSS: any reason	Democrats	2018	65.4	60.8	69.9	689
GSS: any reason	Democrats	2021	70.9	66.4	75.3	637
GSS: any reason	Democrats	2022	83.3	78.3	88.2	476
GSS: any reason	Democrats	2024	80.6	76.7	84.5	810
GSS: any reason	Republicans	1977	41.7	36.7	46.8	447
GSS: any reason	Republicans	1978	36.0	30.5	41.4	482
GSS: any reason	Republicans	1980	38.3	32.8	43.7	440
GSS: any reason	Republicans	1982	41.3	35.8	46.8	480
GSS: any reason	Republicans	1983	37.9	32.7	43.1	512
GSS: any reason	Republicans	1984	41.0	35.6	46.4	514
GSS: any reason	Republicans	1985	37.2	32.5	42.0	597
GSS: any reason	Republicans	1987	42.7	37.7	47.6	537
GSS: any reason	Republicans	1988	32.4	27.1	37.8	353
GSS: any reason	Republicans	1989	36.8	31.6	42.1	402
GSS: any reason	Republicans	1990	39.9	34.5	45.3	402
GSS: any reason	Republicans	1991	41.2	35.9	46.4	409
GSS: any reason	Republicans	1993	37.3	31.9	42.7	399
GSS: any reason	Republicans	1994	41.7	37.7	45.7	722
GSS: any reason	Republicans	1996	38.0	34.0	42.0	683
GSS: any reason	Republicans	1998	31.1	27.0	35.3	603
GSS: any reason	Republicans	2000	31.2	27.1	35.3	593
GSS: any reason	Republicans	2002	29.8	24.1	35.6	322
GSS: any reason	Republicans	2004	31.4	25.7	37.1	324
GSS: any reason	Republicans	2006	30.0	25.8	34.2	652
GSS: any reason	Republicans	2008	35.0	29.8	40.3	437
GSS: any reason	Republicans	2010	31.8	26.3	37.4	408
GSS: any reason	Republicans	2012	28.7	23.0	34.5	386
GSS: any reason	Republicans	2014	32.7	28.1	37.3	515
GSS: any reason	Republicans	2016	34.8	30.1	39.4	588
GSS: any reason	Republicans	2018	34.4	29.1	39.8	500
GSS: any reason	Republicans	2021	34.1	28.4	39.7	407
GSS: any reason	Republicans	2022	29.8	22.5	37.1	355
GSS: any reason	Republicans	2024	33.1	28.5	37.7	724

The Democratic line climbs while the Republican line moves down, so the gap opens mostly from the Democratic side. Bands are 95% confidence intervals and the shaded band marks a change in survey mode. Switch the view for the other survey; the two questions differ, so only direction and timing are comparable.Note: Shaded band: the ANES mode change from 2016 to 2020 in the ANES view, and the GSS move from in-person to web interviews between 2018 and 2021 in the GSS view.Source: ANES Cumulative Data File, 1980-2024, and General Social Survey, 1977-2024; weighted (VCF0009z, which equals 1 for every respondent from 1980 to 1990; weight_use)

The gap widened within every attendance, education and birth-cohort group

A widening gap in the whole electorate could come from one kind of voter. Figure 4 splits the gap by group and compares the 1980s with the 2020s. Among all ANES party identifiers the gap was 3.6 (95% CI 1.5 to 5.7) points in the 1980s and 49.8 (95% CI 47.6 to 52.0) points in the 2020s, and in every group of either survey the 2020s interval lies above the 1980s interval.

Figure 4

The party gap widened in every group

Democratic minus Republican share saying yes, 1980s against 2020s, by group; party identifiers including leaners; 95% intervals

1980s
2020s

Show data

Data for The party gap widened in every group
Survey and item	Group	1980s gap (points)	Lower	Upper	Respondents	2020s gap (points)	Lower	Upper	Respondents
ANES: always allowed	All partisans	3.6	1.5	5.7	7,943	49.8	47.6	52.0	11,668
ANES: always allowed	Weekly attenders	3.8	0.9	6.6	3,018	42.1	37.7	46.6	3,119
ANES: always allowed	Monthly attenders	−0.1	−6.0	5.8	1,060	43.8	36.0	51.5	986
ANES: always allowed	Yearly or never	2.8	−0.4	6.1	3,777	45.7	42.8	48.5	7,502
ANES: always allowed	High school or less	−2.1	−4.9	0.7	4,431	37.9	33.1	42.7	2,406
ANES: always allowed	Some college	10.0	5.6	14.4	1,913	51.8	48.0	55.7	3,798
ANES: always allowed	College graduate	20.7	15.8	25.6	1,544	56.0	53.0	59.0	5,343
ANES: always allowed	Born 1927 to 1942	0.7	−3.9	5.3	1,706	50.8	42.0	59.6	675
ANES: always allowed	Born 1943 to 1958	8.1	4.4	11.7	2,899	49.3	45.2	53.5	3,123
ANES: always allowed	Born 1959 to 1974	10.1	4.5	15.7	1,165	49.8	45.6	54.0	3,016
GSS: any reason	All partisans	−0.4	−2.8	2.0	9,791	45.5	41.3	49.7	3,409
GSS: any reason	Weekly attenders	3.3	−0.3	6.9	3,032	46.0	36.1	55.8	648
GSS: any reason	Monthly attenders	1.8	−3.4	6.9	2,193	42.6	32.2	53.0	573
GSS: any reason	Yearly or less	−5.7	−9.3	−2.0	4,539	38.1	32.7	43.6	2,165
GSS: any reason	No bachelor's degree	−1.1	−3.7	1.6	8,074	38.0	32.4	43.6	1,947
GSS: any reason	Bachelor's or higher	10.8	5.2	16.4	1,694	56.1	49.9	62.4	1,457
GSS: any reason	Born 1943 to 1958	2.7	−1.3	6.7	3,495	47.4	39.4	55.5	834
GSS: any reason	Born 1959 to 1974	3.1	−3.3	9.5	1,420	42.9	35.0	50.7	914

Every row moves to the right: in each group the 2020s gap is larger than the 1980s gap and the two intervals do not overlap. Each dumbbell joins the 1980s gap (open dot) to the 2020s gap (filled dot), and the lines are 95% confidence intervals. Birth cohorts appear only where both periods have enough respondents.Note: The 1980s are the ANES studies of 1980 to 1988 and the GSS surveys of 1980 to 1989; the 2020s are the ANES studies of 2020 and 2024 and the GSS surveys of 2021, 2022 and 2024.Note: ANES attendance groups: weekly or nearly weekly, monthly, and a few times a year or never. The 2024 attendance answers come from the ANES 2024 Time Series because the cumulative file's 2024 attendance leaves out people who said they never attend.Note: GSS attendance groups: weekly or more, monthly to nearly weekly, and a few times a year or less. The GSS degree question has no category for some college without a degree, so the GSS view splits at a bachelor's degree.Source: ANES Cumulative Data File, 1980-2024, with 2024 attendance from the ANES 2024 Time Series; weighted (VCF0009z, which equals 1 for every respondent from 1980 to 1990); General Social Survey, 1980-2024, weighted (weight_use); years combined after rescaling each year's weights to average one

The 2020s gap is not distinguishable between frequent and infrequent attenders in either survey: the difference between the two groups is −3.5 (95% CI −8.8 to 1.7) points in the ANES and +7.8 (95% CI −3.4 to 19.1) points in the GSS. Attendance relates to the level of support in both parties.In the 2020s 48.3% of Democrats and 6.2% of Republicans who attend weekly or almost weekly said abortion should always be allowed, against 79.3% and 33.6% among people who attend a few times a year or never. In the 1980s the GSS gap among infrequent attenders was −5.7 (95% CI −9.3 to −2.0) points, so Republicans were slightly more likely than Democrats to say yes. Chapter 9.1, The god gap in America takes up the political gap between frequent and infrequent worshippers.

Education starts from a different point. College graduates already differed by party in the 1980s, with a gap of 20.7 (95% CI 15.8 to 25.6) points, while the gap among people with a high school education or less was −2.1 (95% CI −4.9 to 0.7) points, an interval that includes zero. The GSS splits at a bachelor's degree and gives the same order in both periods.

Birth cohorts that were adults in both periods show a similar widening, and Figure 5 follows each cohort year by year. Comparing cohorts two at a time within a survey year, the difference has an interval that excludes zero in 3 of 52 comparisons. Each of them involves the cohort born 1927 to 1942, which had the smaller gap in 1986, 1988 and 1994. From 2012 on, the difference between any two cohorts drawn in the same year has an interval that includes zero. The cohorts born 1975 to 1990 and 1991 or later are drawn from 2012 and 2020, because their earlier cells are too small to plot. In the 2020s their gaps are 48.6 (95% CI 44.4 to 52.9) and 50.4 (95% CI 44.3 to 56.5) points, and both intervals overlap those of the older cohorts.

Figure 5

Birth cohorts have similar party gaps in each year

ANES Democratic minus Republican share saying abortion should always be allowed, by birth cohort (years of birth); party identifiers including leaners; 95% intervals in the data table

Show data

Data for Birth cohorts have similar party gaps in each year
Birth cohort	Year	Gap (points)	Lower	Upper	Respondents
1927-42	1980	14.1	3.5	24.8	315
1927-42	1982	n/a	n/a	n/a	n/a
1927-42	1984	0.0	−9.4	9.5	405
1927-42	1986	−2.0	−12.2	8.2	372
1927-42	1988	−3.4	−13.2	6.4	367
1927-42	1990	4.8	−5.7	15.3	323
1927-42	1992	9.2	−0.4	18.8	395
1927-42	1994	3.5	−7.8	14.7	318
1927-42	1996	12.3	1.2	23.5	305
1927-42	1998	n/a	n/a	n/a	n/a
1927-42	2000	19.3	7.6	31.1	274
1927-42	2004	n/a	n/a	n/a	n/a
1927-42	2008	n/a	n/a	n/a	n/a
1927-42	2012	29.9	20.0	39.7	576
1927-42	2016	n/a	n/a	n/a	n/a
1927-42	2020	49.0	38.2	59.8	456
1943-58	1980	11.4	2.3	20.6	465
1943-58	1982	0.0	−10.0	10.1	420
1943-58	1984	9.8	2.4	17.2	688
1943-58	1986	6.9	−0.5	14.3	696
1943-58	1988	9.1	1.6	16.6	630
1943-58	1990	5.1	−3.6	13.8	537
1943-58	1992	18.5	11.1	26.0	688
1943-58	1994	9.9	0.0	19.7	468
1943-58	1996	19.8	9.9	29.6	467
1943-58	1998	25.3	14.3	36.2	358
1943-58	2000	20.6	11.0	30.3	481
1943-58	2004	24.5	12.2	36.8	274
1943-58	2008	n/a	n/a	n/a	n/a
1943-58	2012	31.4	25.3	37.5	1,650
1943-58	2016	35.5	28.7	42.3	1,001
1943-58	2020	49.6	44.5	54.8	1,912
1943-58	2024	48.8	41.8	55.8	1,211
1959-74	1984	11.6	0.2	22.9	278
1959-74	1986	15.4	4.9	25.9	328
1959-74	1988	3.2	−7.0	13.3	365
1959-74	1990	4.8	−4.0	13.6	485
1959-74	1992	17.1	9.2	25.0	611
1959-74	1994	23.3	13.7	32.9	472
1959-74	1996	15.9	5.8	26.0	464
1959-74	1998	12.6	1.1	24.1	317
1959-74	2000	23.1	13.8	32.5	491
1959-74	2004	16.1	3.2	29.1	252
1959-74	2008	n/a	n/a	n/a	n/a
1959-74	2012	25.4	18.6	32.2	1,391
1959-74	2016	29.3	22.2	36.5	936
1959-74	2020	48.0	42.7	53.3	1,789
1959-74	2024	52.6	45.9	59.4	1,227
1975-90	2012	31.6	24.3	38.9	1,174
1975-90	2016	35.0	27.9	42.2	942
1975-90	2020	44.9	39.5	50.3	1,782
1975-90	2024	53.6	46.8	60.4	1,212
After 1990	2020	46.9	38.0	55.8	801
After 1990	2024	53.6	45.2	62.1	697

The cohort lines lie close together in every survey year. Of 52 same-year comparisons between two cohorts, 3 have an interval that excludes zero, all in 1986, 1988 and 1994 and all with the cohort born 1927 to 1942 lower. Each line follows one birth cohort, and the shaded band marks the change in ANES survey mode between 2016 and 2020.Note: Intervals are not drawn because five overlapping bands cannot be read; the 95% confidence intervals are in the data table.Note: Cohort-year cells are drawn only where each party has at least 100 respondents and an effective sample of at least 100, so the cohort born 1975 to 1990 begins in 2012 and the cohort born 1991 or later in 2020. A line breaks where a cohort's cell was left out.Source: ANES Cumulative Data File, 1980-2024; weighted (VCF0009z, which equals 1 for every respondent from 1980 to 1990)

Few Democrats now say abortion should never be permitted

The shares of people out of step with their party fell on both sides. The share of Democrats who said abortion should never be permitted fell from 12.4% (95% CI 10.1% to 14.6%) in 1980 to 1.4% (95% CI 0.7% to 2.0%) in 2024, and the share of Republicans who said it should always be allowed fell from 33.8% to 21.4% (Figure 6). In points the two falls were 11.0 (95% CI 8.6 to 13.4) and 12.4 (95% CI 7.6 to 17.2), and their intervals overlap. The Democratic share was 9.7% (95% CI 6.6% to 12.7%) in 2008, so most of its fall came after that year. Killian and Wilcox followed panel respondents from 1982 to 1997 and found that pro-life Democrats and pro-choice Republicans were more likely than other partisans to switch parties over that period [10].

Figure 6

Each party's out-of-step share fell, to near zero for Democrats

ANES share of Democrats saying abortion should never be permitted and of Republicans saying it should always be allowed; leaners included; 95% intervals

Show data

Data for Each party's out-of-step share fell, to near zero for Democrats
Series	Group and answer counted	Year	Share (%)	Lower	Upper	Respondents
Dem. never	Democrats saying abortion should never be permitted	1980	12.4	10.1	14.6	801
Dem. never	Democrats saying abortion should never be permitted	1982	14.8	12.3	17.4	741
Dem. never	Democrats saying abortion should never be permitted	1984	11.7	9.8	13.7	1,032
Dem. never	Democrats saying abortion should never be permitted	1986	14.2	12.1	16.3	1,061
Dem. never	Democrats saying abortion should never be permitted	1988	13.4	11.2	15.6	938
Dem. never	Democrats saying abortion should never be permitted	1990	11.6	9.6	13.6	993
Dem. never	Democrats saying abortion should never be permitted	1992	9.7	8.0	11.3	1,201
Dem. never	Democrats saying abortion should never be permitted	1994	14.0	11.3	16.6	810
Dem. never	Democrats saying abortion should never be permitted	1996	12.0	9.6	14.4	879
Dem. never	Democrats saying abortion should never be permitted	1998	9.7	7.2	12.2	647
Dem. never	Democrats saying abortion should never be permitted	2000	11.5	9.1	13.9	868
Dem. never	Democrats saying abortion should never be permitted	2004	10.9	7.8	13.9	498
Dem. never	Democrats saying abortion should never be permitted	2008	9.7	6.6	12.7	600
Dem. never	Democrats saying abortion should never be permitted	2012	6.5	5.3	7.7	3,081
Dem. never	Democrats saying abortion should never be permitted	2016	8.4	6.9	9.9	1,915
Dem. never	Democrats saying abortion should never be permitted	2020	4.6	3.7	5.6	3,713
Dem. never	Democrats saying abortion should never be permitted	2024	1.4	0.7	2.0	2,459
Rep. always	Republicans saying abortion should always be allowed	1980	33.8	29.7	37.9	512
Rep. always	Republicans saying abortion should always be allowed	1982	36.1	31.6	40.6	432
Rep. always	Republicans saying abortion should always be allowed	1984	32.8	29.7	36.0	856
Rep. always	Republicans saying abortion should always be allowed	1986	38.2	34.7	41.6	757
Rep. always	Republicans saying abortion should always be allowed	1988	33.5	30.2	36.7	813
Rep. always	Republicans saying abortion should always be allowed	1990	38.6	35.0	42.2	699
Rep. always	Republicans saying abortion should always be allowed	1992	38.7	35.5	41.9	896
Rep. always	Republicans saying abortion should always be allowed	1994	33.9	30.1	37.7	727
Rep. always	Republicans saying abortion should always be allowed	1996	33.7	29.6	37.9	642
Rep. always	Republicans saying abortion should always be allowed	1998	31.7	27.1	36.4	459
Rep. always	Republicans saying abortion should always be allowed	2000	31.0	27.2	34.9	664
Rep. always	Republicans saying abortion should always be allowed	2004	25.7	21.3	30.2	433
Rep. always	Republicans saying abortion should always be allowed	2008	29.3	23.3	35.3	293
Rep. always	Republicans saying abortion should always be allowed	2012	29.0	26.4	31.6	1,987
Rep. always	Republicans saying abortion should always be allowed	2016	27.5	24.9	30.0	1,715
Rep. always	Republicans saying abortion should always be allowed	2020	24.0	22.1	25.9	3,289
Rep. always	Republicans saying abortion should always be allowed	2024	21.4	18.9	23.8	2,207

Both lines end below where they began, by falls of similar size in points, but the Democratic line ends near zero and the Republican line does not. Bands are 95% confidence intervals and the shaded band marks the change in ANES survey mode between 2016 and 2020.Note: The Republican series is the same as the Republican line in the ANES view of the earlier figure on party shares. Shaded band: the ANES mode change from 2016 to 2020.Source: ANES Cumulative Data File, 1980-2024; weighted (VCF0009z, which equals 1 for every respondent from 1980 to 1990)

Abortion views came to track the presidential vote

Among ANES respondents who reported voting for the Democratic or Republican presidential candidate, the correlation between the four-category abortion answer and a Democratic vote was −0.05 (95% CI −0.11 to 0.02) in 1980 and +0.59 (95% CI 0.56 to 0.62) in 2024 (Figure 7). Zero means no straight-line relationship and one a perfect one. The correlation does not say what share of votes the abortion answer predicts, and party identification and ideology also move with the vote, so it does not separate the abortion answer from them. Chapter 1.3, Sort dates by attitude compares when abortion and other attitudes became partisan, and Chapter 9.5, Two moralities (planned) asks whether abortion, gay rights and gender roles form one dimension.

Figure 7

The correlation of abortion views with the vote rose to +0.59

Correlation between the four-category abortion answer (higher = more permissive) and a Democratic rather than Republican presidential vote, ANES presidential-election years, 95% intervals

Show data

Data for The correlation of abortion views with the vote rose to +0.59
Year	Correlation	Lower	Upper	Respondents
1980	−0.05	−0.11	0.02	920
1984	0.11	0.06	0.16	1,341
1988	0.11	0.06	0.17	1,182
1992	0.17	0.12	0.22	1,615
1996	0.23	0.16	0.29	1,096
2000	0.24	0.18	0.30	1,092
2004	0.26	0.18	0.33	796
2008	0.32	0.24	0.40	754
2012	0.37	0.33	0.40	4,168
2016	0.42	0.38	0.46	2,577
2020	0.51	0.49	0.54	5,977
2024	0.59	0.56	0.62	3,630

The line is about zero in 1980, is higher by 1984 and rises in each election after 1988. The band is a 95% confidence interval. The chart shows how closely the abortion answer and the vote move together and does not show that one produces the other.Note: Respondents who reported a vote for a candidate other than the two major-party candidates are left out. Shaded band: the ANES mode change from 2016 to 2020.Source: ANES Cumulative Data File, 1980-2024; weighted (VCF0009z, which equals 1 for every respondent from 1980 to 1990)

What these data cannot show

Each ANES and GSS study interviews a new sample, so the gap between the parties can widen because people change their view, change their party or are replaced by new adults. The surveys cannot say which, and the panel studies cited above [10] [8] address that question with other data; Chapter 1.4, Cohort change in partisan sorting describes ideology by birth cohort, which cannot separate switching from replacement. The estimates of change between studies treat each study as a new sample of people, although the 2020 and 2024 ANES studies re-interviewed some earlier respondents, which makes those intervals conservative. Timing is a second limit. The ANES file used here has no study between 2020 and 2024, none measured before and after the Dobbs decision within one design, and no separate weights by mode for 2020, so the chapter cannot say how much of the 2016 to 2020 change came from the change of mode. The 2022 GSS survey was mostly conducted after the decision but changed mode from 2021, and its samples within each mode are small. Its interviews before the decision were mostly on the web (86.9% of respondents who answered) and those on or after it were less so (39.9%), so a comparison of 2022 respondents before and after the decision mixes timing and mode.

Both questions measure the most permissive position. The ANES answers "permitted only in case of rape, incest or danger to the woman's life" and "permitted for other reasons once need is established" are not plotted, and neither question measures how much respondents care about the issue, which Cassese, Ondercin and Randall study [9]. The data begin in 1977 and 1980, so they cannot say when the divide started; O'Brian dates the link between abortion opposition and conservative views to 1970 [4]. The chapter splits no group by sex, and Chapter 3.5, The gender gap in attitudes and votes compares women and men on gender-role attitudes and the vote.

The group results combine years, so they describe the 1980s and the 2020s rather than any single year. The associations here do not show that abortion opinion changed party loyalty or that party loyalty changed abortion opinion.

Data sources

ANES Cumulative Data FileCumulative Data File 1948-2024, four-category abortion item (VCF0838) in the 17 studies from 1980 to 2024; mode weights for 2012, 2016 and 2024Years: 1948-2024 (33 study years; even years only after 2004; no 2006, 2010 or 2014 study) · Weights: weight (VCF0009z, which equals 1 for every respondent from 1980 to 1990); weight_ftf and weight_web for the mode split · License: Free download after accepting the ANES terms of use; cite the release.
ANES 2024 Time Series2024 Time Series, attendance items V241439 and V241440, used for the 2024 attendance groupsYears: 2024 (pre-election and post-election interviews) · Weights: V240107a, the pre-election weight that the cumulative file uses for 2024 · License: Free download after accepting the ANES terms of use; cite the release.
General Social SurveyCumulative data 1972-2024, abany asked in 29 surveys from 1977 to 2024; interview mode joined from the raw fileYears: 1972-2024 (35 study years) · Weights: weight_use (wtssnrps from 2004, wtssps before) · License: See the provider's terms of use.

All data sources

Methods

Sample. ANES Cumulative Data File respondents who answered the four-category abortion question (VCF0838: 1 abortion never permitted, 2 only for rape, incest or danger to the woman's life, 3 for other reasons once need is established, 4 always allowed as a matter of personal choice; the 'other' answer is missing): 45,017 respondents in 17 studies from 1980 to 2024, of whom 39,618 name or lean toward a party. GSS respondents who answered abany (a legal abortion should be possible if the woman wants it for any reason; asked from 1977 to 2024 and absent from the file before 1977): 41,398 respondents in 29 surveys, 34,037 of them Democratic or Republican identifiers. Party and estimand. Leaners count with their party (ANES pid_7pt 1 to 3 Democratic and 5 to 7 Republican; GSS partyid 0 to 2 and 4 to 6). Independents who lean toward neither party and members of other parties are left out of the party lines and the gap and are counted only in the all-respondent changes given in the text. In each survey year the estimand is the weighted share choosing the most permissive answer (ANES 'always allowed', GSS yes) among Democrats and among Republicans; the party gap is the Democratic share minus the Republican share, in percentage points, which is minus the Republican coefficient in a weighted linear probability model of the answer on a Republican indicator, the form used in earlier versions of this analysis. Weights. ANES weight VCF0009z, which equals 1 for every respondent from 1980 to 1990, so the 1980 endpoint and the 1980s groups are unweighted; GSS weight_use (wtssnrps from 2004, wtssps before). Mode. ANES mode-specific estimates use weight_ftf and weight_web, the weights for in-person and web respondents, which exist for 2008, 2012, 2016 and 2024 (2008 was a single-mode in-person study; 2020 has no separate weights by mode). The 2024 mode groups are the newly recruited in-person and web samples (the ANES calls them fresh samples), 1,017 in person and 1,955 on the web of the 5,021 item respondents; the panel and paper respondents are in neither. The GSS interview mode (in person, telephone, web or other) is joined from the raw file. Intervals. 95% intervals come from the Kish effective sample size with a normal approximation, which treats the weighted sample as a simple random sample and ignores the strata and clusters of the sampling design. For one estimate, the GSS 1977 gap, a design-based interval (Taylor linearisation with the GSS strata vstrat and primary sampling units vpsu) is also reported, because the two methods disagree about whether it lies below zero. Differences between two independent estimates (party gaps, changes over time, mode differences) combine the two standard errors in quadrature. Adjacent ANES studies are treated as independent although the 2020 and 2024 studies re-interviewed some earlier respondents, which makes those intervals conservative. Correlations use Fisher z on the Kish effective sample size. Suppression. A party cell in a yearly series needs at least 150 respondents and an effective sample of at least 100. A party cell in a group, mode or cohort estimate needs at least 100 respondents and an effective sample of at least 100, and a gap needs both cells to clear the floor. Periods and groups. The 1980s combine ANES 1980 to 1988 and GSS 1980 to 1989, and the 2020s combine ANES 2020 and 2024 and GSS 2021, 2022 and 2024, after rescaling each year's weights to average one so that years count in proportion to their respondents. Attendance has three categories (ANES weekly or almost weekly, once or twice a month, a few times a year or never; GSS attend 7 and 8, 4 to 6, and 0 to 3), with the 2024 ANES values rebuilt from the 2024 Time Series (V241439, V241440) because the cumulative file leaves out respondents who said they never attend. Education has three ANES categories and two GSS categories (a bachelor's degree or higher, or not). Birth cohort uses the ANES cohort categories and, for the GSS, survey year minus age. Vote. The weighted Pearson correlation between the four-category answer and a Democratic rather than Republican presidential vote, in presidential-election years. GSS 2022 interview date. The month and day are in dateintv, which has no year, so the interviews dated in January (5 of the respondents who answered the abortion question) are counted as before the decision; counting them as after it changes the weighted share by less than one point.

References

Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
Carmines, Edward G.; Stimson, James A. (1989). Issue Evolution: Race and the Transformation of American Politics. Princeton University Press.
Adams, Greg D. (1997). Abortion: Evidence of an Issue Evolution. American Journal of Political Science 41(3): 718-737.
O'Brian, Neil A. (2020). Before Reagan: The Development of Abortion's Partisan Divide. Perspectives on Politics 18(4): 1031-1047. doi:10.1017/S1537592719003840
American National Election Studies (2021). ANES 2020 Time Series Study Full Release: User Guide and Codebook. Stanford University and the University of Michigan. electionstudies.org/data-center/2020-time-series-study/
Burns, Nancy; Kinder, Donald R.; Rosenstone, Steven J.; Sapiro, Virginia; American National Election Studies (2001). ANES 2000 Time Series Study. Inter-university Consortium for Political and Social Research, ICPSR 3131. doi:10.3886/ICPSR03131.v1
Supreme Court of the United States (2022). Dobbs v. Jackson Women's Health Organization, 597 U.S. 215. Decided 24 June 2022. www.supremecourt.gov/opinions/21pdf/19-1392_6j37.pdf
Lockhart, Mackenzie; Gerber, Alan S.; Huber, Gregory A. (2024). Policy Shapes Partisan Identification: How Dobbs made pre-existing abortion policy preferences relevant to partisanship. Working paper, Yale University. huber.research.yale.edu/writings.html
Cassese, Erin C.; Ondercin, Heather L.; Randall, Jordan (2025). Abortion Attitudes and Polarization in the American Electorate. Elements in Gender and Politics, Cambridge University Press. doi:10.1017/9781009533119
Killian, Mitchell; Wilcox, Clyde (2008). Do Abortion Attitudes Lead to Party Switching?. Political Research Quarterly 61(4): 561-573. doi:10.1177/1065912907312981

How to cite

@misc{singh2026theabortionsort,
  author = {Singh, Vishal},
  title = {The abortion sort},
  year = {2026},
  url = {https://vishalsingh.org/politics/the-abortion-sort},
  note = {In The Public Mind. Published 30 September 2026.}
}

Reading paths

A tour of the book, one chapter per part