Chapter 3.5

The gender gap in attitudes and votes

In 2024 women voted more Democratic than men by 4.1 to 8.1 points, depending on the survey; the gap among voters under thirty cannot be told apart from the older gap; and party separates gender-role attitudes more than gender does in 3 of 4 surveys on the one item each asks; the World Values Survey is the exception, and it compares left-right groups because it has no party variable for the United States.

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

In 2024 women gave the Democratic presidential candidate more of the two-party vote than men did, by 8.1 (95% CI 3.6 to 12.6) points in the American National Election Studies (ANES) and by 4.1 (95% CI 2.6 to 5.7) in the Cooperative Election Study (CES). With the CES weight for respondents matched to voter-file records (lists of registered voters) the CES gap is 8.0 (95% CI 6.3 to 9.6), so the size of the gap depends on the weight as well as on the survey. Among voters under thirty the CES gap is 6.5 (95% CI 0.4 to 12.7), and among voters thirty and older it is 3.9 (95% CI 2.4 to 5.4). The difference is +2.6 (95% CI −3.8 to 8.9), and its confidence interval includes zero, so a larger gap among the young is not shown.

Party separates gender-role attitudes more than gender does in 3 of 4 surveys, each measured on one item chosen for this chapter. In the 2024 General Social Survey (GSS) the share giving the traditional answer on the roles of men and women is higher among Republicans than among Democrats by 25.8 (95% CI 19.6 to 31.9) and higher among men than among women by 4.4 (95% CI −1.1 to 9.9). The World Values Survey (WVS) is the exception, and it compares ideological groups instead of parties: the file has no party variable for the United States, so the left and right sides of its left-right scale stand in for Democrats and Republicans. The gap between the right side and the left side minus the gender gap is 0.3 (95% CI −2.9 to 3.5), an interval that includes zero.

Every gap in this chapter is a difference between two percentages, in points. The vote gap and the party-identification gap are women minus men, so a positive value means women were more Democratic. The attitude gender gap is men minus women and the attitude party gap is Republicans minus Democrats, so a positive value means men or Republicans were more likely to take the traditional or hostile position. The chapter measures the vote gap from 1952 and the attitude gaps as far back as each survey allows, and compares each gap among the young with the gap among older adults. The estimates come from cross-sections of respondents in the United States, so they compare groups within a survey and do not follow individuals. For the parties' coalitions as a whole, see Chapter 3.1, The two coalitions.

The ANES vote gap estimate has been positive since 1964

Figure 1 plots the ANES gap from 1952 to 2024. The estimate is negative in the three elections before 1964, and only the 1956 interval, −6.2 (95% CI −11.6 to −0.8), lies below zero. From 1964 the estimate is positive in 16 of 16 elections, and the interval lies above zero in 11 of them and includes zero in the other 5.The ANES vote estimates use the post-election interview and its weight where the file has one. Every interval in this chapter is a 95% interval on the Kish effective sample size [1], and it treats the weighted sample as a simple random sample, which leaves out the survey's strata and clusters and can move an interval either way. Nationscape is the exception, and its intervals are described in the attitudes section. The highest estimate is 14.9 (95% CI 8.4 to 21.3) in 1996. The change from 1996 to 2024 is −6.8 (95% CI −14.6 to 1.1), an interval that includes zero.

Figure 1

The ANES vote gap estimate has been positive since 1964 and peaked at 14.9 points in 1996

Women's minus men's Democratic share of the two-party presidential vote, points, among voters for a major-party candidate; ANES 1952 to 2024 and CES 2008 to 2024; 95% intervals

Gap in Democratic share of two-party vote (points)

Show data

Data for The ANES vote gap estimate has been positive since 1964 and peaked at 14.9 points in 1996
Election	Survey	Gap (points)	Lower 95%	Upper 95%	Voters
1952	ANES	−2.1	−7.7	3.5	1,181
1956	ANES	−6.2	−11.6	−0.8	1,266
1960	ANES	−5.4	−13.1	2.3	898
1964	ANES	3.7	−1.8	9.3	1,111
1968	ANES	4.6	−1.4	10.7	1,027
1972	ANES	6.6	1.9	11.3	1,587
1976	ANES	0.2	−5.3	5.8	1,322
1980	ANES	6.9	0.7	13.1	958
1984	ANES	7.6	2.4	12.8	1,376
1988	ANES	6.5	0.8	12.2	1,195
1992	ANES	10.1	5.3	14.9	1,658
1996	ANES	14.9	8.4	21.3	1,116
2000	ANES	8.9	2.4	15.4	1,120
2004	ANES	7.1	−0.4	14.6	811
2008	ANES	5.0	−1.2	11.3	1,539
2012	ANES	5.2	1.1	9.3	4,188
2016	ANES	6.0	1.3	10.7	2,579
2020	ANES	4.1	0.6	7.6	6,183
2024	ANES	8.1	3.6	12.6	3,943
2008	CES	6.4	4.8	7.9	23,585
2012	CES	5.3	3.7	6.8	37,789
2016	CES	8.2	6.9	9.5	40,891
2020	CES	8.6	7.4	9.9	43,872
2024	CES	4.1	2.6	5.7	41,757

The CES, a larger survey, sits above the ANES in 2020 by +4.5 and below it in 2024, when the CES gap falls and the ANES gap estimate rises, although the two surveys' intervals overlap at every election. The ANES estimate is below zero in 1952, 1956 and 1960 and above zero from 1964, but its interval includes zero at 5 of the 16 elections since 1964, and its highest estimate is 14.9 in 1996 against 8.1 in 2024. Bands are 95% intervals.Note: The ANES added web interviews in 2012. The gap is 0.2 among face-to-face respondents and 7.8 among web respondents in 2012, and 1.8 and 8.4 in 2024; each face-to-face interval includes zero. The file has no mode weights for 2020.Note: In 2024 the CES weight for respondents matched to voter-file records gives 8.0, and the standard weight on the same respondents gives 3.6.Source: ANES Cumulative Data File, 1952-2024, and CES Cumulative File, 2008-2024; weighted with each survey's survey weights (the ANES post-election weight where the file has one)

Work covering 1952 to 1996 traced the gap to a change in the partisanship of men [2], and a time-series study of 1979 to 2000 tied the growth of the partisan gender gap to a more conservative political climate, a weaker economy and a larger share of economically vulnerable single women [3]. The estimates here describe the gap and do not test either account.

The ANES and CES disagree about 2020 to 2024

The CES has more respondents than the ANES (41,757 voters against 3,943 in 2024), and its intervals are narrower at every election from 2008 to 2024.The CES is an online survey of opt-in panel members whose weights match them to the electorate, so it is not a random sample of adults. Its intervals, like the ANES ones, describe sampling noise and not who joined the panel. The two surveys' intervals overlap at every election from 2008 to 2024, yet the difference between the CES gap and the ANES gap has an interval above zero in 2020, when the CES gap exceeds the ANES gap by +4.5 (95% CI 0.8 to 8.2). From 2020 to 2024 the ANES gap changed by +4.0 (95% CI −1.7 to 9.7) and the CES gap by −4.5 (95% CI −6.4 to −2.5). The two changes differ by −8.5 (95% CI −14.5 to −2.5), so the surveys point in opposite directions for the latest change and the difference between them excludes zero.

The CES weight for respondents matched to voter-file records gives 8.0 (95% CI 6.3 to 9.6) in 2024, against 4.1 with the standard weight. On the same matched respondents the standard weight gives 3.6 (95% CI 2.0 to 5.2) and the matched weight gives 8.0, a difference of +4.3 (95% CI 2.9 to 5.5). The level of the 2024 CES gap therefore depends on the weight, and the age comparisons below report both.The matched weight is the CES weight for validated registered voters, the respondents whose registration the survey could match to voter-file records. The file holds it for 2018, 2020, 2022 and 2024. In 2020 it is the standard weight multiplied by the same number for every respondent, so the two weights give the same gap for the matched respondents, and this chapter compares them in 2024 only. The interval on the difference between the two weights resamples respondents, which keeps both weights on the same people.

The ANES added web interviews in 2012, so its series mixes survey modes after 2008. The face-to-face gap is smaller than the web gap in 2012 and in 2024, but each face-to-face interval includes zero and overlaps the web interval, so a difference of mode cannot be separated from sampling noise (the estimates are in the notes to Figure 1). The file has no mode weights for 2020, so the change from 2020 to 2024 cannot be split by mode.

Voters under thirty against older voters

Between 2020 and 2024 the under-thirty gap minus the gap among voters thirty and older moved by +8.6 (95% CI 1.0 to 16.1). The older gap changed by −5.2 (95% CI −7.2 to −3.2), and the under-thirty gap changed by +3.4 (95% CI −3.9 to 10.7), an interval that includes zero. Figure 2 shows the same pattern for four age groups: the gap fell in each of the three older groups, and the youngest group's change is not distinguishable from zero.

Figure 2

In the CES the vote gap fell after 2020 in every age group except the youngest

Women's minus men's Democratic share of the two-party presidential vote, points, by age group; CES 2020 and 2024; 95% intervals

Change in the gap from 2020 to 2024

2024 gap minus 2020 gap, points, by age group; standard weight

Each dot is the change for one age group with its 95% interval; an interval that stays left of zero is a fall.

Show data

Data for In the CES the vote gap fell after 2020 in every age group except the youngest, Change since 2020
Age group	Gap in 2020	Gap in 2024	Change (points)	Lower 95%	Upper 95%	Voters
Ages 18 to 29	3.2	6.5	3.4	−3.9	10.7	7,338
Ages 30 to 44	7.2	1.0	−6.2	−10.4	−2.0	16,863
Ages 45 to 64	9.1	3.5	−5.6	−8.6	−2.6	34,547
Ages 65 and over	11.3	7.2	−4.0	−7.4	−0.7	26,881

The gap fell in each of the three older age groups, by −5.2 among voters thirty and older, while the youngest group's change, +3.4, has an interval that includes zero.Source: CES Cumulative File, 2020-2024; weighted (survey weights; the weight for respondents matched to voter-file records in the matched-weight rows)

Figure 3 compares the two age groups election by election. In 3 of the 4 earlier elections (2008, 2012 and 2020) the under-thirty gap was smaller than the gap among older voters, with an interval below zero, and in 2016 the difference reached above zero by a small margin. In 2024 the difference is +2.6 (95% CI −3.8 to 8.9), which includes zero, so a larger youth gap is not shown. The interval of the change since 2020 excludes zero by a small margin, and the intervals assume a simple random sample, so intervals that allowed for the survey design could include zero. The evidence that the age pattern changed rests on this one survey.

Figure 3

The under-thirty vote gap was smaller than the older gap in 2008, 2012 and 2020 and not in 2024

Women's minus men's Democratic share of the two-party presidential vote, points, by age group and election; CES 2008 to 2024; 95% intervals

The two gaps, election by election

Gap by age group, points; the open dot is the gap at 30 and over and the filled dot the gap under 30

Age 30 and over
Under 30

Change is the under-thirty gap minus the gap at thirty and over, and the difference tab plots it with its interval.

Show data

Data for The under-thirty vote gap was smaller than the older gap in 2008, 2012 and 2020 and not in 2024, The two gaps
Election	Weight	Gap, 30 and over	Lower 95%	Upper 95%	Gap, under 30	Lower 95%	Upper 95%	Under 30 minus 30 and over	Lower 95%	Upper 95%
2008	Standard weight	8.3	6.7	9.9	−0.6	−4.6	3.4	−8.9	−13.2	−4.6
2012	Standard weight	6.2	4.7	7.8	0.3	−4.6	5.1	−6.0	−11.1	−0.9
2016	Standard weight	8.3	6.9	9.7	4.4	0.2	8.5	−4.0	−8.3	0.4
2020	Standard weight	9.1	7.8	10.4	3.2	−0.7	7.0	−6.0	−10.1	−1.9
2024	Standard weight	3.9	2.4	5.4	6.5	0.4	12.7	2.6	−3.8	8.9
2024, matched	Matched weight	7.3	5.8	8.9	11.6	3.8	19.3	4.2	−3.7	12.1

The under-thirty gap sits left of the older gap through 2020 and right of it in 2024, where the difference, +2.6, has an interval that includes zero. The difference moved by +8.6 between 2020 and 2024, an interval that excludes zero by a small margin. The matched weight, in the bottom row, is compared for 2024 only, because in 2020 the two weights give the same gap.Note: Each gap has its own interval, and the interval of the difference combines the two standard errors as if the two age groups were independent samples.Source: CES Cumulative File, 2008-2024; weighted (survey weights; the weight for respondents matched to voter-file records in the matched-weight rows)

The under-thirty estimate is imprecise. In 2024 the smaller of the women and men groups has an effective sample of 434 among voters under thirty, against 7,955 among voters thirty and older, because the CES holds few voters under thirty and their weights vary. The matched weight gives a larger under-thirty gap, 11.6 (95% CI 3.8 to 19.3), and an older-voter gap of 7.3 (95% CI 5.8 to 8.9). The difference of +4.2 (95% CI −3.7 to 12.1) again includes zero, so the level of the youth gap depends on the weight used.

The second tab of Figure 2 plots the 2024 gap for four age groups under both weights. With the standard weight the oldest group's gap exceeds that of ages thirty to forty-four by +6.3 (95% CI 2.2 to 10.3), and the youngest group's gap exceeds it by +5.5 (95% CI −1.5 to 12.6), an interval that includes zero. The youngest group's interval covers the estimates of both middle groups, and with the matched weight the estimates for the youngest and the oldest group each fall inside the other's interval.

Race-by-gender groups are mostly too small

Whether the youth gap sits mainly among voters of color needs enough respondents in each race-by-gender group (for example Hispanic women under thirty in 2024), and the CES has few. This chapter draws a group only when its women and its men each number at least 100 respondents and have an effective sample of at least 100, a floor that leaves out every group below it. Figure 4 shows the groups that clear the floor. In 2024 only white women and white men under thirty clear it, and the effective samples of the Black, Hispanic and Asian groups run from 23 to 82. Combining the 2020 and 2024 samples brings the Hispanic groups over the floor. The combined-sample gap is 11.2 (95% CI 1.0 to 21.4) among Hispanic voters under thirty and 2.3 (95% CI −2.5 to 7.1) among white voters under thirty, and the difference of +8.9 (95% CI −2.4 to 20.2) includes zero, so a larger gap among Hispanic voters is neither shown nor ruled out.

Pooling can also lower an effective sample, when the weights in one year vary a lot. Among Asian voters under thirty the smaller group has an effective sample of 101 in 2020, which clears the floor, and 64 in the combined sample, which does not. Black voters under thirty also stay below the floor in the combined sample. The estimates of the gap for the pooled Black and Asian groups are positive, and the sign is given only to show that leaving them out reflects the number of respondents and not the estimate. Chapter 3.4, Hispanic, Black and Asian voters (planned) looks at these groups across all ages.

Figure 4

Among voters under thirty, few race-by-gender groups are large enough to compare

Women's minus men's Democratic share of the two-party presidential vote, points, voters aged 18 to 29, by race; single elections and 2020 and 2024 pooled; 95% intervals

Show data

Data for Among voters under thirty, few race-by-gender groups are large enough to compare
Group and period	Gap (points)	Lower 95%	Upper 95%	Voters
All voters, pooled	5.4	1.5	9.2	7,338
White, 2016	1.5	−3.8	6.7	2,601
White, 2020	0.5	−4.4	5.5	2,861
White, 2024	3.5	−4.2	11.2	1,800
White, pooled	2.3	−2.5	7.1	4,661
Hispanic, 2016	10.2	0.8	19.6	645
Hispanic, 2020	8.1	−1.7	17.9	588
Hispanic, pooled	11.2	1.0	21.4	1,092
Asian, 2020	4.0	−5.6	13.6	344

Only white voters have a group large enough in every election shown, and the Hispanic estimates sit to the right of the white ones, but their intervals are wide and overlap the white intervals. The pooled gap is 11.2 among Hispanic voters and 2.3 among white voters, a difference of +8.9. A group is drawn only when its women and its men each have 100 respondents and an effective sample of 100.Note: Black groups miss the floor in every election, Asian groups clear it only in 2020, and Hispanic groups clear it in 2016 and 2020 but not in 2024, when the effective samples of the Black, Hispanic and Asian groups run from 23 to 82.Note: Pooling two elections can lower a group's effective sample when the weights in one year vary a lot: among Asian voters the smaller group has an effective sample of 101 in 2020 and 64 pooled, so the pooled Asian estimate is not drawn. Groups that miss the floor are left out, not shown with wider intervals.Source: CES Cumulative File, 2016-2024; weighted (survey weights)

Figure 5 plots the Democratic share of the two-party vote among voters under thirty in 2020 and 2024. The share changed by −6.3 (95% CI −10.9 to −1.6) for women and by −9.7 (95% CI −15.2 to −4.1) for men. The gap between the two changes is +3.4 (95% CI −3.9 to 10.7), so a faster move by men cannot be separated from a move shared by both. Among white voters under thirty the changes are −2.1 (95% CI −8.0 to 3.8) for women and −5.1 (95% CI −12.1 to 1.9) for men, and both intervals include zero.

Figure 5

Both young women and young men gave Democrats a smaller share in 2024 than in 2020

Democratic share of the two-party presidential vote, percent, voters aged 18 to 29; CES 2020 and 2024; 95% intervals

2020
2024

Show data

Data for Both young women and young men gave Democrats a smaller share in 2024 than in 2020
Group	2020 share	Lower 95%	Upper 95%	2024 share	Lower 95%	Upper 95%
Women	69.3	66.9	71.6	63.0	59.0	67.0
Men	66.1	63.1	69.2	56.5	51.8	61.1
White women	60.9	57.8	64.0	58.8	53.8	63.8
White men	60.3	56.6	64.1	55.2	49.4	61.1

The filled dot (2024) is to the left of the open dot (2020) in every row. The share fell for women from 69.3% to 63.0% and for men from 66.1% to 56.5%, changes of −6.3 and −9.7; the gap between those changes, +3.4, has an interval that includes zero. Among white voters under thirty the changes are −2.1 and −5.1, and both intervals include zero.Source: CES Cumulative File, 2020-2024; weighted (survey weights)

Party identification among adults under thirty

Party identification is observed in every even-numbered year from 2006, and it covers all young respondents who identify with a party, not only voters. Figure 6 plots the under-thirty gap in Democratic identification, women minus men. It was 9.5 (95% CI 5.5 to 13.6) in 2006, the highest estimate of those years, 8.5 (95% CI 4.9 to 12.1) in 2020 and 3.1 (95% CI −2.5 to 8.7) in 2024. The 2006 gap minus the 2020 gap is +1.1 (95% CI −4.3 to 6.5) and the change from 2020 to 2024 is −5.4 (95% CI −12.1 to 1.3), and both intervals include zero, so the highest estimate does not stand out from the later years and the fall to 2024 is not distinguishable from zero. The under-thirty vote gap changed by +3.4 (95% CI −3.9 to 10.7) over the same years, an interval that also includes zero, so the two series do not show party identification and the vote moving apart after 2020. The identification measure counts Democrats and Republicans only, so it rests on a different base from the vote.

Figure 6

The under-thirty party-identification gap was 9.5 in 2006 and 3.1 in 2024, and the 2024 interval includes zero

Women's minus men's Democratic share of respondents who identify as Democrats or Republicans, points, ages 18 to 29, even years 2006 to 2024; CES; 95% intervals

Gap in Democratic share of party identifiers (points)

Show data

Data for The under-thirty party-identification gap was 9.5 in 2006 and 3.1 in 2024, and the 2024 interval includes zero
Year	Gap (points)	Lower 95%	Upper 95%	Respondents
2006	9.5	5.5	13.6	2,726
2008	2.4	−1.4	6.3	3,737
2010	4.2	−0.5	9.0	3,271
2012	1.6	−2.9	6.0	4,737
2014	2.6	−1.4	6.5	5,507
2016	3.9	0.8	7.0	6,611
2018	7.0	3.5	10.6	7,098
2020	8.5	4.9	12.1	6,537
2022	7.1	3.2	11.1	5,285
2024	3.1	−2.5	8.7	5,162

The estimate is highest in 2006, and the bands are wide throughout. The 2006 gap minus the 2020 gap is +1.1 and the change from 2020 to 2024 is −5.4, and both intervals include zero. The gap is 9.5 in 2006, 8.5 in 2020 and 3.1 in 2024. Party identification here excludes independents and leaners and covers all respondents, not only voters.Source: CES Cumulative File, 2006-2024; weighted (survey weights)

Party separates attitudes more than gender does, except in the WVS

Four surveys ask about gender roles in ways that cannot be merged into one scale, so each is analyzed alone, and each contributes one item chosen for this chapter.

Survey and year	Item	Counted as the traditional or hostile position
GSS 2024	Whether it is better if the man is the achiever outside the home and the woman takes care of the home and family	Agree or strongly agree
ANES 2024	Rating of feminists on the feeling thermometer	A rating below fifty
WVS United States 2017	Whether men should have more right to a job than women when jobs are scarce	Agree
Nationscape 2019 to 2021	Whether women who complain about harassment often cause more problems than they solve	Agree

The GSS and WVS items concern roles, the ANES item concerns feeling toward feminists and the Nationscape item concerns hostility toward women who report harassment, so the four gaps measure related but different things. Each item is coded so that a higher share means the traditional or hostile position. The party gap compares Republicans with Democrats and leaves out independents, whereas the gender gap compares men with women among all respondents with an answer. The WVS file used here has no party variable for the United States, so the left and right sides of its left-right scale stand in for party. Figure 7 plots the party gap and the gender gap, the gender gap within each party, and the share for women and men within each party.Nationscape interviewed people weekly from July 2019 to early February 2021 through an opt-in market-research panel [4]. Like the CES, it is not a random sample of adults, and estimates from such panels can be off because of who joins the panel, which no interval measures [5]. Its intervals allow for correlation among respondents interviewed in the same week (clustered standard errors), which makes them wider than intervals that treat respondents as independent. Its party identification leaves out leaners, unlike the GSS and ANES, so its party gap is not comparable in size to theirs.

Figure 7

Party, or left-right group in the WVS, separates the traditional answer more than gender does in 3 of 4 surveys

Points and percent taking the traditional or hostile position on the one item each survey asks; most recent wave of each survey; 95% intervals

Party gap, gender gap and their difference in each survey

Party gap is Republicans minus Democrats (right side minus left side in the WVS) and gender gap is men minus women; points

Party gap
Gender gap
Party gap minus gender gap

GSS 2024: man the achiever, woman at home

ANES 2024: feminists rated below 50

WVS 2017: men first when jobs are scarce

Nationscape 2019 to 2021: harassment complaints

Gap in the traditional or hostile position (points)

The party gap is 25.8 in the GSS against a gender gap of 4.4, 31.6 against 8.8 in the ANES, 19.2 against 9.6 in Nationscape and 3.2 against 2.9 in the WVS.

Show data

Data for Party, or left-right group in the WVS, separates the traditional answer more than gender does in 3 of 4 surveys, Party gap and gender gap
Survey	Measure	Points	Lower 95%	Upper 95%	Respondents
GSS 2024: man the achiever, woman at home	Party gap	25.8	19.6	31.9	1,548
GSS 2024: man the achiever, woman at home	Gender gap	4.4	−1.1	9.9	2,153
GSS 2024: man the achiever, woman at home	Party gap minus gender gap	21.3	13.1	29.6	3,701
ANES 2024: feminists rated below 50	Party gap	31.6	28.3	34.9	5,103
ANES 2024: feminists rated below 50	Gender gap	8.8	5.5	12.1	5,475
ANES 2024: feminists rated below 50	Party gap minus gender gap	22.8	18.1	27.5	10,578
WVS 2017: men first when jobs are scarce	Party gap	3.2	0.8	5.5	2,534
WVS 2017: men first when jobs are scarce	Gender gap	2.9	0.7	5.0	2,591
WVS 2017: men first when jobs are scarce	Party gap minus gender gap	0.3	−2.9	3.5	5,125
Nationscape 2019 to 2021: harassment complaints	Party gap	19.2	18.4	20.0	317,051
Nationscape 2019 to 2021: harassment complaints	Gender gap	9.6	9.1	10.1	480,535
Nationscape 2019 to 2021: harassment complaints	Party gap minus gender gap	9.6	8.7	10.5	797,586

The party gap minus the gender gap is 21.3 in the GSS, 22.8 in the ANES and 9.6 in Nationscape, each with an interval above zero, and 0.3 in the WVS, with an interval that includes zero. The items differ by survey, so levels and gaps are not comparable across panels.Note: GSS: agree or strongly agree that it is much better if the man is the achiever outside the home and the woman takes care of the home and family. ANES: rating feminists below 50 on the 0 to 100 feeling thermometer. WVS: agree that men should have more right to a job than women when jobs are scarce. Nationscape: agree that women who complain about harassment often cause more problems than they solve.Note: Nationscape is an opt-in online panel, not a random sample of adults. Its intervals allow for correlation within the weekly wave, which widens them, and still describe sampling noise among panel members and not who joins the panel.Source: GSS 2024, ANES 2024, WVS United States 2017 and Nationscape 2019-2021; weighted (survey weights; Nationscape intervals cluster on the weekly wave)

The party gap is 25.8 in the GSS, 31.6 in the ANES and 19.2 in Nationscape, against gender gaps of 4.4, 8.8 and 9.6. The party gap minus the gender gap has an interval above zero in each of the three, and in the WVS it is 0.3 (95% CI −2.9 to 3.5), which includes zero. Restricting the gender gap to Democrats and Republicans, the same respondents as the party gap, leaves the count at 3 of 4 surveys.

Within a party, men give the traditional answer more often than women in every group by the estimate. The interval of the men-minus-women difference excludes zero in both parties in the ANES and Nationscape and on the left side of the WVS scale, and includes zero in both parties in the GSS and on the right side of the WVS scale, so the GSS is the survey where the difference is not distinguishable from zero in either party.

Figure 8 follows the two gaps over time. In the GSS the party gap in the traditional answer was −2.0 (95% CI −8.0 to 4.0) in 1977, an interval that includes zero, and 11.3 (95% CI 4.9 to 17.8) in 2018. The estimated gender gap did not exceed 12.1 (95% CI 5.6 to 18.6) in any of the 24 survey years. The GSS moved partly online after 2018, so its later estimates mix modes. In 2024 the party gap is 19.1 (95% CI 8.5 to 29.6) among face-to-face respondents and 32.3 (95% CI 24.4 to 40.2) among web respondents. The face-to-face estimate is above the 2018 estimate, but the intervals overlap, so growth after 2018 and the change of mode cannot be told apart.

Figure 8

The GSS party gap went from −2.0 in 1977 to 11.3 in 2018, and the estimated gender gap never exceeded 12.1

Gap in the share taking the traditional or hostile position, points: Republicans minus Democrats (party gap) and men minus women (gender gap); GSS 1977 to 2024 and ANES 1988 to 2024; 95% intervals

Gap in the traditional or hostile position (points)

Show data

Data for The GSS party gap went from −2.0 in 1977 to 11.3 in 2018, and the estimated gender gap never exceeded 12.1
Survey	Gap	Year	Gap (points)	Lower 95%	Upper 95%	Respondents
GSS	Party gap	1977	−2.0	−8.0	4.0	1,319
GSS	Gender gap	1977	4.8	−0.6	10.2	1,503
GSS	Party gap	1985	1.6	−5.1	8.3	1,340
GSS	Gender gap	1985	7.3	0.9	13.6	1,502
GSS	Party gap	1986	1.4	−4.8	7.6	1,252
GSS	Gender gap	1986	1.9	−3.8	7.7	1,444
GSS	Party gap	1988	5.7	−1.9	13.2	835
GSS	Gender gap	1988	6.4	−0.6	13.3	964
GSS	Party gap	1989	0.9	−6.7	8.4	846
GSS	Gender gap	1989	4.7	−2.3	11.7	977
GSS	Party gap	1990	0.5	−7.1	8.2	792
GSS	Gender gap	1990	7.8	0.5	15.0	901
GSS	Party gap	1991	0.5	−6.9	7.9	866
GSS	Gender gap	1991	1.2	−6.0	8.4	996
GSS	Party gap	1993	5.4	−1.6	12.4	907
GSS	Gender gap	1993	5.8	−0.8	12.3	1,054
GSS	Party gap	1994	5.8	0.5	11.2	1,622
GSS	Gender gap	1994	5.4	0.5	10.4	1,916
GSS	Party gap	1996	8.6	3.7	13.5	1,951
GSS	Gender gap	1996	6.2	1.7	10.6	2,353
GSS	Party gap	1998	6.9	1.3	12.5	1,472
GSS	Gender gap	1998	2.9	−2.2	8.0	1,818
GSS	Party gap	2000	9.8	3.9	15.7	1,393
GSS	Gender gap	2000	3.2	−2.0	8.4	1,804
GSS	Party gap	2002	11.2	2.6	19.8	696
GSS	Gender gap	2002	6.2	−1.5	13.8	898
GSS	Party gap	2004	7.5	−0.8	15.8	739
GSS	Gender gap	2004	2.1	−5.5	9.7	894
GSS	Party gap	2006	10.9	5.1	16.6	1,553
GSS	Gender gap	2006	4.7	−0.4	9.8	1,968
GSS	Party gap	2008	14.0	7.0	21.0	1,080
GSS	Gender gap	2008	5.2	−1.1	11.4	1,308
GSS	Party gap	2010	11.3	4.2	18.4	1,132
GSS	Gender gap	2010	7.7	1.4	13.9	1,418
GSS	Party gap	2012	18.5	11.0	26.0	1,004
GSS	Gender gap	2012	12.1	5.6	18.6	1,285
GSS	Party gap	2014	12.4	6.3	18.5	1,276
GSS	Gender gap	2014	2.8	−2.5	8.2	1,655
GSS	Party gap	2016	15.4	9.9	21.0	1,501
GSS	Gender gap	2016	0.2	−4.8	5.2	1,872
GSS	Party gap	2018	11.3	4.9	17.8	1,186
GSS	Gender gap	2018	5.0	−0.6	10.7	1,545
GSS	Party gap	2021	20.8	16.3	25.3	2,054
GSS	Gender gap	2021	5.6	1.6	9.5	2,647
GSS	Party gap	2022	17.1	10.6	23.6	1,710
GSS	Gender gap	2022	9.9	4.3	15.4	2,335
GSS	Party gap	2024	25.8	19.6	31.9	1,548
GSS	Gender gap	2024	4.4	−1.1	9.9	2,153
ANES	Party gap	1988	20.6	16.0	25.2	1,448
ANES	Gender gap	1988	1.0	−3.4	5.5	1,627
ANES	Party gap	1992	27.0	22.9	31.1	1,834
ANES	Gender gap	1992	0.9	−3.0	4.8	2,085
ANES	Party gap	2000	18.4	13.1	23.7	1,255
ANES	Gender gap	2000	1.7	−3.3	6.7	1,427
ANES	Party gap	2002	20.0	13.9	26.0	1,166
ANES	Gender gap	2002	4.7	−1.4	10.7	1,276
ANES	Party gap	2004	25.1	19.2	31.0	917
ANES	Gender gap	2004	8.2	2.5	13.9	1,028
ANES	Party gap	2008	21.4	16.2	26.5	1,723
ANES	Gender gap	2008	8.2	3.5	12.8	1,972
ANES	Party gap	2012	28.5	24.9	32.0	4,657
ANES	Gender gap	2012	9.4	6.1	12.8	5,376
ANES	Party gap	2016	29.8	26.2	33.5	3,098
ANES	Gender gap	2016	6.0	2.4	9.6	3,537
ANES	Party gap	2020	34.2	31.5	36.9	6,453
ANES	Gender gap	2020	7.1	4.3	9.8	7,279
ANES	Party gap	2024	31.6	28.3	34.9	5,103
ANES	Gender gap	2024	8.8	5.5	12.1	5,475

The party gap estimate reaches 25.8 in 2024, in the shaded years when the GSS is partly or largely online: it is 19.1 among face-to-face respondents and 32.3 among web respondents, against 11.3 in 2018. The face-to-face interval overlaps the 2018 interval, so growth after 2018 cannot be separated from the change of mode. The gender gap is 4.8 in 1977 and 4.4 in 2024, its highest estimate is 12.1 in 2012, and its interval lies above zero in 8 of 24 survey years.Note: GSS item: agree or strongly agree that it is much better if the man is the achiever outside the home and the woman takes care of the home and family. ANES item: rating feminists below 50 on the 0 to 100 feeling thermometer.Note: The GSS 2021, 2022 and 2024 samples mix interview modes; the ANES has web interviews from 2012. Lines are not drawn across the years in which the item was not asked (GSS 1978 to 1984 and 2019 to 2020, ANES 1996), so the GSS 1977 estimate stands alone; its interval is in the tooltip and the data table.Source: GSS 1977-2024, weighted (survey weights); ANES Cumulative Data File, 1988-2024, weighted (survey weights)

The ANES has a different history. Its party gap in rating feminists as cold was already 20.6 (95% CI 16.0 to 25.2) in 1988 and is 31.6 (95% CI 28.3 to 34.9) in 2024. Its gender gap went from 1.0 in 1988 to 8.8 in 2024, a change of +7.8 (95% CI 2.2 to 13.3). The gender gap estimates run from 0.9 to 1.7 through 2000 and from 6.0 to 9.4 from 2004 on, and the change from 2000 to 2004 alone is +6.5 (95% CI −1.1 to 14.1), an interval that includes zero, so the estimates show a higher range from 2004 without dating the rise to a single election.

Figure 9 compares the gender gap in these attitudes among the young with the gap among older adults in each survey. In Nationscape the gap among adults under thirty exceeds the gap among older adults by +6.5 (95% CI 5.3 to 7.7), an interval that excludes zero. In the ANES the difference is −9.3 (95% CI −19.3 to 0.7) and in the WVS +1.5 (95% CI −4.4 to 7.3), and both intervals include zero, with the young estimate the smaller in the ANES and the larger in the WVS. The GSS groups under thirty fall below the floor, and the gap among older adults there is 2.4 (95% CI −3.5 to 8.3). Only Nationscape's difference excludes zero, and it is the opt-in panel.

Figure 9

Only Nationscape shows a larger attitude gender gap among adults under thirty

Men's minus women's share taking the traditional or hostile position, points, by age group, and the difference between the two age groups; most recent wave of each survey; 95% intervals

All adults
30 and over
Under 30
Under 30 minus 30 and over

GSS 2024: man the achiever, woman at home

ANES 2024: feminists rated below 50

WVS 2017: men first when jobs are scarce

Nationscape 2019 to 2021: harassment complaints

Men minus women, points

Show data

Data for Only Nationscape shows a larger attitude gender gap among adults under thirty
Survey	Age group	Gap (points)	Lower 95%	Upper 95%	Respondents
GSS 2024: man the achiever, woman at home	All adults	4.4	−1.1	9.9	2,153
GSS 2024: man the achiever, woman at home	30 and over	2.4	−3.5	8.3	1,798
ANES 2024: feminists rated below 50	All adults	8.8	5.5	12.1	5,475
ANES 2024: feminists rated below 50	Under 30	1.4	−8.0	10.7	545
ANES 2024: feminists rated below 50	30 and over	10.7	7.1	14.2	4,672
ANES 2024: feminists rated below 50	Under 30 minus 30 and over	−9.3	−19.3	0.7	5,217
WVS 2017: men first when jobs are scarce	All adults	2.9	0.7	5.0	2,591
WVS 2017: men first when jobs are scarce	Under 30	4.3	−1.1	9.6	630
WVS 2017: men first when jobs are scarce	30 and over	2.8	0.5	5.2	1,961
WVS 2017: men first when jobs are scarce	Under 30 minus 30 and over	1.5	−4.4	7.3	2,591
Nationscape 2019 to 2021: harassment complaints	All adults	9.6	9.1	10.1	480,535
Nationscape 2019 to 2021: harassment complaints	Under 30	14.8	13.7	16.0	106,828
Nationscape 2019 to 2021: harassment complaints	30 and over	8.3	7.8	8.8	373,707
Nationscape 2019 to 2021: harassment complaints	Under 30 minus 30 and over	6.5	5.3	7.7	480,535

In Nationscape the gap among adults under thirty, 14.8, exceeds the gap among older adults, 8.3, by +6.5, an interval that excludes zero. In the ANES the young estimate is smaller (1.4 against 10.7, a difference of −9.3) and in the WVS it is larger (4.3 against 2.8, a difference of +1.5), and both differences have intervals that include zero. The GSS groups under thirty fall below the floor of 100 for the effective sample, so that panel has no age comparison.Note: Nationscape is an opt-in online panel; its intervals allow for correlation within the weekly wave and describe sampling noise among panel members only.Source: GSS 2024, ANES 2024, WVS United States 2017 and Nationscape 2019-2021; weighted (survey weights; Nationscape intervals cluster on the weekly wave)

What these data cannot show

The estimates describe groups in each year and cannot show that any person changed. A rise in the under-thirty gap between two elections compares different people, so age cannot be separated from generation here, which Chapter 3.6, Age and generation takes up. The vote gaps use self-reported votes, and the two surveys and the two CES weights differ on the 2024 level and on the change since 2020, so the data cannot say which estimate is closest to the electorate. Both the CES and Nationscape are opt-in online panels, so their intervals leave out the error that comes from who joined the panel. The CES intervals also treat the weighted sample as a simple random sample, so allowing for the survey design could widen or narrow them. Results whose interval ends close to zero, such as the under-thirty gap in 2024 and the change in the age difference since 2020, could include zero if the intervals were wider. Most race-by-gender groups are too small to estimate, and a gap that is not distinguishable from zero in a small group is not evidence that the gap is zero.

The attitude items differ in wording, scale and window, and their levels cannot be compared across surveys or merged into one index. Each survey contributes one item chosen for this chapter, so the count of surveys in which the party gap exceeds the gender gap describes these items, and a different item could change it. Within a survey, the comparisons of party with gender and of young with old are informative, and the party gap minus the gender gap ignores how the two gaps move together, because both are computed on the same respondents.

All the data are from the United States. A widely read Financial Times analysis described a gender divide among the young in several countries [6], and a comparative study of gender attitudes examined many societies [7], but nothing here tests either claim outside the United States. The files record two genders, so nonbinary respondents cannot be examined. Finally, the data hold no measure of why the gaps arise. They cannot separate differences in policy attitudes from differences in the salience of issues, in candidates or in party strategy, and Chapter 9.5, Two moralities (planned) places gender-role attitudes among other cultural attitudes.

Data sources

ANES Cumulative Data FileCumulative Data File, presidential vote 1952-2024 and feminist thermometer 1988-2024Years: 1948-2024 (33 study years; even years only after 2004; no 2006, 2010 or 2014 study) · Weights: post-election weight where available, else VCF0009z · License: Free download after accepting the ANES terms of use; cite the release.
Cooperative Election StudyCumulative File, presidential vote 2008-2024 and party identification 2006-2024Years: 2006-2024 (every year; self-reported turnout is available in the 10 even years) · Weights: weight_cumulative; vvweight for the matched-weight rows (2024) · License: See the provider's terms of use.
General Social Survey1977-2024, gender-role item fefam, with interview mode from the raw fileYears: 1972-2024 (35 study years) · Weights: weight_use · License: See the provider's terms of use.
World Values Survey and EVSUnited States, wave 7 (2017)Years: 1981-2023 (7 unified waves) · Weights: weight (S017) · License: Free download with registration; users must cite the release.
NationscapeWeekly waves, July 2019 to February 2021Years: 2019-07-18 to 2021-04-15 (85 weekly waves) · Weights: weight; errors clustered on the weekly wave · License: Free after registration and acceptance of the Voter Study Group terms; also on Harvard Dataverse.

All data sources

Methods

Samples: the vote gap uses respondents who reported voting for the Democratic or Republican presidential candidate. ANES Cumulative Data File, presidential elections 1952 to 2024 (post-election weight where the file has one, otherwise the full-sample weight VCF0009z); CES Cumulative File, 2008 to 2024 (weight_cumulative). The CES is an online survey of opt-in panel members whose weights match them to the electorate, so it is not a random sample of adults. The matched-weight rows use vvweight, the weight for respondents matched to voter-file records, which the file holds for 2018, 2020, 2022 and 2024; in 2020 it is a constant multiple of weight_cumulative, so the matched weight is compared with the standard weight in 2024 only. Party identification in the CES uses even years 2006 to 2024 and counts Democrats and Republicans only, leaners excluded, among all respondents and not only voters. Attitude items, one per survey and chosen by the author: GSS agree or strongly agree that it is much better if the man is the achiever outside the home and the woman takes care of the home and family (1977 to 2024, weight_use, party and sex from the subset); ANES feminist feeling thermometer below 50 (1988 to 2024, VCF0009z); WVS United States wave 7 (2017, weight S017), agreement that men should have more right to a job when jobs are scarce, with the left side (1 to 5) and right side (6 to 10) of the left-right scale standing in for party because the cleaned file has no party variable; Nationscape (weekly waves from July 2019 to early February 2021, plus four later interviews in the file, weight), agreement that women who complain about harassment often cause more problems than they solve, party identification without leaners. Party is Democrat or Republican with leaners included in the GSS and ANES; the party gap therefore leaves out independents (and, in Nationscape, leaners), while the gender gap uses all respondents with an answer, and a check restricts the gender gap to Democrats and Republicans. Gender is the two-category variable in each file; ANES respondents coded Other (11) are dropped. Age groups are 18 to 29, 30 to 44, 45 to 64 and 65 and over. Estimands: a gap is a difference of two weighted shares, women minus men for the vote and for party identification, men minus women and Republicans minus Democrats for the attitude items; a change is a difference of two gaps from different samples. Intervals: 95% normal intervals on the Kish effective sample size, model based, treating the weighted sample as a simple random sample and ignoring clustering; differences of independent estimates combine the two standard errors in quadrature (the party gap minus the gender gap ignores how the two gaps move together, because they are computed on overlapping respondents). The difference between the matched-weight and standard-weight gaps for the same 2024 respondents is a percentile bootstrap over respondents (400 replicates, seed 12345), because the two gaps share people. Nationscape estimates are weighted regressions with errors clustered on the 85 weekly waves, wider than the Kish interval, and still describe sampling noise among opt-in panel members. Suppression: a race, age or party cell needs 100 respondents and an effective sample of 100 (subgroup-by-year floor); timeline cells built on one item need 150 respondents and an effective sample of 100; a gap is published only when both cells clear the floor. Pooling two elections can lower a cell's effective sample when the weights in one year vary a lot (among Asian voters under thirty the smaller cell has 101 in 2020 and 64 pooled), so a pooled group can miss the floor that its 2020 cells clear. Modes: the ANES mode split uses weight_ftf and weight_web (2012, 2016 and 2024; the file has no mode weights for 2020); the GSS mode split uses the interview mode from the raw file (2022 and 2024 mix in-person and web interviews, 2021 is mostly web).

References

Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
Kaufmann, Karen M.; Petrocik, John R. (1999). The Changing Politics of American Men: Understanding the Sources of the Gender Gap. American Journal of Political Science 43(3): 864-887. doi:10.2307/2991838
Box-Steffensmeier, Janet M.; De Boef, Suzanna; Lin, Tse-min (2004). The Dynamics of the Partisan Gender Gap. American Political Science Review 98(3): 515-528. doi:10.1017/S0003055404001315
Tausanovitch, Chris; Vavreck, Lynn (2021). Democracy Fund + UCLA Nationscape Project (version 20211215). Harvard Dataverse. doi:10.7910/DVN/CQFP3Z
Mercer, Andrew W.; Kreuter, Frauke; Keeter, Scott; Stuart, Elizabeth A. (2017). Theory and Practice in Nonprobability Surveys: Parallels between Causal Inference and Survey Inference. Public Opinion Quarterly 81(S1): 250-271. academic.oup.com/poq/article/81/S1/250/3749176
Burn-Murdoch, John (2024). A new global gender divide is emerging. Financial Times, 26 January 2024.
Inglehart, Ronald; Norris, Pippa (2003). Rising Tide: Gender Equality and Cultural Change Around the World. Cambridge University Press.

How to cite

@misc{singh2026thegendergap,
  author = {Singh, Vishal},
  title = {The gender gap in attitudes and votes},
  year = {2026},
  url = {https://vishalsingh.org/politics/the-gender-gap},
  note = {In The Public Mind. Published 30 September 2026.}
}

The ANES vote gap estimate has been positive since 1964

The ANES and CES disagree about 2020 to 2024

Voters under thirty against older voters

Race-by-gender groups are mostly too small

The Democratic share fell among young women and young men

Party identification among adults under thirty

Party separates attitudes more than gender does, except in the WVS

What these data cannot show

Data sources

Methods

References

How to cite