Chapter 3.2

The diploma divide

In the ANES college graduates went from voting less Democratic than other voters to voting more Democratic, a gap of +23.3 points in 2024 against +12.3 in the CES, while their differences from non-graduates on cultural, racial and immigration issues were already present in the 1990s.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 3 October 2026

DraftFigures and numbers may change as the analysis is rebuilt.what is still being checked

The reasons the ANES and the CES give different education gaps in 2008, 2012 and 2024 are still being checked.

In 2024 the gap between voters with a four-year degree and other voters in the Democratic share of the two-party presidential vote was +12.3 (95% CI 10.8 to 13.8) points in the Cooperative Election Study (CES) and +23.3 (95% CI 19.0 to 27.7) points in the American National Election Studies (ANES), with graduates the more Democratic group in both. The ANES gap runs from negative to positive. It was −0.8 (95% CI −5.0 to 3.5) points in 2012, an interval that includes zero, and in 8 of the 14 elections from 1956 to 2012 that have an estimate the interval lies below zero, so graduates were the less Democratic group. The CES gap is above zero in every presidential election since 2008, +3.7 (95% CI 2.2 to 5.2) points in 2008. The ANES sample of 2024 voters is more Democratic than the certified count, and against the CES the excess is concentrated among graduates (Figure 3), so the ANES level of the 2024 gap needs the CES beside it.

The differences on issues are older than the difference in the vote. On a cultural index (abortion, gay rights and moral traditionalism), graduates' answers sat +0.30 (95% CI 0.25 to 0.36) standard deviations nearer the liberal end than other respondents' in the ANES studies of 1992 to 2000 and +0.36 (95% CI 0.32 to 0.39) in 2016 to 2024, pooling in-person and web interviews. The racial index gives +0.21 (95% CI 0.16 to 0.27) and +0.31 (95% CI 0.27 to 0.35), and the immigration item +0.36 (95% CI 0.30 to 0.41) and +0.37 (95% CI 0.33 to 0.41). The economic index is the exception, with a gap of −0.32 (95% CI −0.37 to −0.28) in the first period and +0.02 (95% CI −0.02 to 0.06) in the second, an interval that includes zero. Holding party identification fixed shrinks the cultural gap in 2016 to 2024 to +0.26 (95% CI 0.22 to 0.29) and leaves positive gaps on the racial (+0.18 (95% CI 0.15 to 0.21)) and immigration (+0.30 (95% CI 0.26 to 0.34)) items, and it turns the economic gap negative (−0.10 (95% CI −0.13 to −0.07)). These are descriptions of how groups differ. They cannot say what a degree does to a person's politics.

This chapter compares people with a four-year degree or more, called graduates, with everyone else. A graduate holds a college or advanced degree in the ANES, a four-year or postgraduate degree in the CES (24.7% of respondents in 2008 and 34.8% in 2024) and a bachelor's or graduate degree in the General Social Survey (GSS), and an associate degree counts as no four-year degree in all three. Education here is a credential, the highest degree a person holds. Income, which Chapter 3.3, The income ladder takes up, is a separate measure of household resources, and occupation is not measured in this chapter; a later section checks whether income rank accounts for the vote gap. Graduates are a growing part of the samples, 6.4% (95% CI 5.2% to 7.5%) of ANES respondents in 1952 and 34.4% (95% CI 32.6% to 36.2%) in 2024, so graduates and non-graduates are different slices of the population in different years. Party identification includes leaners, and white means non-Hispanic white. A gap is the value for graduates minus the value for non-graduates. Every estimate carries a 95% confidence interval that treats the weighted sample as a random sample; the methods note gives the details.

In the ANES graduates went from the less Democratic group to the more Democratic group

Figure 1 plots the Democratic share of the two-party vote by degree. The share among non-graduates was above the share among graduates in 12 of the 14 elections from 1956 to 2012 that have an estimate, the gap's interval includes zero in 6 of the 14, and in none does it lie above zero. From 2012 to 2016 the gap rose by +12.7 (95% CI 6.5 to 18.9) points, to +11.9 (95% CI 7.4 to 16.4), and it was +14.9 (95% CI 11.6 to 18.1) in 2020 and +23.3 (95% CI 19.0 to 27.7) in 2024. From 2012 to 2024 the graduate share changed by +16.0 (95% CI 11.5 to 20.5) points and the non-graduate share by −8.2 (95% CI −12.2 to −4.1) points.

Figure 1

Graduates voted less Democratic than other voters in most elections through 2012 and more Democratic from 2016

Democratic share of the two-party presidential vote by degree, ANES, 1956 to 2024; voters who reported a major-party vote, all voters (the button shows white non-Hispanic voters); 95% intervals from the Kish effective sample size

Democratic share of the two-party vote (%)

Show data

Data for Graduates voted less Democratic than other voters in most elections through 2012 and more Democratic from 2016
Population	Group	Election	Democratic share (%)	95% CI low	95% CI high	n voters
All voters	Four-year degree or more	1956	31.5	23.2	39.7	124
All voters	No four-year degree	1956	41.2	38.3	44.1	1,136
All voters	Four-year degree or more	1964	50.0	41.7	58.3	142
All voters	No four-year degree	1964	69.9	67.1	72.8	965
All voters	Four-year degree or more	1968	37.5	30.0	45.0	160
All voters	No four-year degree	1968	48.1	44.6	51.7	750
All voters	Four-year degree or more	1972	39.3	33.5	45.1	275
All voters	No four-year degree	1972	34.9	32.4	37.5	1,311
All voters	Four-year degree or more	1976	37.3	31.2	43.4	250
All voters	No four-year degree	1976	54.4	51.4	57.5	1,069
All voters	Four-year degree or more	1980	35.0	28.0	41.9	183
All voters	No four-year degree	1980	45.9	42.2	49.6	691
All voters	Four-year degree or more	1984	41.3	35.8	46.8	305
All voters	No four-year degree	1984	41.9	38.9	44.8	1,063
All voters	Four-year degree or more	1988	42.2	36.8	47.5	325
All voters	No four-year degree	1988	48.8	45.5	52.2	850
All voters	Four-year degree or more	1992	51.8	46.9	56.7	401
All voters	No four-year degree	1992	61.9	58.7	65.0	922
All voters	Four-year degree or more	1996	49.5	44.0	55.0	360
All voters	No four-year degree	1996	62.2	58.2	66.3	672
All voters	Four-year degree or more	2000	50.4	45.3	55.4	429
All voters	No four-year degree	2000	53.0	48.9	57.1	689
All voters	Four-year degree or more	2004	50.3	44.2	56.5	289
All voters	No four-year degree	2004	49.7	45.0	54.4	522
All voters	Four-year degree or more	2008	50.8	45.0	56.6	400
All voters	No four-year degree	2008	56.8	53.2	60.4	1,133
All voters	Four-year degree or more	2012	53.0	49.6	56.4	1,471
All voters	No four-year degree	2012	53.8	51.2	56.4	2,673
All voters	Four-year degree or more	2016	60.4	57.2	63.6	1,175
All voters	No four-year degree	2016	48.5	45.3	51.6	1,416
All voters	Four-year degree or more	2020	64.9	62.8	67.0	2,993
All voters	No four-year degree	2020	50.0	47.6	52.5	3,036
All voters	Four-year degree or more	2024	69.0	66.1	71.9	2,013
All voters	No four-year degree	2024	45.7	42.5	48.8	1,905
White non-Hispanic voters	Four-year degree or more	1956	30.6	22.3	38.8	121
White non-Hispanic voters	No four-year degree	1956	40.1	37.1	43.0	1,086
White non-Hispanic voters	Four-year degree or more	1964	46.6	38.1	55.1	133
White non-Hispanic voters	No four-year degree	1964	67.0	63.9	70.2	877
White non-Hispanic voters	Four-year degree or more	1968	33.8	26.2	41.3	151
White non-Hispanic voters	No four-year degree	1968	42.6	38.9	46.4	664
White non-Hispanic voters	Four-year degree or more	1972	34.5	28.6	40.4	252
White non-Hispanic voters	No four-year degree	1972	29.4	26.8	32.0	1,168
White non-Hispanic voters	Four-year degree or more	1976	36.6	30.3	42.9	237
White non-Hispanic voters	No four-year degree	1976	48.9	45.6	52.1	940
White non-Hispanic voters	Four-year degree or more	1980	31.5	24.4	38.6	165
White non-Hispanic voters	No four-year degree	1980	37.5	33.5	41.4	574
White non-Hispanic voters	Four-year degree or more	1984	39.6	33.8	45.5	270
White non-Hispanic voters	No four-year degree	1984	34.4	31.3	37.5	875
White non-Hispanic voters	Four-year degree or more	1988	38.6	32.9	44.3	280
White non-Hispanic voters	No four-year degree	1988	39.1	35.4	42.8	665
White non-Hispanic voters	Four-year degree or more	1992	48.0	42.7	53.4	335
White non-Hispanic voters	No four-year degree	1992	54.8	51.1	58.6	693
White non-Hispanic voters	Four-year degree or more	1996	46.6	40.8	52.5	318
White non-Hispanic voters	No four-year degree	1996	53.0	48.2	57.7	514
White non-Hispanic voters	Four-year degree or more	2000	48.1	42.5	53.6	357
White non-Hispanic voters	No four-year degree	2000	44.6	39.9	49.2	533
White non-Hispanic voters	Four-year degree or more	2004	45.7	38.8	52.5	233
White non-Hispanic voters	No four-year degree	2004	39.6	34.1	45.1	368
White non-Hispanic voters	Four-year degree or more	2008	43.3	36.5	50.0	258
White non-Hispanic voters	No four-year degree	2008	43.8	39.1	48.6	523
White non-Hispanic voters	Four-year degree or more	2012	47.0	43.1	50.8	1,039
White non-Hispanic voters	No four-year degree	2012	40.3	37.2	43.4	1,489
White non-Hispanic voters	Four-year degree or more	2016	55.2	51.6	58.8	937
White non-Hispanic voters	No four-year degree	2016	34.1	30.6	37.6	1,010
White non-Hispanic voters	Four-year degree or more	2020	60.2	57.8	62.6	2,339
White non-Hispanic voters	No four-year degree	2020	35.6	32.9	38.3	2,230
White non-Hispanic voters	Four-year degree or more	2024	63.9	60.6	67.3	1,591
White non-Hispanic voters	No four-year degree	2024	35.8	32.3	39.4	1,390

The graduate line starts below the line for other voters and ends above it, and the two cross between 2012 and 2016. Each line is the weighted Democratic share of voters who reported a major-party vote, with a 95% interval; in 2024 the share was 69.0% among graduates and 45.7% among other voters, and the 2024 gap is +23.3 points here against +12.3 in the CES. A graduate holds a college or advanced degree.Note: A cell needs at least 100 respondents and a Kish effective sample size of at least 100; cells below that are left out. This leaves out 1952 and 1960, and the line joins the neighboring elections.Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Source: ANES Cumulative Data File, 1952-2024; weighted (VCF0009z; the file gives every respondent the same weight in most studies before 1992)

The white non-Hispanic view shows the same shape with larger gaps. Among white voters the gap was +6.6 (95% CI 1.7 to 11.6) points in 2012, the first election with an interval above zero, and +28.1 (95% CI 23.2 to 33.0) in 2024; Chapter 3.4, Hispanic, Black and Asian voters (planned) follows the vote of other racial and ethnic groups. The vote is the respondent's report of a major-party choice (32,098 ANES respondents in the elections plotted). Two elections, 1952 and 1960, are left out because the graduate cell has fewer than 100 respondents or a Kish effective sample size [1] under 100.

Party identification moved later than the vote. Among ANES respondents who identify with a party, the Democratic share was lower among graduates in 25 of the 27 studies from 1952 to 2012 that have an estimate with an interval below zero, and the interval includes zero in the rest. The gap was −5.8 (95% CI −9.7 to −1.9) points in 2012 and +1.6 (95% CI −2.2 to 5.5) in 2016, an interval that includes zero, and then +8.9 (95% CI 5.9 to 11.9) in 2020 and +17.5 (95% CI 13.6 to 21.4) in 2024. Among white respondents the gap was −0.4 (95% CI −5.0 to 4.2) in 2012 and +9.5 (95% CI 5.1 to 13.9) in 2016 (Figure 2).

Figure 2

Graduates became more Democratic than other respondents in vote and party identification, and in the ANES the party-identification gap followed the vote gap

Democratic share among graduates minus the share among other respondents, points, ANES 1952 to 2024 and CES 2008 to 2024; presidential vote (voters with a major-party vote) and party identification (identifiers, leaners included); all respondents (the button shows white non-Hispanic respondents)

Graduate minus other respondents (points)

Show data

Data for Graduates became more Democratic than other respondents in vote and party identification, and in the ANES the party-identification gap followed the vote gap
Population	Measure	Survey	Study year	Gap (points)	95% CI low	95% CI high	n respondents
All respondents	Presidential vote	ANES	1956	−9.7	−18.4	−1.1	1,260
All respondents	Presidential vote	ANES	1964	−19.9	−28.7	−11.2	1,107
All respondents	Presidential vote	ANES	1968	−10.6	−19.0	−2.3	910
All respondents	Presidential vote	ANES	1972	+4.3	−2.0	10.7	1,586
All respondents	Presidential vote	ANES	1976	−17.1	−24.0	−10.3	1,319
All respondents	Presidential vote	ANES	1980	−10.9	−18.8	−3.0	874
All respondents	Presidential vote	ANES	1984	−0.6	−6.8	5.7	1,368
All respondents	Presidential vote	ANES	1988	−6.7	−13.0	−0.3	1,175
All respondents	Presidential vote	ANES	1992	−10.0	−15.8	−4.2	1,323
All respondents	Presidential vote	ANES	1996	−12.8	−19.6	−5.9	1,032
All respondents	Presidential vote	ANES	2000	−2.6	−9.1	3.9	1,118
All respondents	Presidential vote	ANES	2004	+0.7	−7.1	8.4	811
All respondents	Presidential vote	ANES	2008	−6.0	−12.9	0.9	1,533
All respondents	Presidential vote	ANES	2012	−0.8	−5.0	3.5	4,144
All respondents	Presidential vote	ANES	2016	+11.9	7.4	16.4	2,591
All respondents	Presidential vote	ANES	2020	+14.9	11.6	18.1	6,029
All respondents	Presidential vote	ANES	2024	+23.3	19.0	27.7	3,918
All respondents	Presidential vote	CES	2008	+3.7	2.2	5.2	23,585
All respondents	Presidential vote	CES	2012	+5.6	4.1	7.1	37,789
All respondents	Presidential vote	CES	2016	+12.3	11.1	13.5	40,891
All respondents	Presidential vote	CES	2020	+15.3	14.1	16.5	43,872
All respondents	Presidential vote	CES	2024	+12.3	10.8	13.8	42,008
All respondents	Party identification	ANES	1952	−18.5	−28.1	−8.9	1,602
All respondents	Party identification	ANES	1956	−12.0	−20.9	−3.0	1,528
All respondents	Party identification	ANES	1958	−22.2	−31.9	−12.5	1,266
All respondents	Party identification	ANES	1962	−8.8	−17.6	0.0	1,132
All respondents	Party identification	ANES	1964	−18.1	−26.3	−9.9	1,407
All respondents	Party identification	ANES	1966	−21.2	−30.8	−11.6	1,101
All respondents	Party identification	ANES	1968	−22.6	−30.2	−14.9	1,364
All respondents	Party identification	ANES	1970	−16.2	−24.4	−8.0	1,290
All respondents	Party identification	ANES	1972	−14.6	−20.4	−8.7	2,293
All respondents	Party identification	ANES	1974	−12.7	−21.6	−3.9	1,290
All respondents	Party identification	ANES	1976	−17.9	−24.1	−11.7	1,869
All respondents	Party identification	ANES	1978	−17.3	−23.2	−11.4	1,893
All respondents	Party identification	ANES	1980	−12.5	−19.5	−5.6	1,365
All respondents	Party identification	ANES	1982	−11.5	−18.6	−4.5	1,216
All respondents	Party identification	ANES	1984	−8.3	−14.2	−2.5	1,938
All respondents	Party identification	ANES	1986	−8.3	−13.9	−2.8	1,845
All respondents	Party identification	ANES	1988	−11.3	−17.0	−5.5	1,751
All respondents	Party identification	ANES	1990	−9.9	−15.7	−4.1	1,702
All respondents	Party identification	ANES	1992	−11.8	−16.7	−6.8	2,099
All respondents	Party identification	ANES	1994	−17.8	−23.8	−11.8	1,555
All respondents	Party identification	ANES	1996	−11.1	−17.0	−5.1	1,546
All respondents	Party identification	ANES	1998	−8.0	−14.9	−1.1	1,119
All respondents	Party identification	ANES	2000	−13.3	−19.0	−7.6	1,566
All respondents	Party identification	ANES	2002	−12.4	−18.8	−6.1	1,363
All respondents	Party identification	ANES	2004	−2.4	−9.3	4.5	1,074
All respondents	Party identification	ANES	2008	−9.3	−15.6	−3.0	2,006
All respondents	Party identification	ANES	2012	−5.8	−9.7	−1.9	5,034
All respondents	Party identification	ANES	2016	+1.6	−2.2	5.5	3,643
All respondents	Party identification	ANES	2020	+8.9	5.9	11.9	7,181
All respondents	Party identification	ANES	2024	+17.5	13.6	21.4	5,055
All respondents	Party identification	CES	2008	−1.6	−3.0	−0.2	28,600
All respondents	Party identification	CES	2009	+0.5	−1.8	2.8	11,436
All respondents	Party identification	CES	2010	+1.9	0.6	3.2	47,861
All respondents	Party identification	CES	2011	+0.2	−2.1	2.4	16,696
All respondents	Party identification	CES	2012	+0.8	−0.7	2.2	45,847
All respondents	Party identification	CES	2013	+0.7	−1.8	3.1	12,834
All respondents	Party identification	CES	2014	+1.5	0.2	2.8	44,776
All respondents	Party identification	CES	2015	+3.6	1.0	6.2	11,559
All respondents	Party identification	CES	2016	+4.5	3.4	5.6	51,986
All respondents	Party identification	CES	2017	+4.8	2.5	7.2	14,360
All respondents	Party identification	CES	2018	+8.2	7.0	9.4	49,327
All respondents	Party identification	CES	2019	+9.1	7.0	11.2	14,767
All respondents	Party identification	CES	2020	+12.3	11.1	13.5	49,769
All respondents	Party identification	CES	2021	+12.1	10.4	13.9	20,584
All respondents	Party identification	CES	2022	+11.3	10.0	12.6	49,520
All respondents	Party identification	CES	2023	+8.9	7.0	10.7	19,715
All respondents	Party identification	CES	2024	+9.4	8.0	10.9	50,847
White non-Hispanic respondents	Presidential vote	ANES	1956	−9.5	−18.2	−0.7	1,207
White non-Hispanic respondents	Presidential vote	ANES	1964	−20.4	−29.5	−11.4	1,010
White non-Hispanic respondents	Presidential vote	ANES	1968	−8.8	−17.3	−0.4	815
White non-Hispanic respondents	Presidential vote	ANES	1972	+5.2	−1.3	11.6	1,420
White non-Hispanic respondents	Presidential vote	ANES	1976	−12.3	−19.3	−5.2	1,177
White non-Hispanic respondents	Presidential vote	ANES	1980	−5.9	−14.1	2.2	739
White non-Hispanic respondents	Presidential vote	ANES	1984	+5.2	−1.4	11.9	1,145
White non-Hispanic respondents	Presidential vote	ANES	1988	−0.5	−7.3	6.3	945
White non-Hispanic respondents	Presidential vote	ANES	1992	−6.8	−13.3	−0.3	1,028
White non-Hispanic respondents	Presidential vote	ANES	1996	−6.3	−13.8	1.2	832
White non-Hispanic respondents	Presidential vote	ANES	2000	+3.5	−3.7	10.7	890
White non-Hispanic respondents	Presidential vote	ANES	2004	+6.1	−2.7	14.8	601
White non-Hispanic respondents	Presidential vote	ANES	2008	−0.6	−8.8	7.7	781
White non-Hispanic respondents	Presidential vote	ANES	2012	+6.6	1.7	11.6	2,528
White non-Hispanic respondents	Presidential vote	ANES	2016	+21.1	16.1	26.1	1,947
White non-Hispanic respondents	Presidential vote	ANES	2020	+24.6	21.0	28.2	4,569
White non-Hispanic respondents	Presidential vote	ANES	2024	+28.1	23.2	33.0	2,981
White non-Hispanic respondents	Presidential vote	CES	2008	+9.2	7.6	10.8	19,889
White non-Hispanic respondents	Presidential vote	CES	2012	+10.2	8.6	11.9	30,356
White non-Hispanic respondents	Presidential vote	CES	2016	+18.6	17.2	20.0	30,727
White non-Hispanic respondents	Presidential vote	CES	2020	+22.7	21.5	24.0	32,857
White non-Hispanic respondents	Presidential vote	CES	2024	+16.1	14.5	17.7	31,857
White non-Hispanic respondents	Party identification	ANES	1952	−19.4	−29.2	−9.5	1,464
White non-Hispanic respondents	Party identification	ANES	1956	−12.7	−21.8	−3.7	1,415
White non-Hispanic respondents	Party identification	ANES	1958	−22.6	−32.6	−12.6	1,164
White non-Hispanic respondents	Party identification	ANES	1962	−7.1	−16.2	2.0	1,036
White non-Hispanic respondents	Party identification	ANES	1964	−16.4	−24.8	−7.9	1,258
White non-Hispanic respondents	Party identification	ANES	1966	−18.6	−28.3	−8.9	978
White non-Hispanic respondents	Party identification	ANES	1968	−21.5	−29.4	−13.6	1,206
White non-Hispanic respondents	Party identification	ANES	1970	−13.8	−22.2	−5.4	1,148
White non-Hispanic respondents	Party identification	ANES	1972	−14.6	−20.7	−8.4	2,018
White non-Hispanic respondents	Party identification	ANES	1974	−9.9	−19.1	−0.8	1,138
White non-Hispanic respondents	Party identification	ANES	1976	−14.7	−21.1	−8.2	1,607
White non-Hispanic respondents	Party identification	ANES	1978	−15.1	−21.3	−8.9	1,613
White non-Hispanic respondents	Party identification	ANES	1980	−11.8	−19.2	−4.3	1,136
White non-Hispanic respondents	Party identification	ANES	1982	−7.8	−15.3	−0.4	1,031
White non-Hispanic respondents	Party identification	ANES	1984	−5.4	−11.7	0.9	1,568
White non-Hispanic respondents	Party identification	ANES	1986	−4.1	−10.2	1.9	1,431
White non-Hispanic respondents	Party identification	ANES	1988	−6.3	−12.5	−0.1	1,321
White non-Hispanic respondents	Party identification	ANES	1990	−7.6	−14.0	−1.2	1,288
White non-Hispanic respondents	Party identification	ANES	1992	−8.3	−13.8	−2.8	1,606
White non-Hispanic respondents	Party identification	ANES	1994	−12.4	−19.0	−5.8	1,200
White non-Hispanic respondents	Party identification	ANES	1996	−6.3	−12.9	0.3	1,201
White non-Hispanic respondents	Party identification	ANES	1998	−9.3	−17.1	−1.5	834
White non-Hispanic respondents	Party identification	ANES	2000	−9.8	−16.2	−3.3	1,169
White non-Hispanic respondents	Party identification	ANES	2002	−8.3	−15.3	−1.2	1,065
White non-Hispanic respondents	Party identification	ANES	2004	+0.7	−7.3	8.7	764
White non-Hispanic respondents	Party identification	ANES	2008	−4.7	−12.3	2.8	990
White non-Hispanic respondents	Party identification	ANES	2012	−0.4	−5.0	4.2	2,981
White non-Hispanic respondents	Party identification	ANES	2016	+9.5	5.1	13.9	2,636
White non-Hispanic respondents	Party identification	ANES	2020	+18.1	14.8	21.5	5,290
White non-Hispanic respondents	Party identification	ANES	2024	+23.5	19.1	27.9	3,682
White non-Hispanic respondents	Party identification	CES	2008	+3.8	2.3	5.3	23,608
White non-Hispanic respondents	Party identification	CES	2009	+6.8	4.2	9.4	8,366
White non-Hispanic respondents	Party identification	CES	2010	+8.0	6.5	9.5	35,480
White non-Hispanic respondents	Party identification	CES	2011	+5.6	3.1	8.2	12,698
White non-Hispanic respondents	Party identification	CES	2012	+6.8	5.1	8.4	33,444
White non-Hispanic respondents	Party identification	CES	2013	+7.0	4.2	9.8	9,468
White non-Hispanic respondents	Party identification	CES	2014	+8.1	6.6	9.6	32,357
White non-Hispanic respondents	Party identification	CES	2015	+11.1	8.2	14.1	8,233
White non-Hispanic respondents	Party identification	CES	2016	+11.6	10.3	12.9	36,202
White non-Hispanic respondents	Party identification	CES	2017	+12.2	9.5	15.0	9,458
White non-Hispanic respondents	Party identification	CES	2018	+16.5	15.2	17.8	36,733
White non-Hispanic respondents	Party identification	CES	2019	+16.6	14.3	18.9	10,414
White non-Hispanic respondents	Party identification	CES	2020	+20.8	19.5	22.1	35,942
White non-Hispanic respondents	Party identification	CES	2021	+21.3	19.3	23.4	13,665
White non-Hispanic respondents	Party identification	CES	2022	+18.2	16.8	19.6	34,004
White non-Hispanic respondents	Party identification	CES	2023	+15.6	13.5	17.8	12,444
White non-Hispanic respondents	Party identification	CES	2024	+14.8	13.2	16.5	34,772

In the ANES the vote gap is above zero from 2016 and the party-identification gap from 2020, and the CES lines sit above the ANES lines before 2016 and below them in 2024. The ANES vote gap was +23.3 in 2024 against +12.3 in the CES, and the party-identification gap +17.5 against +9.4. Each point is a gap between two weighted Democratic shares, with a 95% interval.Note: A cell needs at least 100 respondents and a Kish effective sample size of at least 100; cells below that are left out. The line joins the neighboring studies where a study is left out.Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Note: The CES is an opt-in online panel, so its intervals are model based and ignore how respondents came to be in the panel.Source: ANES Cumulative Data File, 1952-2024; weighted (VCF0009z; the file gives every respondent the same weight in most studies before 1992); Cooperative Election Study cumulative file, 2008-2024; weighted (weight_cumulative)

The CES and the ANES agree in 2016 and 2020 and differ in the other years

The CES has 188,145 respondents with a major-party vote in 2008 to 2024, against 0 in the ANES over the same years. Its vote gap among all voters was +3.7 (95% CI 2.2 to 5.2) points in 2008, +12.3 (95% CI 11.1 to 13.5) in 2016, +15.3 (95% CI 14.1 to 16.5) in 2020 and +12.3 (95% CI 10.8 to 13.8) in 2024. Among white voters it was +9.2 (95% CI 7.6 to 10.8) in 2008 and +16.1 (95% CI 14.5 to 17.7) in 2024. Its party-identification gap among all respondents was −1.6 (95% CI −3.0 to −0.2) in 2008, with an interval below zero, and +12.3 (95% CI 11.1 to 13.5) in 2020 and +9.4 (95% CI 8.0 to 10.9) in 2024.

The two surveys agree where their intervals overlap, in 2016 and 2020, and disagree before and after. The ANES gap minus the CES gap was −9.7 (95% CI −16.7 to −2.6) points in 2008, −6.4 (95% CI −10.9 to −1.9) in 2012 and +11.0 (95% CI 6.5 to 15.6) in 2024. They also disagree about the latest movement: the ANES gap changed by +8.5 (95% CI 3.1 to 13.9) points from 2020 to 2024, and the CES gap by −3.0 (95% CI −4.8 to −1.1) points. Both surveys record the respondent's own report of a vote. The ANES is a probability sample, and the CES is an opt-in online panel whose weights match it to the population.

Certified returns give a benchmark for the level of the Democratic share, though not for the gap (Figure 3). In 2024 the ANES voters were 55.3% (95% CI 53.0% to 57.5%) Democratic, the CES voters 49.0% (95% CI 48.2% to 49.7%) and the certified two-party count 49.2%. In 2020 the shares were 56.0% (95% CI 54.3% to 57.7%), 52.5% (95% CI 51.9% to 53.1%) and 52.3%. The ANES share minus the count was +6.0 (95% CI 3.8 to 8.3) points in 2024 and +3.7 (95% CI 2.0 to 5.4) in 2020, while the CES share is within 1.1 points of the count in every election from 2008 to 2024. Against the CES, the ANES Democratic share in 2024 is +12.9 (95% CI 9.8 to 16.0) points higher among graduates (69.0% against 56.1%) and +1.8 (95% CI −1.5 to 5.2) among non-graduates (45.7% against 43.8%). The ANES sample of 2024 voters is therefore more Democratic than the count, and the excess among graduates accounts for most of the ANES-CES difference of +11.0 points in the gap. The count has no breakdown by education, so it cannot show how far either survey's shares for graduates and non-graduates are from the truth.

Figure 3

The ANES voters were more Democratic than the certified count in 1992, 1996, 2020 and 2024, and the CES voters were within 1.1 points of it in every election

Democratic share of the two-party presidential vote among survey voters minus the certified national share, points, ANES 1976 to 2024 and CES 2008 to 2024; all voters who reported a major-party vote, whether or not education is known (button: the graduate gap under each survey weight); 95% intervals

ANES
CES

Show data

Data for The ANES voters were more Democratic than the certified count in 1992, 1996, 2020 and 2024, and the CES voters were within 1.1 points of it in every election
View	Row	Survey or weight	Difference or gap (points)	95% CI low	95% CI high	n respondents
certified	1976	ANES	0.0	−2.7	2.8	1,322
certified	1980	ANES	−1.2	−4.5	2.1	877
certified	1984	ANES	+0.9	−1.7	3.5	1,376
certified	1988	ANES	+0.9	−1.9	3.8	1,195
certified	1992	ANES	+4.9	2.3	7.5	1,357
certified	1996	ANES	+3.6	0.3	6.9	1,034
certified	2000	ANES	+1.8	−1.5	5.0	1,120
certified	2004	ANES	+1.1	−2.6	4.9	811
certified	2008	ANES	+1.1	−2.0	4.2	1,539
certified	2012	ANES	+1.6	−0.4	3.6	4,188
certified	2016	ANES	+1.8	−0.5	4.1	2,609
certified	2020	ANES	+3.7	2.0	5.4	6,119
certified	2024	ANES	+6.0	3.8	8.3	3,948
certified	2008	CES	+0.1	−0.7	0.9	23,585
certified	2012	CES	+1.1	0.3	1.8	37,789
certified	2016	CES	−0.2	−0.8	0.5	40,891
certified	2020	CES	+0.3	−0.4	0.9	43,872
certified	2024	CES	−0.3	−1.0	0.5	42,008
weights	ANES 2016	Main weight	+11.9	7.4	16.4	2,591
weights	ANES 2016	Post-election weight	+12.0	7.5	16.5	2,571
weights	ANES 2020	Main weight	+14.9	11.6	18.1	6,029
weights	ANES 2020	Post-election weight	+15.4	12.1	18.8	5,987
weights	ANES 2024	Main weight	+23.3	19.0	27.7	3,918
weights	ANES 2024	Post-election weight	+22.9	18.7	27.2	3,903
weights	CES 2012	Main weight	+5.6	4.1	7.1	37,789
weights	CES 2012	Post-election weight	+4.7	3.2	6.1	37,789
weights	CES 2016	Main weight	+12.3	11.1	13.5	40,891
weights	CES 2016	Post-election weight	+10.1	8.8	11.4	40,891
weights	CES 2020	Main weight	+15.3	14.1	16.5	43,872
weights	CES 2020	Post-election weight	+12.9	11.6	14.2	43,871
weights	CES 2020	Validated-vote weight	+16.2	14.9	17.5	35,746
weights	CES 2024	Main weight	+12.3	10.8	13.8	42,008
weights	CES 2024	Validated-vote weight	+12.4	10.9	14.0	32,948

The ANES dots lie right of zero in 1992, 1996, 2020 and 2024 and the CES dots sit close to the zero line. In 2024 the ANES voters were 55.3% Democratic, the CES voters 49.0% and the certified count 49.2%. Against the CES, the ANES Democratic share in 2024 is +12.9 points higher among graduates (69.0% against 56.1%) and +1.8 among non-graduates (45.7% against 43.8%). The certified count has no breakdown by education, so this comparison shows the level and not the gap. The weights view gives the graduate gap under each weight defined for the year; within a survey the weights move the gap by at most 3.4 points, against an ANES-CES difference of +11.0 points in 2024.Note: The certified share is the Democratic share of the Democratic and Republican candidates' votes in the state returns, summed over states; it is a full count and has no interval. The survey interval is shifted by the count.Note: ANES 2024 uses the pre-election weight VCF0009z; the weights view shows the post-election weight. The CES weights are the cumulative-file weight, the post-election weight (2012, 2016 and 2020) and the validated-vote weight (2020 and 2024).Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Note: The CES is an opt-in online panel, so its intervals are model based and ignore how respondents came to be in the panel.Source: ANES Cumulative Data File, 1976-2024, and CES cumulative file, 2008-2024; weighted (VCF0009z, weight_postelection, weight_cumulative, weight_post and vvweight, as labelled); MIT Election Data and Science Lab state presidential returns, 1976-2024; full counts

The choice of weight matters less than the choice of survey. The ANES 2024 gap uses the full-sample weight, which for 2024 is the pre-election weight, and is +23.3 (95% CI 19.0 to 27.7) points; with the post-election weight it is +22.9 (95% CI 18.7 to 27.2). In the CES the gap is +12.3 in 2016 with the cumulative-file weight and +10.1 (95% CI 8.8 to 11.4) with the post-election weight, and in 2020 +15.3, +12.9 (95% CI 11.6 to 14.2) and, with the validated-vote weight, +16.2 (95% CI 14.9 to 17.5). Within a survey the weights move the gap by at most 3.4 points, less than the +11.0 points between the surveys in 2024.

Interview mode is another candidate, and the ANES can speak to it only in part. It interviewed respondents in person and on the web in 2012, 2016 and 2024 and gives each mode its own weight (Figure 4), but in 2024 the mode groups cover 2,045 of the 3,918 voters. Panel and paper cases are in neither group, and their gap is +18.5 (95% CI 12.6 to 24.5) points. In both modes the 2012 interval includes zero and the 2016 and 2024 intervals lie above it. The 2024 vote gap was +30.5 (95% CI 20.8 to 40.1) points in person and +20.2 (95% CI 13.5 to 26.9) on the web. The ANES web gap minus the CES gap is +7.9 (95% CI 1.0 to 14.7) points, although the CES is also online, so mode does not account for the difference between the surveys. The web minus in-person difference within the ANES in 2024 is −10.3 (95% CI −22.0 to 1.5) points, an interval that includes zero and reaches past the ANES-CES difference, so the data do not rule out a mode effect. The 2020 ANES was collected without in-person interviews and has no mode weights.

Figure 4

The vote gap rose after 2012 in both interview modes, and the in-person estimates are the larger ones in 2016 and 2024

ANES graduate gap in the vote and in party identification by interview mode, 2012, 2016 and 2024, points (button: issue gaps by mode, standard deviations); 95% intervals

In person
Web
Panel and paper

Show data

Data for The vote gap rose after 2012 in both interview modes, and the in-person estimates are the larger ones in 2016 and 2024
View	Row	Mode	Gap	95% CI low	95% CI high	n respondents
vote	Vote, 2012	In person	−4.15	−11.99	3.68	1,327
vote	Vote, 2012	Web	+1.16	−3.84	6.16	2,817
vote	Vote, 2016	In person	+15.94	7.16	24.72	745
vote	Vote, 2016	Web	+10.36	5.16	15.56	1,846
vote	Vote, 2024	In person	+30.45	20.77	40.14	688
vote	Vote, 2024	Web	+20.19	13.49	26.88	1,357
vote	Vote, 2024	Panel and paper	+18.55	12.57	24.53	1,873
vote	Party identification, 2012	In person	−7.86	−14.84	−0.88	1,800
vote	Party identification, 2012	Web	−4.61	−9.34	0.12	3,234
vote	Party identification, 2016	In person	+5.19	−2.35	12.74	1,040
vote	Party identification, 2016	Web	+0.38	−4.04	4.79	2,603
vote	Party identification, 2024	In person	+24.64	16.14	33.14	963
vote	Party identification, 2024	Web	+15.20	9.41	20.99	1,840
vote	Party identification, 2024	Panel and paper	+12.71	7.13	18.29	2,252
issue	Economic index, ANES 2016 and 2024	In person	+0.13	0.01	0.24	2,196
issue	Economic index, ANES 2016 and 2024	Web	−0.08	−0.15	−0.02	5,094
issue	Cultural index, ANES 2016 and 2024	In person	+0.46	0.35	0.57	2,195
issue	Cultural index, ANES 2016 and 2024	Web	+0.34	0.27	0.41	5,077
issue	Racial index, ANES 2016 and 2024	In person	+0.47	0.35	0.59	1,936
issue	Racial index, ANES 2016 and 2024	Web	+0.22	0.15	0.29	4,291
issue	Immigration item, ANES 2016 and 2024	In person	+0.43	0.31	0.54	1,934
issue	Immigration item, ANES 2016 and 2024	Web	+0.31	0.24	0.38	4,292
issue	Economic index, GSS 2022 and 2024	In person	−0.03	−0.18	0.13	1,724
issue	Economic index, GSS 2022 and 2024	Web	−0.04	−0.16	0.08	2,278
issue	Cultural index, GSS 2022 and 2024	In person	+0.11	−0.02	0.25	2,541
issue	Cultural index, GSS 2022 and 2024	Web	+0.15	0.02	0.28	1,952
issue	Racial index, GSS 2022 and 2024	In person	+0.29	0.15	0.44	2,177
issue	Racial index, GSS 2022 and 2024	Web	+0.29	0.18	0.40	2,782

In both modes the vote gap includes zero in 2012 and lies above zero in 2016 and 2024, the party-identification gap lies above zero in 2024, and the in-person dot lies to the right of the web dot in 2016 and 2024. In 2024 the vote gap was +30.5 points in person and +20.2 on the web. Each mode uses its own ANES weight. The mode groups cover 2,045 of the 3,918 voters of 2024, and the panel and paper cases, which have no mode weight, show a 2024 vote gap of +18.5 points. The issue view compares the cultural, racial, economic and immigration gaps by mode in the ANES and the GSS.Note: The ANES has no mode weights for 2020, so 2020 has no row. The panel and paper group uses the full-sample weight and appears for 2024 only, because in 2012 and 2016 every respondent is in a mode group. A cell needs at least 100 respondents and a Kish effective sample size of at least 100; cells below that are left out.Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Note: In the issue view each mode is standardized within itself, and the GSS in-person and web groups exclude the small residual interview modes.Source: ANES Cumulative Data File, 2012, 2016 and 2024; weighted (mode weights weight_ftf and weight_web); General Social Survey, 2022 and 2024; weighted (weight_use)

In the CES the steps between education levels grew

A split at the four-year degree hides the steps on either side of it. Figure 5 follows four education levels in the CES. In 2008 the step from some college to a four-year degree was −0.1 (95% CI −2.0 to 1.8) points, an interval that includes zero, and the step from a four-year degree to a postgraduate degree, +6.0 (95% CI 3.6 to 8.4) points, had the largest point estimate of the three. By 2024 every step is above zero: +7.2 (95% CI 5.1 to 9.3) points from a high school education or less to some college, +5.4 (95% CI 3.5 to 7.3) from some college to a four-year degree, and +9.3 (95% CI 7.2 to 11.4) from a four-year degree to a postgraduate degree. Postgraduates were 9.0 (95% CI 6.6 to 11.4) points more Democratic than voters with a high school education or less in 2008 and 21.9 (95% CI 19.6 to 24.2) points in 2024. Among white voters the span was 27.0 (95% CI 24.6 to 29.4) points in 2024.

Figure 5

In the CES the Democratic share rose at each step of education by 2016, while in 2008 a four-year degree did not differ from some college

Democratic share of the two-party presidential vote by highest degree, CES, 2008 to 2024; voters who reported a major-party vote, all voters (the button shows white non-Hispanic voters)

Democratic share of the two-party vote (%)

Show data

Data for In the CES the Democratic share rose at each step of education by 2016, while in 2008 a four-year degree did not differ from some college
Population	Education	Election	Democratic share (%)	95% CI low	95% CI high	n voters
All voters	Four-year degree	2008	54.4	53.0	55.8	5,430
All voters	High school or less	2008	51.3	49.9	52.8	8,004
All voters	Postgraduate degree	2008	60.4	58.4	62.3	2,680
All voters	Some college	2008	54.5	53.2	55.8	7,471
All voters	Four-year degree	2012	54.8	53.3	56.3	9,203
All voters	High school or less	2012	50.8	49.3	52.2	9,248
All voters	Postgraduate degree	2012	60.8	59.0	62.7	5,907
All voters	Some college	2012	52.0	50.7	53.2	13,431
All voters	Four-year degree	2016	56.8	55.7	57.9	10,205
All voters	High school or less	2016	43.9	42.5	45.3	9,580
All voters	Postgraduate degree	2016	64.0	62.7	65.4	6,347
All voters	Some college	2016	50.7	49.7	51.7	14,759
All voters	Four-year degree	2020	58.5	57.4	59.6	11,567
All voters	High school or less	2020	43.8	42.4	45.2	10,646
All voters	Postgraduate degree	2020	67.3	66.0	68.6	7,301
All voters	Some college	2020	49.2	48.2	50.2	14,358
All voters	Four-year degree	2024	52.6	51.2	53.9	10,429
All voters	High school or less	2024	40.0	38.3	41.6	10,255
All voters	Postgraduate degree	2024	61.9	60.3	63.4	6,565
All voters	Some college	2024	47.2	45.8	48.5	14,759
White non-Hispanic voters	Four-year degree	2008	49.6	48.1	51.1	4,676
White non-Hispanic voters	High school or less	2008	41.9	40.4	43.4	6,826
White non-Hispanic voters	Postgraduate degree	2008	58.0	55.9	60.0	2,406
White non-Hispanic voters	Some college	2008	45.4	44.0	46.8	5,981
White non-Hispanic voters	Four-year degree	2012	50.8	49.1	52.5	7,351
White non-Hispanic voters	High school or less	2012	42.2	40.6	43.7	7,719
White non-Hispanic voters	Postgraduate degree	2012	58.9	56.9	61.0	4,867
White non-Hispanic voters	Some college	2012	45.0	43.6	46.4	10,419
White non-Hispanic voters	Four-year degree	2016	52.6	51.3	53.9	7,165
White non-Hispanic voters	High school or less	2016	33.2	31.9	34.6	7,960
White non-Hispanic voters	Postgraduate degree	2016	61.7	60.1	63.2	4,893
White non-Hispanic voters	Some college	2016	41.9	40.8	43.1	10,709
White non-Hispanic voters	Four-year degree	2020	54.0	52.8	55.2	8,433
White non-Hispanic voters	High school or less	2020	30.1	28.8	31.3	8,684
White non-Hispanic voters	Postgraduate degree	2020	64.3	62.8	65.7	5,637
White non-Hispanic voters	Some college	2020	39.8	38.8	40.9	10,103
White non-Hispanic voters	Four-year degree	2024	48.7	47.1	50.2	7,929
White non-Hispanic voters	High school or less	2024	31.8	30.1	33.6	8,226
White non-Hispanic voters	Postgraduate degree	2024	58.8	57.1	60.5	4,888
White non-Hispanic voters	Some college	2024	40.5	39.1	41.9	10,814

The four lines fan out: in 2008 the lines for some college and for a four-year degree overlap and postgraduates sit above them, and by 2024 the lines are separated at every level. Postgraduates were 9.0 points more Democratic than voters with a high school education or less in 2008 and 21.9 points in 2024. Each point is a weighted Democratic share with a 95% interval.Note: The CES is an opt-in online panel, so its intervals are model based and ignore how respondents came to be in the panel.Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Source: Cooperative Election Study cumulative file, 2008-2024; weighted (weight_cumulative)

The graduate comparison used in the rest of the chapter is a cut of this gradient at the four-year degree.

Graduates differ most on cultural, racial and immigration items

Figure 6 gives the graduate gap on each item in the ANES studies of 1992 to 2000 and of 2016 to 2024. Every item is oriented so that a higher value is the more liberal answer: more government action on the economic and racial items, the more permissive answer on the cultural items, and more immigration. The gap is the weighted mean answer of graduates minus that of other respondents, in standard deviations of the item among all respondents in the period. The racial resentment index is the mean of the items from Kinder and Sanders [2], some worded so that agreement is resentful and the rest so that agreement is sympathetic, which cancels a tendency to agree with any statement, and Chapter 5.2, Racial attitudes and ideology defines it. Moral traditionalism is the mean of a pair of items, one worded in each direction.On the abortion item, a late gap of +0.32 (95% CI 0.28 to 0.36) standard deviations is a difference of +14.9 (95% CI 13.0 to 16.9) points in the share saying abortion should always be allowed, 58.2% of graduates against 43.2% of other respondents. On immigration, 26.4% of graduates and 42.0% of other respondents wanted immigration decreased in 2016 to 2024.

Figure 6

The economic gap moved toward zero between 1992 to 2000 and 2016 to 2024, while most cultural, racial and immigration gaps stayed positive

Graduate minus other respondents on each item, standard deviations of the item among all respondents; items oriented so that a higher value is the more liberal answer; ANES studies of 1992 to 2000 and 2016 to 2024 (buttons: ANES in person only, GSS, CES); 95% intervals

1992 to 2000
2016 to 2024

Graduate minus non-graduate gap (standard deviations; positive = more liberal answers among graduates)

Show data

Data for The economic gap moved toward zero between 1992 to 2000 and 2016 to 2024, while most cultural, racial and immigration gaps stayed positive
Survey view	Domain	Item	Early gap (SD)	95% CI low early	95% CI high early	n early	Late gap (SD)	95% CI low late	95% CI high late	n late	Change (SD)
ANES	Economic items	Government should guarantee jobs	−0.23	−0.28	−0.18	7,299	+0.03	−0.02	0.07	15,657	+0.25
ANES	Economic items	Government health insurance	−0.23	−0.29	−0.17	6,153	+0.09	0.05	0.13	15,720	+0.32
ANES	Economic items	More government services and spending	−0.29	−0.34	−0.24	6,985	−0.04	−0.08	0.00	15,279	+0.25
ANES	Economic items	More federal spending on the poor	−0.36	−0.42	−0.30	5,810	−0.04	−0.08	0.00	17,496	+0.31
ANES	Economic items	More federal spending on welfare	−0.10	−0.15	−0.05	7,492	+0.09	0.05	0.13	17,475	+0.19
ANES	Cultural items	Abortion always a personal choice	+0.37	0.32	0.42	8,731	+0.32	0.28	0.36	16,947	−0.05
ANES	Cultural items	Gay couples may adopt	+0.32	0.24	0.40	3,741	+0.24	0.21	0.28	17,295	−0.08
ANES	Cultural items	Job protection for gay people	+0.17	0.11	0.24	4,989	+0.24	0.20	0.28	17,386	+0.07
ANES	Cultural items	Moral traditionalism (two items)	+0.07	0.01	0.12	8,200	+0.22	0.18	0.27	15,550	+0.15
ANES	Racial items	Government effort to help Black Americans	+0.18	0.13	0.24	7,419	+0.23	0.19	0.27	15,764	+0.05
ANES	Racial items	Affirmative action in hiring and promotion	−0.06	−0.11	−0.01	7,977	+0.05	0.01	0.09	15,261	+0.11
ANES	Racial items	Racial resentment (four items)	+0.44	0.38	0.50	6,692	+0.48	0.44	0.52	15,748	+0.04
ANES	Immigration	More immigrants admitted	+0.36	0.30	0.41	6,543	+0.37	0.33	0.41	15,694	+0.02
ANES face to face	Economic items	Government should guarantee jobs	−0.23	−0.28	−0.17	6,391	+0.14	0.03	0.25	1,998	+0.37
ANES face to face	Economic items	Government health insurance	−0.24	−0.30	−0.18	5,219	+0.21	0.10	0.33	2,058	+0.45
ANES face to face	Economic items	More government services and spending	−0.28	−0.33	−0.22	6,131	+0.04	−0.08	0.16	1,914	+0.32
ANES face to face	Economic items	More federal spending on the poor	−0.38	−0.46	−0.31	4,044	−0.07	−0.18	0.05	2,154	+0.32
ANES face to face	Economic items	More federal spending on welfare	−0.10	−0.16	−0.04	5,723	+0.20	0.09	0.32	2,149	+0.30
ANES face to face	Cultural items	Abortion always a personal choice	+0.37	0.31	0.42	6,974	+0.43	0.32	0.54	2,160	+0.06
ANES face to face	Cultural items	Gay couples may adopt	+0.37	0.26	0.47	2,088	+0.27	0.17	0.38	2,102	−0.09
ANES face to face	Cultural items	Job protection for gay people	+0.20	0.12	0.27	3,519	+0.29	0.19	0.40	2,132	+0.10
ANES face to face	Cultural items	Moral traditionalism (two items)	+0.07	0.01	0.13	6,659	+0.31	0.18	0.43	1,952	+0.24
ANES face to face	Racial items	Government effort to help Black Americans	+0.19	0.13	0.25	6,496	+0.36	0.24	0.47	1,954	+0.17
ANES face to face	Racial items	Affirmative action in hiring and promotion	−0.01	−0.07	0.05	6,351	+0.20	0.08	0.32	1,864	+0.21
ANES face to face	Racial items	Racial resentment (four items)	+0.44	0.38	0.51	5,153	+0.61	0.49	0.72	1,961	+0.16
ANES face to face	Immigration	More immigrants admitted	+0.36	0.30	0.41	6,543	+0.43	0.31	0.54	1,934	+0.07
GSS	Economic items	Government should improve the living standards of the poor	−0.22	−0.27	−0.17	8,451	−0.01	−0.06	0.05	10,473	+0.21
GSS	Economic items	Government should help pay for medical care	−0.12	−0.17	−0.06	8,483	−0.01	−0.06	0.05	10,496	+0.11
GSS	Economic items	Too little is spent on welfare	−0.03	−0.09	0.03	6,221	+0.06	0.00	0.12	7,822	+0.09
GSS	Cultural items	Legal abortion for any reason	+0.30	0.25	0.36	8,282	+0.33	0.27	0.39	7,928	+0.03
GSS	Cultural items	Sex between two adults of the same sex is not wrong	+0.44	0.38	0.49	8,101	+0.26	0.21	0.32	10,267	−0.17
GSS	Cultural items	Sex before marriage is not wrong	+0.13	0.08	0.18	8,425	+0.09	0.04	0.15	10,480	−0.04
GSS	Cultural items	Marijuana should be legal	+0.11	0.05	0.16	8,268	+0.06	−0.02	0.14	5,266	−0.04
GSS	Racial items	Government has a special obligation to help Black Americans	+0.06	0.00	0.11	8,414	+0.21	0.16	0.27	10,343	+0.16
GSS	Racial items	Too little is spent on improving conditions of Black Americans	+0.09	0.02	0.15	5,844	+0.12	0.05	0.18	7,587	+0.03
GSS	Racial items	Preferential hiring and promotion of Black Americans	0.00	−0.06	0.06	6,628	+0.11	0.06	0.17	10,071	+0.11
GSS	Racial items	Disagree that Black Americans should work up without special favors	+0.39	0.33	0.46	6,980	+0.49	0.43	0.54	10,528	+0.09
GSS	Racial items	Black-white differences are mainly due to discrimination	0.00	−0.05	0.06	8,212	+0.23	0.17	0.28	8,114	+0.22
CES	Economic items	Oppose repealing the Affordable Care Act	+0.13	0.11	0.15	55,829	+0.26	0.24	0.28	60,856	+0.13
CES	Economic items	Increase spending on welfare	−0.01	−0.03	0.02	48,648	+0.12	0.09	0.14	51,547	+0.12
CES	Economic items	Increase spending on health care	−0.07	−0.09	−0.04	48,626	+0.04	0.02	0.07	51,547	+0.11
CES	Cultural items	Always allow abortion	+0.07	0.05	0.10	55,678	+0.18	0.16	0.20	60,982	+0.11
CES	Cultural items	Oppose banning abortion after 20 weeks	+0.16	0.14	0.19	55,421	+0.33	0.31	0.36	60,976	+0.17
CES	Cultural items	Oppose prohibiting public funds for abortion	+0.12	0.10	0.15	55,465	+0.20	0.18	0.22	60,982	+0.08
CES	Immigration	Conditional legal status for undocumented immigrants	+0.14	0.12	0.17	56,200	+0.07	0.04	0.09	60,944	−0.08
CES	Immigration	Oppose increasing border security	+0.07	0.05	0.10	56,200	+0.23	0.21	0.26	60,952	+0.16

In the default view the economic items start left of zero and end near it, while the cultural, racial and immigration items are right of zero in both periods, with affirmative action the exception in the first period. The ring is the earlier period and the dot the later period. In the ANES the economic index went from −0.32 to +0.02 and the cultural index from +0.30 to +0.36. The buttons switch to in-person ANES samples only, to the GSS and to the CES.Note: Positive gaps mean graduates gave the more liberal answer: more government action on economic and racial items, the more permissive answer on cultural items and more immigration.Note: The ANES late period mixes in-person and web interviews and has no mode weights for 2020. The in-person view uses the 1992 to 1998 studies and the in-person samples of 2016 and 2024. The racial resentment index and the moral traditionalism item are averages of four and two standardized items.Note: The CES has no racial item after 2014. The GSS extract has no immigration item.Note: Items were not asked in every study of a window: gay adoption in 1992 and 2000, job protection in 1992, 1996 and 2000, immigration in 1992 to 1998, the health, poor and welfare items in four of the five early studies, and the resentment items in all but 1996 and part of 1998. The common-items view of the within-party figure uses only items asked in every study.Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Source: ANES Cumulative Data File, 1992-2024, General Social Survey, 1993-2024, and Cooperative Election Study, 2014 and 2020, as the buttons show; weighted (VCF0009z, weight_use and weight_cumulative)

The economic items have gaps below zero in 1992 to 2000, between −0.36 and −0.10 standard deviations, so graduates were more conservative than other respondents on guaranteed jobs, health insurance, government services and spending on the poor and on welfare. In 2016 to 2024 their gaps lie between −0.04 and +0.09. Of the cultural, racial and immigration items, all but one have gaps above zero in both periods. Affirmative action is the exception, at −0.06 (95% CI −0.11 to −0.01) in the first period and +0.05 (95% CI 0.01 to 0.09) in the second. The racial resentment index has the largest point estimate in both periods, +0.44 (95% CI 0.38 to 0.50) and +0.48 (95% CI 0.44 to 0.52), and immigration is +0.36 (95% CI 0.30 to 0.41) and +0.37 (95% CI 0.33 to 0.41). The abortion gap went from +0.37 (95% CI 0.32 to 0.42) to +0.32 (95% CI 0.28 to 0.36), a change whose interval includes zero.

Not every item was asked in every study of a period. Gay adoption was asked in two of the early studies, job protection in three, immigration in 1992 to 1998, and the resentment items are missing in 1996 and in part of 1998. With only the items asked in every study of both periods (jobs and government services, abortion and moral traditionalism, government help and affirmative action), the economic gap went from −0.30 (95% CI −0.36 to −0.25) to +0.02 (95% CI −0.02 to 0.06), the cultural gap from +0.27 (95% CI 0.22 to 0.33) to +0.32 (95% CI 0.28 to 0.37) and the racial gap from +0.11 (95% CI 0.05 to 0.17) to +0.19 (95% CI 0.15 to 0.24) (the common-items button of Figure 8). The racial level depends on the resentment items, which are not in this check: with them the racial index is +0.21 (95% CI 0.16 to 0.27) and +0.31 (95% CI 0.27 to 0.35). The immigration item has no counterpart because it was not asked in 2000.

The late period mixes interview modes, and mode matters for the size of the gaps and, on economic items, for their sign. Respondents interviewed in person in 2016 and 2024 show larger gaps, as point estimates, than respondents interviewed on the web in the same years on every domain: on the cultural index +0.46 (95% CI 0.35 to 0.57) against +0.34 (95% CI 0.27 to 0.41), on the racial index +0.47 (95% CI 0.35 to 0.59) against +0.22 (95% CI 0.15 to 0.29), and on the economic index +0.13 (95% CI 0.01 to 0.24) against −0.08 (95% CI −0.15 to −0.02). The late economic gap is therefore above zero in person and below zero on the web, and the pooled value near zero averages the two. The face-to-face view of Figure 6 compares in-person studies with in-person samples. In it the cultural index changes by +0.16 (95% CI 0.03 to 0.28), the racial index by +0.24 (95% CI 0.11 to 0.37) and the economic index by +0.46 (95% CI 0.34 to 0.59), and the immigration index by +0.07 (95% CI −0.06 to 0.20). The all-interview change in the cultural index is +0.05 (95% CI −0.01 to 0.12), an interval that includes zero but allows a rise of as much as 40% of the 1992 to 2000 gap, so the data do not establish that the cultural gap was constant. In the in-person subsample the cultural and racial gaps are larger in 2016 and 2024 than in the 1990s. The immigration change is +0.02 (95% CI −0.05 to 0.09) for all interviews and +0.07 (95% CI −0.06 to 0.20) in person, intervals that include zero and are wide enough to allow a rise.

The GSS and the CES ask other questions (use the buttons on the figure). The GSS agrees in direction on the economic and racial indices: its economic index went from −0.20 (95% CI −0.25 to −0.14) to +0.01 (95% CI −0.04 to 0.07) and its racial index from +0.15 (95% CI 0.10 to 0.20) to +0.29 (95% CI 0.24 to 0.34). Its cultural index went from +0.30 (95% CI 0.26 to 0.35) to +0.23 (95% CI 0.18 to 0.28), a change of −0.07 (95% CI −0.14 to 0.00) that just reaches zero. The item on sex between two adults of the same sex fell the most, +0.44 (95% CI 0.38 to 0.49) to +0.26 (95% CI 0.21 to 0.32), a change of −0.17 (95% CI −0.25 to −0.09), and the items on sex before marriage and on legal marijuana also fell by amounts whose intervals include zero. The CES battery has no racial items after 2014. Its economic, cultural and immigration indices rose from 2014 to 2020, to +0.18 (95% CI 0.16 to 0.21), +0.30 (95% CI 0.28 to 0.32) and +0.18 (95% CI 0.16 to 0.21). The CES economic index is above zero in 2020 while the pooled ANES and GSS estimates are near zero. The CES items (the Affordable Care Act, spending on welfare and on health care) differ from the ANES and GSS items, and the data cannot say whether the difference in level comes from the items or from the sample. The GSS late period includes the 2021 web survey and the mixed 2022 and 2024 surveys. In 2022 and 2024 the GSS gaps in person and on the web differ by at most 0.03 standard deviations on every domain, but each interval for that difference extends at least 0.18 from its estimate, which covers the ANES in-person minus web difference on the cultural index, +0.12 (95% CI −0.01 to 0.25), so the GSS does not rule out a mode difference of that size.

In the ANES and GSS the economic gap moved to about zero

Figure 7 follows one long-running item for each domain, so that the item set does not change from year to year. The ANES jobs item (whether the government should guarantee jobs and a good standard of living) has a negative point estimate in every study from 1974 to 2016, and its interval lies below zero in 15 of those 18 studies. The gap was −0.10 (95% CI −0.17 to −0.02) in 2016, +0.05 (95% CI −0.01 to 0.11) in 2020 and +0.07 (95% CI −0.01 to 0.15) in 2024, and the last two intervals include zero. The abortion gap lies between +0.27 and +0.52 in every study since 1980. The gap on government help for Black Americans was +0.04 (95% CI −0.04 to 0.11) in 2012, an interval that includes zero, and +0.25 (95% CI 0.17 to 0.33) in 2024. The immigration gap lies between +0.28 and +0.61 in every study since 1992.

Figure 7

The graduate gap on guaranteed jobs rose from below zero to about zero by 2020, while the gaps on abortion and immigration stayed positive in every study

Graduate minus other respondents on one long-running item per domain, standard deviations of the item among all respondents in the study year; ANES 1970 to 2024 (the button shows the GSS); 95% intervals

Graduate minus other respondents (standard deviations)

Show data

Data for The graduate gap on guaranteed jobs rose from below zero to about zero by 2020, while the gaps on abortion and immigration stayed positive in every study
Survey	Item	Study year	Gap (SD)	95% CI low	95% CI high	n respondents
ANES	Cultural: abortion always a personal choice	1980	+0.46	0.34	0.59	1,545
ANES	Cultural: abortion always a personal choice	1982	+0.52	0.40	0.64	1,358
ANES	Cultural: abortion always a personal choice	1984	+0.44	0.34	0.55	2,164
ANES	Cultural: abortion always a personal choice	1986	+0.38	0.28	0.48	2,102
ANES	Cultural: abortion always a personal choice	1988	+0.30	0.19	0.40	1,960
ANES	Cultural: abortion always a personal choice	1990	+0.44	0.34	0.54	1,904
ANES	Cultural: abortion always a personal choice	1992	+0.39	0.31	0.48	2,346
ANES	Cultural: abortion always a personal choice	1994	+0.33	0.22	0.44	1,697
ANES	Cultural: abortion always a personal choice	1996	+0.37	0.26	0.48	1,676
ANES	Cultural: abortion always a personal choice	1998	+0.36	0.23	0.49	1,255
ANES	Cultural: abortion always a personal choice	2000	+0.38	0.27	0.48	1,757
ANES	Cultural: abortion always a personal choice	2004	+0.39	0.25	0.53	1,047
ANES	Cultural: abortion always a personal choice	2008	+0.33	0.16	0.49	1,026
ANES	Cultural: abortion always a personal choice	2012	+0.27	0.20	0.34	5,787
ANES	Cultural: abortion always a personal choice	2016	+0.32	0.25	0.39	4,176
ANES	Cultural: abortion always a personal choice	2020	+0.30	0.24	0.35	7,788
ANES	Cultural: abortion always a personal choice	2024	+0.35	0.28	0.43	4,983
ANES	Economic: government should guarantee jobs	1972	+0.03	−0.08	0.13	2,128
ANES	Economic: government should guarantee jobs	1974	−0.09	−0.24	0.07	1,297
ANES	Economic: government should guarantee jobs	1976	−0.11	−0.22	0.00	1,784
ANES	Economic: government should guarantee jobs	1978	−0.10	−0.21	0.00	1,816
ANES	Economic: government should guarantee jobs	1980	−0.18	−0.30	−0.06	1,176
ANES	Economic: government should guarantee jobs	1982	−0.23	−0.35	−0.10	1,209
ANES	Economic: government should guarantee jobs	1984	−0.24	−0.34	−0.14	1,908
ANES	Economic: government should guarantee jobs	1986	−0.14	−0.27	0.00	1,000
ANES	Economic: government should guarantee jobs	1988	−0.20	−0.30	−0.10	1,690
ANES	Economic: government should guarantee jobs	1990	−0.24	−0.34	−0.13	1,664
ANES	Economic: government should guarantee jobs	1992	−0.17	−0.26	−0.07	2,106
ANES	Economic: government should guarantee jobs	1994	−0.32	−0.43	−0.20	1,616
ANES	Economic: government should guarantee jobs	1996	−0.26	−0.37	−0.15	1,551
ANES	Economic: government should guarantee jobs	1998	−0.20	−0.33	−0.08	1,118
ANES	Economic: government should guarantee jobs	2000	−0.23	−0.36	−0.10	908
ANES	Economic: government should guarantee jobs	2004	−0.11	−0.24	0.01	1,103
ANES	Economic: government should guarantee jobs	2008	−0.17	−0.33	−0.01	1,020
ANES	Economic: government should guarantee jobs	2012	−0.20	−0.27	−0.13	5,414
ANES	Economic: government should guarantee jobs	2016	−0.10	−0.17	−0.02	3,742
ANES	Economic: government should guarantee jobs	2020	+0.05	−0.01	0.11	7,213
ANES	Economic: government should guarantee jobs	2024	+0.07	−0.01	0.15	4,702
ANES	Immigration: more immigrants admitted	1992	+0.28	0.18	0.38	2,113
ANES	Immigration: more immigrants admitted	1994	+0.40	0.27	0.52	1,694
ANES	Immigration: more immigrants admitted	1996	+0.41	0.29	0.52	1,496
ANES	Immigration: more immigrants admitted	1998	+0.36	0.23	0.49	1,240
ANES	Immigration: more immigrants admitted	2004	+0.61	0.48	0.74	1,047
ANES	Immigration: more immigrants admitted	2008	+0.46	0.34	0.59	2,048
ANES	Immigration: more immigrants admitted	2012	+0.39	0.31	0.46	5,333
ANES	Immigration: more immigrants admitted	2016	+0.33	0.26	0.41	3,597
ANES	Immigration: more immigrants admitted	2020	+0.38	0.32	0.44	7,241
ANES	Immigration: more immigrants admitted	2024	+0.37	0.29	0.45	4,856
ANES	Racial: government help for Black Americans	1970	+0.24	0.09	0.39	1,346
ANES	Racial: government help for Black Americans	1972	+0.30	0.19	0.41	1,999
ANES	Racial: government help for Black Americans	1974	+0.30	0.15	0.45	1,333
ANES	Racial: government help for Black Americans	1976	+0.25	0.14	0.36	1,845
ANES	Racial: government help for Black Americans	1978	+0.09	−0.02	0.19	2,033
ANES	Racial: government help for Black Americans	1980	+0.21	0.08	0.34	1,200
ANES	Racial: government help for Black Americans	1982	+0.29	0.17	0.42	1,194
ANES	Racial: government help for Black Americans	1984	+0.13	0.02	0.25	1,937
ANES	Racial: government help for Black Americans	1986	+0.20	0.10	0.30	1,931
ANES	Racial: government help for Black Americans	1988	+0.11	0.01	0.21	1,748
ANES	Racial: government help for Black Americans	1990	+0.21	0.10	0.32	1,751
ANES	Racial: government help for Black Americans	1992	+0.28	0.19	0.38	2,168
ANES	Racial: government help for Black Americans	1994	+0.13	0.02	0.25	1,618
ANES	Racial: government help for Black Americans	1996	+0.13	0.01	0.24	1,560
ANES	Racial: government help for Black Americans	1998	+0.16	0.03	0.29	1,150
ANES	Racial: government help for Black Americans	2000	+0.14	0.00	0.28	923
ANES	Racial: government help for Black Americans	2004	+0.20	0.07	0.34	1,073
ANES	Racial: government help for Black Americans	2008	+0.08	−0.04	0.20	1,910
ANES	Racial: government help for Black Americans	2012	+0.04	−0.04	0.11	5,131
ANES	Racial: government help for Black Americans	2016	+0.15	0.07	0.22	3,724
ANES	Racial: government help for Black Americans	2020	+0.24	0.18	0.30	7,294
ANES	Racial: government help for Black Americans	2024	+0.25	0.17	0.33	4,746
GSS	Cultural: legal abortion for any reason	1977	+0.48	0.31	0.64	1,473
GSS	Cultural: legal abortion for any reason	1978	+0.50	0.31	0.70	1,481
GSS	Cultural: legal abortion for any reason	1980	+0.53	0.36	0.70	1,402
GSS	Cultural: legal abortion for any reason	1982	+0.37	0.19	0.55	1,752
GSS	Cultural: legal abortion for any reason	1983	+0.58	0.41	0.75	1,514
GSS	Cultural: legal abortion for any reason	1984	+0.53	0.37	0.69	1,417
GSS	Cultural: legal abortion for any reason	1985	+0.48	0.31	0.64	1,481
GSS	Cultural: legal abortion for any reason	1987	+0.47	0.32	0.62	1,722
GSS	Cultural: legal abortion for any reason	1988	+0.62	0.44	0.81	935
GSS	Cultural: legal abortion for any reason	1989	+0.36	0.18	0.54	983
GSS	Cultural: legal abortion for any reason	1990	+0.31	0.13	0.50	873
GSS	Cultural: legal abortion for any reason	1991	+0.44	0.26	0.61	947
GSS	Cultural: legal abortion for any reason	1993	+0.34	0.18	0.50	1,009
GSS	Cultural: legal abortion for any reason	1994	+0.24	0.13	0.36	1,931
GSS	Cultural: legal abortion for any reason	1996	+0.35	0.23	0.46	1,814
GSS	Cultural: legal abortion for any reason	1998	+0.29	0.16	0.41	1,773
GSS	Cultural: legal abortion for any reason	2000	+0.34	0.22	0.47	1,755
GSS	Cultural: legal abortion for any reason	2002	+0.26	0.08	0.44	899
GSS	Cultural: legal abortion for any reason	2004	+0.44	0.27	0.61	852
GSS	Cultural: legal abortion for any reason	2006	+0.36	0.23	0.48	1,939
GSS	Cultural: legal abortion for any reason	2008	+0.39	0.25	0.54	1,298
GSS	Cultural: legal abortion for any reason	2010	+0.39	0.24	0.54	1,230
GSS	Cultural: legal abortion for any reason	2012	+0.45	0.30	0.59	1,248
GSS	Cultural: legal abortion for any reason	2014	+0.45	0.34	0.57	1,653
GSS	Cultural: legal abortion for any reason	2016	+0.35	0.23	0.47	1,808
GSS	Cultural: legal abortion for any reason	2018	+0.38	0.24	0.51	1,524
GSS	Cultural: legal abortion for any reason	2021	+0.35	0.23	0.48	1,322
GSS	Cultural: legal abortion for any reason	2022	+0.28	0.10	0.45	1,136
GSS	Cultural: legal abortion for any reason	2024	+0.28	0.16	0.40	2,138
GSS	Economic: government should improve the living standards of the poor	1975	−0.25	−0.40	−0.10	1,447
GSS	Economic: government should improve the living standards of the poor	1983	−0.28	−0.42	−0.13	1,528
GSS	Economic: government should improve the living standards of the poor	1984	−0.27	−0.41	−0.13	1,407
GSS	Economic: government should improve the living standards of the poor	1986	−0.23	−0.36	−0.10	1,429
GSS	Economic: government should improve the living standards of the poor	1987	−0.11	−0.25	0.03	1,754
GSS	Economic: government should improve the living standards of the poor	1988	−0.05	−0.21	0.11	971
GSS	Economic: government should improve the living standards of the poor	1989	−0.21	−0.38	−0.04	997
GSS	Economic: government should improve the living standards of the poor	1990	−0.20	−0.37	−0.04	874
GSS	Economic: government should improve the living standards of the poor	1991	−0.29	−0.45	−0.12	980
GSS	Economic: government should improve the living standards of the poor	1993	−0.29	−0.44	−0.14	1,025
GSS	Economic: government should improve the living standards of the poor	1994	−0.26	−0.37	−0.15	1,933
GSS	Economic: government should improve the living standards of the poor	1996	−0.23	−0.34	−0.12	1,845
GSS	Economic: government should improve the living standards of the poor	1998	−0.19	−0.31	−0.07	1,836
GSS	Economic: government should improve the living standards of the poor	2000	−0.16	−0.28	−0.04	1,812
GSS	Economic: government should improve the living standards of the poor	2002	−0.07	−0.23	0.09	888
GSS	Economic: government should improve the living standards of the poor	2004	−0.22	−0.38	−0.05	864
GSS	Economic: government should improve the living standards of the poor	2006	−0.14	−0.26	−0.02	1,958
GSS	Economic: government should improve the living standards of the poor	2008	−0.16	−0.29	−0.02	1,332
GSS	Economic: government should improve the living standards of the poor	2010	−0.18	−0.32	−0.05	1,340
GSS	Economic: government should improve the living standards of the poor	2012	−0.24	−0.38	−0.10	1,294
GSS	Economic: government should improve the living standards of the poor	2014	−0.14	−0.26	−0.03	1,638
GSS	Economic: government should improve the living standards of the poor	2016	+0.02	−0.10	0.13	1,884
GSS	Economic: government should improve the living standards of the poor	2018	+0.03	−0.10	0.17	1,519
GSS	Economic: government should improve the living standards of the poor	2021	+0.03	−0.06	0.12	2,621
GSS	Economic: government should improve the living standards of the poor	2022	+0.02	−0.11	0.15	2,322
GSS	Economic: government should improve the living standards of the poor	2024	−0.13	−0.25	−0.01	2,127
GSS	Racial: government obligation to help Black Americans	1975	+0.27	0.11	0.43	1,447
GSS	Racial: government obligation to help Black Americans	1983	+0.17	0.01	0.32	1,524
GSS	Racial: government obligation to help Black Americans	1984	+0.09	−0.06	0.25	1,397
GSS	Racial: government obligation to help Black Americans	1986	+0.25	0.11	0.40	1,426
GSS	Racial: government obligation to help Black Americans	1987	+0.25	0.12	0.39	1,757
GSS	Racial: government obligation to help Black Americans	1988	+0.28	0.10	0.45	963
GSS	Racial: government obligation to help Black Americans	1989	+0.24	0.06	0.41	995
GSS	Racial: government obligation to help Black Americans	1990	+0.30	0.16	0.43	1,318
GSS	Racial: government obligation to help Black Americans	1991	+0.01	−0.16	0.18	977
GSS	Racial: government obligation to help Black Americans	1993	−0.07	−0.22	0.08	1,017
GSS	Racial: government obligation to help Black Americans	1994	+0.05	−0.05	0.16	1,933
GSS	Racial: government obligation to help Black Americans	1996	−0.01	−0.12	0.11	1,847
GSS	Racial: government obligation to help Black Americans	1998	+0.14	0.02	0.26	1,828
GSS	Racial: government obligation to help Black Americans	2000	+0.09	−0.03	0.21	1,789
GSS	Racial: government obligation to help Black Americans	2002	+0.14	−0.03	0.31	890
GSS	Racial: government obligation to help Black Americans	2004	+0.06	−0.11	0.23	858
GSS	Racial: government obligation to help Black Americans	2006	+0.08	−0.04	0.20	1,948
GSS	Racial: government obligation to help Black Americans	2008	+0.10	−0.04	0.23	1,307
GSS	Racial: government obligation to help Black Americans	2010	−0.06	−0.20	0.08	1,322
GSS	Racial: government obligation to help Black Americans	2012	+0.07	−0.08	0.21	1,282
GSS	Racial: government obligation to help Black Americans	2014	+0.05	−0.07	0.17	1,622
GSS	Racial: government obligation to help Black Americans	2016	+0.16	0.04	0.28	1,872
GSS	Racial: government obligation to help Black Americans	2018	+0.17	0.02	0.31	1,493
GSS	Racial: government obligation to help Black Americans	2021	+0.32	0.23	0.41	2,590
GSS	Racial: government obligation to help Black Americans	2022	+0.31	0.18	0.44	2,283
GSS	Racial: government obligation to help Black Americans	2024	+0.06	−0.06	0.19	2,105

The jobs gap is negative in every study from 1974 to 2016 and is close to zero in 2020 and 2024, while the abortion and immigration gaps stay above zero. The jobs gap was −0.10 in 2016, +0.05 in 2020 and +0.07 in 2024. Each panel is one item, so the item set does not change over time. The GSS view shows the economic, cultural and racial anchors in that survey.Note: A cell needs at least 100 respondents and a Kish effective sample size of at least 100; cells below that are left out.Note: Items are oriented so that a positive gap means graduates gave the more liberal answer. GSS 2021 was collected mostly on the web and GSS 2022 and 2024 mixed in-person and web interviews.Note: Intervals are 95% intervals from the Kish effective sample size; they treat the weighted sample as a random sample and ignore the stages of the sample design.Source: ANES Cumulative Data File, 1970-2024; General Social Survey, 1975-2024; weighted (VCF0009z and weight_use)

The GSS agrees through 2022. Its economic anchor (government should improve the standard of living of the poor) went from −0.24 (95% CI −0.38 to −0.10) in 2012 to +0.02 (95% CI −0.11 to 0.15) in 2022, and then to −0.13 (95% CI −0.25 to −0.01) in 2024. The change from 2022 to 2024 is −0.14 (95% CI −0.32 to 0.03) standard deviations, an interval that includes zero, so the data do not show whether the 2024 GSS reading is a reversal or sampling variation. The ANES jobs gap does not fall in 2024. In 2016 and 2024 the web samples show lower economic gaps than the in-person samples, so the pooled ANES values near zero average over groups with different gaps.

Within party the gaps are largest among Democrats

If graduates differed from other respondents only because they had sorted into different parties, the gaps within a party would be zero. Figure 8 separates the two. It repeats the early and late comparison on each domain index for all respondents, for Democrats, for Republicans, and net of party identification, which is the graduate coefficient from a weighted regression of the index on a graduate indicator and the seven categories of the party-identification scale.

Figure 8

Graduate Democrats stay more liberal than other Democrats on cultural, racial and immigration items, and graduate Republicans are more conservative than other Republicans on economic items

Graduate minus other respondents on each domain index, standard deviations, for all respondents, Democrats, Republicans and net of party identification; ANES 1992 to 2000 and 2016 to 2024 (buttons: white non-Hispanic respondents, in-person samples, items asked in every study, GSS, CES); 95% intervals

1992 to 2000
2016 to 2024

Graduate minus non-graduate gap (standard deviations; positive = more liberal answers among graduates)

Show data

Data for Graduate Democrats stay more liberal than other Democrats on cultural, racial and immigration items, and graduate Republicans are more conservative than other Republicans on economic items
Survey view	Index	Group	Early gap (SD)	95% CI low early	95% CI high early	n early	Late gap (SD)	95% CI low late	95% CI high late	n late	Change (SD)
ANES	Economic index	All respondents	−0.32	−0.37	−0.28	8,626	+0.02	−0.02	0.06	17,568	+0.34
ANES	Economic index	Democrats	0.00	−0.07	0.06	4,308	+0.09	0.04	0.14	8,237	+0.09
ANES	Economic index	Republicans	−0.41	−0.48	−0.35	3,316	−0.27	−0.32	−0.23	7,419	+0.14
ANES	Economic index	Net of party identification	−0.21	−0.25	−0.16	8,577	−0.10	−0.13	−0.07	17,506	+0.11
ANES	Cultural index	All respondents	+0.30	0.25	0.36	8,405	+0.36	0.32	0.39	17,532	+0.05
ANES	Cultural index	Democrats	+0.67	0.60	0.75	4,181	+0.39	0.35	0.43	8,234	−0.28
ANES	Cultural index	Republicans	+0.06	−0.02	0.13	3,237	+0.15	0.09	0.21	7,406	+0.09
ANES	Cultural index	Net of party identification	+0.39	0.34	0.44	8,351	+0.26	0.22	0.29	17,476	−0.14
ANES	Racial index	All respondents	+0.21	0.16	0.27	7,974	+0.31	0.27	0.35	15,664	+0.09
ANES	Racial index	Democrats	+0.55	0.47	0.64	3,952	+0.32	0.27	0.38	7,404	−0.23
ANES	Racial index	Republicans	+0.05	−0.01	0.11	3,124	+0.06	0.01	0.10	6,613	+0.01
ANES	Racial index	Net of party identification	+0.29	0.24	0.34	7,930	+0.18	0.15	0.21	15,621	−0.11
ANES	Immigration item	All respondents	+0.36	0.30	0.41	6,543	+0.37	0.33	0.41	15,694	+0.02
ANES	Immigration item	Democrats	+0.49	0.41	0.58	3,282	+0.40	0.34	0.45	7,418	−0.10
ANES	Immigration item	Republicans	+0.23	0.14	0.31	2,531	+0.23	0.17	0.29	6,627	0.00
ANES	Immigration item	Net of party identification	+0.36	0.31	0.42	6,510	+0.30	0.26	0.34	15,647	−0.07
ANES white	Economic index	All respondents	−0.25	−0.30	−0.20	6,581	+0.14	0.09	0.18	12,630	+0.39
ANES white	Economic index	Democrats	+0.10	0.02	0.18	2,911	+0.18	0.12	0.24	5,197	+0.08
ANES white	Economic index	Republicans	−0.38	−0.44	−0.31	2,913	−0.24	−0.29	−0.20	6,252	+0.13
ANES white	Economic index	Net of party identification	−0.16	−0.20	−0.11	6,545	−0.08	−0.12	−0.05	12,602	+0.07
ANES white	Cultural index	All respondents	+0.31	0.25	0.37	6,446	+0.39	0.34	0.43	12,617	+0.08
ANES white	Cultural index	Democrats	+0.69	0.60	0.77	2,842	+0.31	0.27	0.36	5,199	−0.38
ANES white	Cultural index	Republicans	+0.08	0.00	0.16	2,859	+0.14	0.07	0.21	6,248	+0.06
ANES white	Cultural index	Net of party identification	+0.39	0.33	0.44	6,407	+0.19	0.15	0.23	12,591	−0.20
ANES white	Racial index	All respondents	+0.35	0.30	0.40	6,173	+0.47	0.43	0.52	11,396	+0.12
ANES white	Racial index	Democrats	+0.74	0.65	0.83	2,717	+0.52	0.45	0.58	4,750	−0.22
ANES white	Racial index	Republicans	+0.12	0.06	0.18	2,771	+0.11	0.07	0.15	5,611	−0.01
ANES white	Racial index	Net of party identification	+0.39	0.34	0.44	6,139	+0.27	0.23	0.31	11,375	−0.12
ANES white	Immigration item	All respondents	+0.41	0.34	0.47	5,042	+0.48	0.43	0.53	11,410	+0.08
ANES white	Immigration item	Democrats	+0.58	0.49	0.67	2,245	+0.46	0.39	0.53	4,752	−0.12
ANES white	Immigration item	Republicans	+0.26	0.17	0.35	2,249	+0.26	0.20	0.33	5,621	0.00
ANES white	Immigration item	Net of party identification	+0.41	0.34	0.47	5,018	+0.34	0.29	0.38	11,386	−0.07
ANES face to face	Economic index	All respondents	−0.34	−0.39	−0.28	6,858	+0.13	0.01	0.24	2,196	+0.46
ANES face to face	Economic index	Democrats	+0.01	−0.06	0.08	3,435	+0.20	0.07	0.34	1,047	+0.20
ANES face to face	Economic index	Republicans	−0.43	−0.50	−0.36	2,651	−0.28	−0.40	−0.15	953	+0.15
ANES face to face	Economic index	Net of party identification	−0.21	−0.26	−0.16	6,824	−0.03	−0.12	0.06	2,170	+0.18
ANES face to face	Cultural index	All respondents	+0.30	0.24	0.36	6,629	+0.46	0.35	0.57	2,195	+0.16
ANES face to face	Cultural index	Democrats	+0.70	0.62	0.78	3,306	+0.48	0.36	0.61	1,047	−0.22
ANES face to face	Cultural index	Republicans	+0.03	−0.05	0.12	2,566	+0.16	−0.01	0.33	953	+0.13
ANES face to face	Cultural index	Net of party identification	+0.39	0.34	0.45	6,589	+0.32	0.23	0.42	2,169	−0.07
ANES face to face	Racial index	All respondents	+0.23	0.17	0.29	6,412	+0.47	0.35	0.59	1,936	+0.24
ANES face to face	Racial index	Democrats	+0.61	0.51	0.70	3,190	+0.49	0.33	0.64	929	−0.12
ANES face to face	Racial index	Republicans	+0.03	−0.03	0.10	2,517	+0.07	−0.06	0.19	844	+0.03
ANES face to face	Racial index	Net of party identification	+0.31	0.25	0.37	6,379	+0.29	0.19	0.39	1,917	−0.03
ANES face to face	Immigration item	All respondents	+0.36	0.30	0.41	6,543	+0.43	0.31	0.54	1,934	+0.07
ANES face to face	Immigration item	Democrats	+0.49	0.41	0.58	3,282	+0.45	0.30	0.60	928	−0.04
ANES face to face	Immigration item	Republicans	+0.23	0.14	0.31	2,531	+0.21	0.04	0.37	842	−0.02
ANES face to face	Immigration item	Net of party identification	+0.36	0.31	0.42	6,510	+0.34	0.23	0.45	1,915	−0.03
ANES common items	Economic index	All respondents	−0.30	−0.36	−0.25	6,540	+0.02	−0.02	0.06	14,092	+0.32
ANES common items	Economic index	Democrats	0.00	−0.07	0.07	3,227	+0.01	−0.04	0.06	6,755	+0.01
ANES common items	Economic index	Republicans	−0.40	−0.47	−0.33	2,655	−0.22	−0.27	−0.17	6,079	+0.18
ANES common items	Economic index	Net of party identification	−0.19	−0.24	−0.14	6,510	−0.11	−0.15	−0.08	14,058	+0.08
ANES common items	Cultural index	All respondents	+0.27	0.22	0.33	8,023	+0.32	0.28	0.37	14,986	+0.05
ANES common items	Cultural index	Democrats	+0.63	0.56	0.71	3,989	+0.38	0.33	0.43	7,139	−0.25
ANES common items	Cultural index	Republicans	+0.04	−0.03	0.12	3,117	+0.05	−0.01	0.11	6,309	+0.01
ANES common items	Cultural index	Net of party identification	+0.36	0.31	0.41	7,980	+0.21	0.17	0.24	14,952	−0.15
ANES common items	Racial index	All respondents	+0.11	0.05	0.17	6,714	+0.19	0.15	0.24	13,651	+0.08
ANES common items	Racial index	Democrats	+0.41	0.31	0.50	3,320	+0.18	0.12	0.24	6,566	−0.23
ANES common items	Racial index	Republicans	0.00	−0.06	0.06	2,694	−0.02	−0.07	0.02	5,830	−0.02
ANES common items	Racial index	Net of party identification	+0.19	0.13	0.24	6,682	+0.07	0.03	0.10	13,625	−0.12
GSS	Economic index	All respondents	−0.20	−0.25	−0.14	8,455	+0.01	−0.04	0.07	10,477	+0.21
GSS	Economic index	Democrats	+0.04	−0.04	0.11	3,868	+0.06	0.00	0.12	4,526	+0.03
GSS	Economic index	Republicans	−0.34	−0.42	−0.26	3,121	−0.21	−0.30	−0.13	3,419	+0.13
GSS	Economic index	Net of party identification	−0.14	−0.19	−0.09	8,253	−0.09	−0.14	−0.05	10,100	+0.04
GSS	Cultural index	All respondents	+0.30	0.26	0.35	12,545	+0.23	0.18	0.28	12,069	−0.07
GSS	Cultural index	Democrats	+0.58	0.51	0.65	5,772	+0.38	0.32	0.44	5,290	−0.20
GSS	Cultural index	Republicans	+0.11	0.04	0.18	4,579	−0.04	−0.14	0.05	3,920	−0.15
GSS	Cultural index	Net of party identification	+0.34	0.30	0.39	12,278	+0.18	0.13	0.23	11,608	−0.16
GSS	Racial index	All respondents	+0.15	0.10	0.20	10,162	+0.29	0.24	0.34	13,038	+0.14
GSS	Racial index	Democrats	+0.38	0.30	0.45	4,706	+0.38	0.32	0.44	5,626	+0.01
GSS	Racial index	Republicans	−0.01	−0.07	0.06	3,670	+0.01	−0.06	0.07	4,248	+0.01
GSS	Racial index	Net of party identification	+0.19	0.14	0.24	9,957	+0.18	0.14	0.22	12,564	−0.01
CES	Economic index	All respondents	+0.03	0.01	0.06	48,781	+0.18	0.16	0.21	51,551	+0.15
CES	Economic index	Democrats	+0.17	0.14	0.20	22,211	+0.17	0.14	0.19	25,217	0.00
CES	Economic index	Republicans	−0.16	−0.19	−0.14	16,996	−0.15	−0.18	−0.12	17,944	+0.01
CES	Economic index	Net of party identification	−0.01	−0.03	0.01	46,694	0.00	−0.02	0.02	50,405	+0.01
CES	Cultural index	All respondents	+0.16	0.13	0.18	55,757	+0.30	0.28	0.32	60,988	+0.14
CES	Cultural index	Democrats	+0.36	0.33	0.39	25,526	+0.40	0.37	0.42	29,351	+0.04
CES	Cultural index	Republicans	−0.07	−0.10	−0.04	18,935	−0.07	−0.10	−0.05	20,412	0.00
CES	Cultural index	Net of party identification	+0.14	0.12	0.16	53,057	+0.17	0.15	0.19	58,620	+0.02
CES	Immigration item	All respondents	+0.13	0.11	0.16	56,200	+0.18	0.16	0.21	60,910	+0.05
CES	Immigration item	Democrats	+0.29	0.26	0.32	25,737	+0.23	0.21	0.26	29,318	−0.06
CES	Immigration item	Republicans	−0.03	−0.06	0.00	19,039	−0.09	−0.12	−0.06	20,377	−0.06
CES	Immigration item	Net of party identification	+0.12	0.10	0.14	53,459	+0.06	0.04	0.08	58,543	−0.06

Among Democrats the cultural and racial gaps are right of zero in both periods and narrower in the later period, among Republicans they are smaller, and the economic gap among Republicans is left of zero in both periods. The last row of each block is the graduate coefficient from a regression that holds the seven party-identification categories fixed. For the cultural index the overall gap was +0.36 in 2016 to 2024, against +0.39 among Democrats, +0.15 among Republicans and +0.26 net of party identification.Note: Democrats and Republicans include leaners; pure independents are in the all-respondents row and the net row only. Each index is standardized among all respondents in the period before the group is selected.Note: The net row comes from a weighted linear regression with HC3 errors; the other rows use Kish intervals. The CES view compares 2014 with 2020 and has no racial index; the GSS view has no immigration index.Note: The in-person view uses the 1992 to 1998 studies and the in-person samples of 2016 and 2024. The common-items view uses jobs and government services (economic), abortion and moral traditionalism (cultural) and government help and affirmative action (racial), the items asked in every study of both windows; it has no immigration index.Note: A cell needs at least 100 respondents and a Kish effective sample size of at least 100; cells below that are left out.Source: ANES Cumulative Data File, 1992-2024, General Social Survey, 1993-2024, and Cooperative Election Study, 2014 and 2020, as the buttons show; weighted (VCF0009z, weight_use and weight_cumulative)

Among Democrats the ANES gap on the cultural index was +0.67 (95% CI 0.60 to 0.75) in 1992 to 2000 and +0.39 (95% CI 0.35 to 0.43) in 2016 to 2024. Among Republicans it was +0.06 (95% CI −0.02 to 0.13) in the first period, an interval that includes zero, and +0.15 (95% CI 0.09 to 0.21) in the second. The overall cultural gap changed by +0.05 (95% CI −0.01 to 0.12), and the gap within the Democratic Party by −0.28 (95% CI −0.36 to −0.20). The overall gap mixes the two within-party gaps with the difference in party mix between graduates and other respondents, and that difference changed over the same years (Figure 2). The racial index shows the same pattern, +0.55 (95% CI 0.47 to 0.64) falling to +0.32 (95% CI 0.27 to 0.38) among Democrats, against +0.05 (95% CI −0.01 to 0.11) among Republicans in the first period, an interval that includes zero, and +0.06 (95% CI 0.01 to 0.10) in the second. On immigration the gap persists in both parties, +0.40 (95% CI 0.34 to 0.45) among Democrats and +0.23 (95% CI 0.17 to 0.29) among Republicans in 2016 to 2024, with no change between the periods whose interval excludes zero in either party.

Economic items differ in sign. In 2016 to 2024 the economic gap was +0.09 (95% CI 0.04 to 0.14) among Democrats and −0.27 (95% CI −0.32 to −0.23) among Republicans, so within the Republican Party graduates are more conservative on economic items. The overall economic gap, +0.02 (95% CI −0.02 to 0.06), combines a small positive gap among Democrats, a larger negative gap among Republicans and a difference in party mix. Net of party identification the economic gap is −0.10 (95% CI −0.13 to −0.07), against −0.21 (95% CI −0.25 to −0.16) in 1992 to 2000.

Among white respondents net of party identification the late gaps were +0.19 (95% CI 0.15 to 0.23) on the cultural index, +0.27 (95% CI 0.23 to 0.31) on the racial index, +0.34 (95% CI 0.29 to 0.38) on immigration and −0.08 (95% CI −0.12 to −0.05) on economic items. The GSS and the CES give positive cultural gaps net of party, +0.18 (95% CI 0.13 to 0.23) and +0.17 (95% CI 0.15 to 0.19). The CES economic gap net of party is 0.00 (95% CI −0.02 to 0.02). The surveys disagree about the Republican cultural gap. It is +0.15 (95% CI 0.09 to 0.21) in the ANES, −0.04 (95% CI −0.14 to 0.05) in the GSS and −0.07 (95% CI −0.10 to −0.05) in the CES, and the ANES abortion item alone gives +0.11 (95% CI 0.05 to 0.17) among Republicans, against negative gaps on the CES abortion items. The data do not settle whether graduate Republicans are more liberal than other Republicans on cultural items. Party identification is itself shaped by the attitudes in question, so these estimates describe where graduates and non-graduates of the same party stand and do not isolate an effect of education. Chapter 1.3, Sort dates by attitude dates when party gaps on issues became large, and Chapter 7.2, Economic and cultural dimensions examines the economic and cultural dimensions themselves.

Income rank accounts for part of the rise in the vote gap

The ANES records family income as a percentile group within each year, which sets aside the changing value of the dollar. Figure 9 compares the vote gap with the gap from a weighted linear probability model that adds the five income groups, and for all voters adds race. In 2024 the gap was +23.6 (95% CI 19.1 to 28.0) points before the adjustment and +22.1 (95% CI 17.2 to 27.0) after it, and in 2020 +15.1 (95% CI 11.8 to 18.4) and +17.5 (95% CI 13.9 to 21.0). In every one of the 17 elections before 2024 the adjusted gap is higher than the unadjusted gap, by 1.3 to 8.3 points: among voters with the same income rank, graduates were more Democratic relative to non-graduates than the raw comparison shows. In 1992, for example, the gap was −9.8 (95% CI −15.8 to −3.8) raw and −4.1 (95% CI −10.5 to 2.4) net of income rank. The adjustment lowers the 2024 end of the series and raises the earlier years, so it shrinks the rise: from 2012 to 2024 the gap rose by +24.5 (95% CI 18.3 to 30.7) points unadjusted, +18.6 (95% CI 11.9 to 25.4) net of income rank and +16.9 (95% CI 10.4 to 23.4) net of income rank and race, and from 2008 by +30.0 (95% CI 21.6 to 38.4) and +20.3 (95% CI 11.4 to 29.2).

Figure 9

Holding income rank fixed shrinks the rise in the vote gap from 2012 to 2024 from +24.5 to +18.6 points and leaves most of it

Graduate minus other voters in the Democratic share of the two-party presidential vote, points, ANES 1956 to 2024, from weighted linear probability models on voters with a reported family income; all voters (the button shows white non-Hispanic voters); 95% HC3 intervals

Show data

Data for Holding income rank fixed shrinks the rise in the vote gap from 2012 to 2024 from +24.5 to +18.6 points and leaves most of it
Population	Specification	Election	Gap (points)	95% CI low	95% CI high	n voters
All voters	Graduate indicator alone	1956	−9.4	−18.3	−0.5	1,210
All voters	Graduate indicator alone	1960	−12.2	−24.6	0.3	891
All voters	Graduate indicator alone	1964	−21.3	−30.2	−12.4	1,070
All voters	Graduate indicator alone	1968	−9.8	−18.3	−1.4	892
All voters	Graduate indicator alone	1972	+4.6	−1.8	11.0	1,546
All voters	Graduate indicator alone	1976	−15.1	−22.2	−8.1	1,239
All voters	Graduate indicator alone	1980	−11.8	−20.0	−3.6	793
All voters	Graduate indicator alone	1984	−1.7	−8.2	4.9	1,245
All voters	Graduate indicator alone	1988	−7.5	−14.0	−0.9	1,082
All voters	Graduate indicator alone	1992	−9.8	−15.8	−3.8	1,238
All voters	Graduate indicator alone	1996	−11.6	−18.8	−4.4	941
All voters	Graduate indicator alone	2000	−0.3	−7.4	6.8	941
All voters	Graduate indicator alone	2004	+1.3	−6.8	9.5	729
All voters	Graduate indicator alone	2008	−6.5	−13.6	0.7	1,442
All voters	Graduate indicator alone	2012	−0.9	−5.2	3.4	4,040
All voters	Graduate indicator alone	2016	+12.1	7.6	16.6	2,542
All voters	Graduate indicator alone	2020	+15.1	11.8	18.4	5,902
All voters	Graduate indicator alone	2024	+23.6	19.1	28.0	3,570
All voters	Net of income rank	1956	−4.4	−13.8	5.0	1,210
All voters	Net of income rank	1960	−8.3	−21.7	5.0	891
All voters	Net of income rank	1964	−16.9	−26.4	−7.5	1,070
All voters	Net of income rank	1968	−8.5	−17.4	0.4	892
All voters	Net of income rank	1972	+9.7	3.1	16.3	1,546
All voters	Net of income rank	1976	−8.3	−15.6	−0.9	1,239
All voters	Net of income rank	1980	−5.8	−14.3	2.7	793
All voters	Net of income rank	1984	+5.4	−1.2	12.1	1,245
All voters	Net of income rank	1988	−2.4	−9.3	4.5	1,082
All voters	Net of income rank	1992	−4.1	−10.5	2.4	1,238
All voters	Net of income rank	1996	−5.3	−13.0	2.5	941
All voters	Net of income rank	2000	+5.1	−2.5	12.8	941
All voters	Net of income rank	2004	+7.9	−0.9	16.7	729
All voters	Net of income rank	2008	+1.8	−5.6	9.2	1,442
All voters	Net of income rank	2012	+3.5	−1.1	8.1	4,040
All voters	Net of income rank	2016	+15.3	10.4	20.1	2,542
All voters	Net of income rank	2020	+17.5	13.9	21.0	5,902
All voters	Net of income rank	2024	+22.1	17.2	27.0	3,570
All voters	Net of income rank and race	1956	−4.5	−13.8	4.9	1,210
All voters	Net of income rank and race	1960	−8.0	−21.5	5.4	891
All voters	Net of income rank and race	1964	−17.2	−26.4	−8.1	1,070
All voters	Net of income rank and race	1968	−7.5	−16.0	1.0	892
All voters	Net of income rank and race	1972	+9.3	3.0	15.5	1,546
All voters	Net of income rank and race	1976	−6.6	−13.9	0.7	1,239
All voters	Net of income rank and race	1980	−5.0	−13.4	3.3	793
All voters	Net of income rank and race	1984	+6.3	−0.3	12.9	1,245
All voters	Net of income rank and race	1988	−1.4	−8.1	5.3	1,082
All voters	Net of income rank and race	1992	−3.6	−10.0	2.8	1,238
All voters	Net of income rank and race	1996	−3.4	−11.0	4.2	941
All voters	Net of income rank and race	2000	+6.4	−1.2	14.1	941
All voters	Net of income rank and race	2004	+10.1	1.7	18.4	729
All voters	Net of income rank and race	2008	+4.2	−3.0	11.4	1,442
All voters	Net of income rank and race	2012	+5.8	1.3	10.3	4,040
All voters	Net of income rank and race	2016	+16.8	12.1	21.5	2,542
All voters	Net of income rank and race	2020	+18.4	15.0	21.7	5,902
All voters	Net of income rank and race	2024	+22.6	17.9	27.4	3,570
White non-Hispanic voters	Graduate indicator alone	1956	−9.1	−18.1	−0.2	1,158
White non-Hispanic voters	Graduate indicator alone	1964	−22.1	−31.3	−12.9	978
White non-Hispanic voters	Graduate indicator alone	1968	−8.3	−16.8	0.3	803
White non-Hispanic voters	Graduate indicator alone	1972	+5.3	−1.2	11.8	1,384
White non-Hispanic voters	Graduate indicator alone	1976	−10.5	−17.8	−3.2	1,108
White non-Hispanic voters	Graduate indicator alone	1980	−7.1	−15.6	1.3	676
White non-Hispanic voters	Graduate indicator alone	1984	+3.8	−3.1	10.8	1,043
White non-Hispanic voters	Graduate indicator alone	1988	−1.6	−8.7	5.4	864
White non-Hispanic voters	Graduate indicator alone	1992	−6.7	−13.4	0.0	964
White non-Hispanic voters	Graduate indicator alone	1996	−5.2	−13.2	2.7	760
White non-Hispanic voters	Graduate indicator alone	2000	+6.3	−1.6	14.1	751
White non-Hispanic voters	Graduate indicator alone	2004	+6.4	−2.8	15.6	544
White non-Hispanic voters	Graduate indicator alone	2008	−0.7	−9.1	7.7	743
White non-Hispanic voters	Graduate indicator alone	2012	+6.5	1.5	11.5	2,466
White non-Hispanic voters	Graduate indicator alone	2016	+21.6	16.5	26.7	1,910
White non-Hispanic voters	Graduate indicator alone	2020	+25.0	21.4	28.7	4,469
White non-Hispanic voters	Graduate indicator alone	2024	+28.5	23.3	33.6	2,738
White non-Hispanic voters	Net of income rank	1956	−4.4	−13.9	5.0	1,158
White non-Hispanic voters	Net of income rank	1964	−18.9	−28.6	−9.1	978
White non-Hispanic voters	Net of income rank	1968	−8.2	−17.2	0.8	803
White non-Hispanic voters	Net of income rank	1972	+8.6	1.9	15.4	1,384
White non-Hispanic voters	Net of income rank	1976	−5.5	−13.1	2.0	1,108
White non-Hispanic voters	Net of income rank	1980	−3.6	−12.4	5.3	676
White non-Hispanic voters	Net of income rank	1984	+8.7	1.7	15.7	1,043
White non-Hispanic voters	Net of income rank	1988	+1.4	−5.9	8.8	864
White non-Hispanic voters	Net of income rank	1992	−2.2	−9.3	5.0	964
White non-Hispanic voters	Net of income rank	1996	−1.4	−9.8	7.0	760
White non-Hispanic voters	Net of income rank	2000	+9.9	1.4	18.3	751
White non-Hispanic voters	Net of income rank	2004	+11.1	1.3	20.9	544
White non-Hispanic voters	Net of income rank	2008	+6.7	−1.9	15.3	743
White non-Hispanic voters	Net of income rank	2012	+8.1	2.9	13.4	2,466
White non-Hispanic voters	Net of income rank	2016	+21.6	16.3	27.0	1,910
White non-Hispanic voters	Net of income rank	2020	+25.4	21.5	29.4	4,469
White non-Hispanic voters	Net of income rank	2024	+26.7	21.1	32.4	2,738

The adjusted line lies above the unadjusted line in every election before 2024 and the two lines meet in 2024, so the adjusted rise is smaller: from 2008 to 2024 it is +20.3 points net of income rank against +30.0 unadjusted. In 2016, 2020 and 2024 the two estimates differ by at most 3.2 points. Income is the respondent's family income percentile group within the study year, and the adjusted lines add the five income groups (and, for all voters, a white indicator) to a graduate indicator.Note: Elections where either education group has fewer than 100 voters with a reported income are left out (1952).Note: Every line comes from the same rows within a view, those with a reported income. Intervals use HC3 robust errors.Source: ANES Cumulative Data File, 1952-2024; weighted (VCF0009z; the file gives every respondent the same weight in most studies before 1992)

Income rank as measured here shrinks the rise in the vote gap by 23.9% since 2012 and 32.3% since 2008 and leaves most of it. Chapter 3.3, The income ladder shows that in the CES the highest income bracket became more Democratic after 2008 and that income and education overlap. A study using European survey data argues that workplace experience shapes political views on redistribution, immigration and abortion [3], and this chapter has no measure of occupation to separate that from the degree. A study of Western democracies finds that higher-educated voters moved toward the left while high-income voters stayed on the right [4], and a book-length account of the American case argues that Democrats became the party of highly educated voters with progressive social views and Republicans the party of white voters without college degrees [5]. Chapter 12.2, The diploma map measures the same education gradient across counties, and Chapter 7.6, Education and ideology (planned) asks how education relates to ideology across countries.

What these data cannot show

Every survey here interviews a new sample each year, so a rising gap can come from people changing their votes, from new voters replacing older ones, or from changes in who has a degree. The graduate share of respondents rose from 6.4% in 1952 to 34.4% in 2024, and the data cannot separate a change in what graduates think from a change in who counts as a graduate. The chapter compares groups and not people. The vote gap is a difference in the Democratic share among voters who chose a major party, and it leaves out nonvoters and third-party voters; Chapter 3.1, The two coalitions takes up the composition of each party's voters.

The two vote surveys differ by as much as 11.0 points in the gap in a year. The ANES voters of 2024 are more Democratic than the certified count, with the excess among graduates, and the CES matches the count nationally. The count has no breakdown by degree, so it cannot show which survey is closer on graduates or on the gap, and a national match does not show that the CES is accurate within education groups. The vote is a self-report in both surveys, and no validated vote is available in either file. The ANES mode split has wide intervals, leaves out the panel and paper respondents of 2024 and does not exist for 2020. The in-person ANES samples give larger gaps than the web samples, the GSS cannot rule out a mode difference of that size, and the CES, which is online in every year, has smaller vote gaps than the in-person ANES in 2024.

The issue gaps rest on a few items per domain, and the sets differ across surveys and across studies within a period. A standardized gap depends on how answers spread in each period, so the same difference in the share giving an answer can be a different number of standard deviations in two periods. The CES policy battery ends in 2021 and has no racial item after 2014, and the GSS extract has no immigration item, so the ANES is the only survey here that covers all four domains. The orientation of each item comes from its wording, and a check against the party gap in every year can flag a reversed code but cannot detect a change in what Democrats believe.

Finally, the within-party results describe partisans who have already sorted. Party identification follows attitudes as well as preceding them, so a gap within a party is not the effect of education on that party's voters, and the comparison cannot say whether graduate Democrats held their views before they joined the party or acquired them in it. The chapter holds only party, income rank and race fixed. Age, religion, region, occupation and the kind of place a person lives in go with education and are left free.

Data sources

ANES Cumulative Data FileCumulative Data File 1952-2024: presidential vote, party identification, income rank and issue itemsYears: 1948-2024 (33 study years; even years only after 2004; no 2006, 2010 or 2014 study) · Weights: VCF0009z; weight_ftf and weight_web for mode splits · License: Free download after accepting the ANES terms of use; cite the release.
Cooperative Election StudyCumulative Common Content 2008-2024: presidential vote and party identification; policy items for 2014 and 2020Years: 2006-2024 (every year; self-reported turnout is available in the 10 even years) · Weights: weight_cumulative · License: See the provider's terms of use.
General Social Survey1993-2024 waves with the raw-file interview mode: issue items by degreeYears: 1972-2024 (35 study years) · Weights: weight_use · License: See the provider's terms of use.
State presidential returnsMIT Election Lab state presidential returns 1976-2024: certified two-party Democratic share for the survey checkYears: 1976-2024 (13 elections) · Weights: none (full counts) · License: See the provider's terms of use.

All data sources

Methods

Sample: three surveys. The American National Election Studies (ANES) Cumulative Data File, 1952 to 2024: 34,173 respondents with a Democratic or Republican presidential vote and a known education, and 63,120 respondents with a party identification and a known education. The Cooperative Election Study (CES) cumulative file: 188,145 respondents with a Democratic or Republican presidential vote in 2008, 2012, 2016, 2020 and 2024, and 540,578 respondents with a party identification in 2008 to 2024; the policy items are used for 2014 and 2020 only, because the CES policy battery ends in 2021 and the items used here were asked in both years. The General Social Survey (GSS), 1993 to 2024, with the interview mode joined from the raw file; it supplies issue items only. Education: ANES college_grad, a college or advanced degree (VCF0110 = 4); CES education_4cat, with College and Post-grad as graduates; GSS degree of 3 (bachelor's) or 4 (graduate). A graduate is a respondent with a four-year degree or more; everyone else, an associate degree included, is a non-graduate. Party: ANES pid_7pt, CES pid7_num and GSS partyid, with 1 to 3 (0 to 2 in the GSS) Democratic and 5 to 7 (4 to 6) Republican, leaners included; pure independents are in neither party. White: ANES race_4cat White non-Hispanic and CES race5 White (non-Hispanic); the ANES codes Hispanic origin only where the studies asked it, and the GSS is not split by race. Vote: the share of respondents reporting a Democratic vote among those reporting a Democratic or Republican presidential vote (self-reported; third-party voters and nonvoters excluded). Party identification: the Democratic share of respondents who are Democrats or Republicans. Weights: ANES weight (VCF0009z; the cumulative file gives every respondent a weight of one in most studies before 1992, so those studies are unweighted), weight_ftf (VCF0009x) and weight_web (VCF0009y) for the mode splits of 2012, 2016 and 2024; CES weight_cumulative; GSS weight_use. The ANES 2020 study has no mode weights. For 2024 the ANES weight VCF0009z is the pre-election weight (V240107a), although the vote is a post-election item; the post-election weight (weight_postelection, V240107b) is reported as a sensitivity for 2016, 2020 and 2024, together with the CES post-election weight (2012, 2016 and 2020) and the validated-vote weight (2020 and 2024). In the ANES mode splits the face-to-face and web groups use their mode weights; in 2024 they cover respondents in neither group (panel and paper cases), who are reported separately with the full-sample weight. Certified benchmark: the Democratic share of Democratic plus Republican votes in the MIT Election Data and Science Lab state presidential returns, summed over states (full counts, no interval); the survey share uses every respondent with a major-party vote, whether or not education is known, and the interval is the survey Kish interval shifted by the count. Issue items: every item is oriented so that a higher value is the more liberal answer (more government action on economic and racial items, the more permissive answer on cultural items, more immigration). The direction is a coding fact. The ANES directions were read from the value labels stored next to each item in the cumulative file, and each clean column was checked against its raw VCF code. The GSS directions were read from the question text in the GSS dictionary and the CES directions from the variable descriptions. Each item was then checked against the Democratic-minus-Republican gap, in standard deviations, in every study year in which it was asked: 181 ANES item-years, 320 GSS item-years and 65 CES item-years, and the build refuses any item-year below 0.25 standard deviations. The smallest was −0.09 standard deviations (ANES abortion, 1982) and −0.16 in the GSS (abany, 1983) and +0.52 in the CES; 4 ANES, 7 GSS and 0 CES item-years have a negative point estimate and 0 ANES, 2 GSS and 0 CES have an interval wholly below zero, all of them early studies of items that cut across the parties. A separate scan of all 52 CES policy items in the cumulative file, by year and with the 2014 ACA recode applied, found a sign change in 3 items (spending_vs_tax_0_100, support_gay_marriage_ban, support_raise_fuel_eff) and a level shift of more than 0.4 standard deviations between adjacent years in 5 items (incometax_vs_salestax_0_100, spending_vs_tax_0_100, support_gay_marriage_ban, support_raise_fuel_eff, support_tariff_china); this chapter uses none of them, and NOTES.md lists the years. Standardized gaps: an item or index is standardized as a weighted z-score among all respondents of the survey in the same unit (a study year, or a pooled window of studies); the gap is the weighted mean z of graduates minus that of non-graduates. Items were asked in different studies within a window (gay adoption in 1992 and 2000, job protection in 1992, 1996 and 2000, immigration in 1992 to 1998, the health, poor and welfare items in four of five early studies, the resentment items in all early studies but 1996 and part of 1998); a check uses only the items asked in every study of both windows (jobs and government services, abortion and moral traditionalism, government help and affirmative action). The windows are the ANES and GSS studies of 1992 to 2000 and of 2016 to 2024 (for the in-person view, the 1992 to 1998 studies and the in-person samples of 2016 and 2024), and the CES years 2014 and 2020; windows pool studies with their weights as given. Group restrictions (Democrats, Republicans, white respondents) are applied after standardizing. Indices: the mean of a domain's standardized items for respondents who answered at least two of them (one for the immigration item), standardized again. The racial resentment index is the mean of four standardized items (VCF9039 to VCF9042, two in each direction) for respondents who answered at least two, and moral traditionalism the mean of two items (VCF0852 and VCF0853) for respondents who answered both. Net of party identification: the graduate coefficient from a weighted linear regression of the index on a graduate indicator and the seven party-identification categories (HC3 errors). Intervals: 95%. Shares, means and gaps use the Kish effective sample size and a normal approximation, with the standard errors of independent groups combined in quadrature; changes between studies or windows combine the two gap standard errors in quadrature. The net-of-party rows and the income models are weighted regressions with HC3 robust errors. All intervals are model based: the weighted sample is treated as a random sample and the stages of each sample design are ignored. The CES is an opt-in panel, so its intervals ignore how respondents came to be in the panel. Income: the ANES family income percentile group within the study year (VCF0114, five groups) as a categorical control. Suppression: a cell needs at least 100 respondents and a Kish effective sample size of at least 100, and a gap needs both cells to clear it; this leaves out the ANES vote gap in 1952 and 1960 and the party-identification gap in 1954 and 1960. Rounding: displays round half up from the unrounded estimate and differences are computed before rounding. Survey mode: the ANES has in-person and web interviews in 2012, 2016 and 2024 and web, telephone and video interviews in 2020; the GSS has in-person and web interviews in 2022 and 2024 and a mostly web sample in 2021; the CES is online in every year.

References

Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
Kinder, Donald R.; Sanders, Lynn M. (1996). Divided by Color: Racial Politics and Democratic Ideals. University of Chicago Press.
Kitschelt, Herbert; Rehm, Philipp (2014). Occupations as a Site of Political Preference Formation. Comparative Political Studies 47(12): 1670-1706. doi:10.1177/0010414013516066
Gethin, Amory; Martínez-Toledano, Clara; Piketty, Thomas (2022). Brahmin Left Versus Merchant Right: Changing Political Cleavages in 21 Western Democracies, 1948-2020. Quarterly Journal of Economics 137(1): 1-48. doi:10.1093/qje/qjab036
Grossmann, Matt; Hopkins, David A. (2024). Polarized by Degrees: How the Diploma Divide and the Culture War Transformed American Politics. Cambridge University Press. doi:10.1017/9781009057141

How to cite

@misc{singh2026thediplomadivide,
  author = {Singh, Vishal},
  title = {The diploma divide},
  year = {2026},
  url = {https://vishalsingh.org/politics/the-diploma-divide},
  note = {In The Public Mind. Published 3 October 2026.}
}

Same question at another grain

These chapters take up the same axis with a different unit of analysis: individuals, places or countries.

Education
- Chapter 1.6, Political knowledge and constraint (planned) Individuals
- Chapter 12.2, The diploma map Places
- Chapter 7.6, Education and ideology (planned) Countries

Reading paths

Education and the vote