Chapter 3.3

The income ladder

From 2008 to 2024 the highest income bracket's Democratic share minus the lowest bracket's went from −9.1 to +5.7 points, and by 2024 the middle brackets were less Democratic than either end, but within education groups the 2024 gap includes zero.

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

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

The plain-language definition of the validated-vote weight is still being checked against the CES documentation.

In 2008 the lowest of five household-income brackets was the most Democratic; in 2024 the highest was. In the Cooperative Election Study, the Democratic share of the two-party presidential vote in the lowest bracket fell from 62.4% in 2008 to 51.5% in 2024, while the share in the highest bracket rose from 53.3% to 57.2%. The gap between the two, the highest bracket's Democratic share minus the lowest bracket's, measured in points, was −9.1 (95% CI −12.9 to −5.4) in 2008 and +5.7 (95% CI 1.1 to 10.3) in 2024. That is a change of +14.8 (95% CI 10.8 to 18.9) points, larger than sampling error.

The change stays above zero in five alternative estimates (Figure 7), but within education groups the 2024 gap includes zero, at −3.3 (95% CI −9.3 to 2.7) points among voters without a degree and +0.4 (95% CI −5.2 to 6.0) among degree holders (Figure 4), so the reversal combines income and education.

The 2024 profile is U-shaped. The three brackets between the ends, which cover household incomes from $30,000 to $150,000, are less Democratic than either end when pooled: the lowest bracket's share minus that of the middle three is +4.0 (95% CI 1.7 to 6.3) points, and the same difference for the highest bracket is +9.7 (95% CI 7.5 to 11.8) points.

The Cooperative Election Study (CES) is an online panel survey fielded by YouGov around each election. This chapter uses the presidential years 2008, 2012, 2016, 2020 and 2024: 188,145 respondents who reported voting for one of the two major-party candidates, of whom 170,053 also reported household income.Except where a figure says otherwise, estimates use the cumulative-file survey weight, the one weight the file defines for all five elections. Intervals come from the Kish effective sample size [1] or from standard errors clustered by state, and each figure subtitle says which; the methods note in the appendix gives the rule. The file groups the twelve income bands of the questionnaire into five brackets in nominal dollars: under $30,000, $30,000 to $60,000, $60,000 to $100,000, $100,000 to $150,000, and $150,000 or more. High income means the highest bracket everywhere in this chapter. The Democratic share is the weighted share of these voters who reported a Democratic vote, and the gap is the Democratic share in one group minus that in another.

The most Democratic bracket went from lowest to highest

Figure 1 compares the two elections bracket by bracket. In 2008 the Democratic share fell with income, from 62.4% (95% CI 60.5% to 64.4%) in the lowest bracket to 54.1%, 50.4% and 50.0% in the next three, and then rose to 53.3% in the highest. The rise at the top was small. The highest bracket was +1.3 (95% CI −1.7 to 4.3) points more Democratic than the three middle brackets pooled, and that interval includes zero.

Figure 1

The most Democratic income bracket in 2008 was the lowest, and in 2024 it was the highest

Democratic share of the two-party presidential vote by household-income bracket, 2008 and 2024 (the buttons show other elections); two-party voters who reported income; 95% intervals from the Kish effective sample size

2008
2024

Show data

Data for The most Democratic income bracket in 2008 was the lowest, and in 2024 it was the highest
Later election	Income bracket	Share 2008 (%)	95% CI low 2008	95% CI high 2008	n 2008	Share later (%)	95% CI low later	95% CI high later	n later	Change (points)	95% CI low change	95% CI high change
2024	$150k or more	53.3	50.4	56.1	1,634	57.2	55.2	59.1	4,832	+3.9	0.4	7.3
2024	$100k to $150k	50.0	48.0	51.9	3,244	49.3	47.5	51.0	6,416	−0.7	−3.3	2.0
2024	$60k to $100k	50.4	49.0	51.8	6,261	46.9	45.4	48.4	9,646	−3.5	−5.6	−1.4
2024	$30k to $60k	54.1	52.7	55.4	7,172	46.9	45.3	48.4	10,250	−7.2	−9.3	−5.1
2024	Under $30k	62.4	60.5	64.4	3,900	51.5	49.4	53.6	7,449	−10.9	−13.8	−8.1
2020	$150k or more	53.3	50.4	56.1	1,634	59.3	57.5	61.0	3,908	+6.0	2.6	9.4
2020	$100k to $150k	50.0	48.0	51.9	3,244	53.7	52.2	55.2	6,062	+3.8	1.3	6.2
2020	$60k to $100k	50.4	49.0	51.8	6,261	50.3	49.1	51.5	10,534	−0.1	−2.0	1.7
2020	$30k to $60k	54.1	52.7	55.4	7,172	51.6	50.4	52.9	11,283	−2.4	−4.3	−0.6
2020	Under $30k	62.4	60.5	64.4	3,900	56.6	55.0	58.2	7,867	−5.8	−8.3	−3.3
2016	$150k or more	53.3	50.4	56.1	1,634	58.4	56.2	60.7	2,739	+5.1	1.5	8.8
2016	$100k to $150k	50.0	48.0	51.9	3,244	52.8	51.2	54.5	5,428	+2.9	0.3	5.5
2016	$60k to $100k	50.4	49.0	51.8	6,261	50.4	49.1	51.6	10,089	−0.1	−2.0	1.8
2016	$30k to $60k	54.1	52.7	55.4	7,172	49.1	47.9	50.4	11,291	−4.9	−6.8	−3.1
2016	Under $30k	62.4	60.5	64.4	3,900	55.1	53.4	56.8	7,025	−7.4	−10.0	−4.8
2012	$150k or more	53.3	50.4	56.1	1,634	56.2	53.1	59.3	2,206	+2.9	−1.3	7.1
2012	$100k to $150k	50.0	48.0	51.9	3,244	52.0	49.8	54.2	4,300	+2.0	−0.9	5.0
2012	$60k to $100k	50.4	49.0	51.8	6,261	50.7	49.1	52.2	8,686	+0.2	−1.9	2.3
2012	$30k to $60k	54.1	52.7	55.4	7,172	53.8	52.4	55.2	10,959	−0.3	−2.3	1.7
2012	Under $30k	62.4	60.5	64.4	3,900	59.6	57.9	61.4	6,872	−2.8	−5.4	−0.2

In the default view the three lowest brackets have moved toward the Republican candidate since 2008, the lowest bracket furthest (a change of −10.9 points), and the highest bracket has moved toward the Democratic candidate (a change of +3.9 points); the fourth bracket has not moved by more than sampling error. The ring is 2008 and the dot is the later election, and the buttons switch the later election.Note: Brackets are in the nominal dollars of each survey year; the lowest and highest brackets hold a different share of voters in different years.Note: Intervals are model based (Kish effective sample size, normal approximation) and ignore clustering in the sample design.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

In 2024 the five shares, from the lowest bracket to the highest, were 51.5%, 46.9%, 46.9%, 49.3% and 57.2%. Between the two elections the Democratic share changed by −10.9 points in the lowest bracket and by −7.2, −3.5 and −0.7 in the next three, the last with an interval that includes zero. In the highest bracket it changed by +3.9 (95% CI 0.4 to 7.3) points, toward the Democratic candidate. Among all two-party voters in the survey, whether or not they reported income, the Democratic share changed by −4.9 (95% CI −6.0 to −3.9) points over the same years, so the highest bracket moved toward the Democratic candidate while the survey's voters as a whole moved away.

Work that covers a longer period found that Democratic candidates began winning a growing share of affluent voters in the 1990s, a trend concentrated among, but not limited to, college-educated voters and visible only in larger metropolitan areas [2]. The CES covers the recent end of that history. The next sections ask whether the recent change is a reversal of the income ladder or a different shape.

By 2024 both ends voted more Democratic than the middle

Figure 2 breaks the change into two arms. The highest bracket's edge over the pooled middle three brackets rose from +1.3 points in 2008 to +8.2 in 2016 and +9.7 (95% CI 7.5 to 11.8) in 2024. The lowest bracket's edge over the same middle brackets fell from +10.4 to +4.0 (95% CI 1.7 to 6.3) and stayed above zero. The fourth bracket taken alone is not distinguishable from the lowest, with a difference of +2.2 (95% CI −0.5 to 5.0) points, so the U shape rests on the two pooled contrasts and not on every bracket. A ladder turned upside down would have the Democratic share rise steadily with income. In 2024 it does not, because the lowest bracket is more Democratic than the middle three. The reversal applies only to the comparison of the two end brackets.

Figure 2

The highest bracket's edge over the middle three rose from +1.3 to +9.7 points, while the lowest bracket's fell from +10.4 to +4.0

Difference in the Democratic share of the two-party presidential vote between an end bracket and the pooled middle three brackets, percentage points, by presidential election; two-party voters who reported income; 95% intervals from the Kish effective sample size

Difference in Democratic share (percentage points)

Show data

Data for The highest bracket's edge over the middle three rose from +1.3 to +9.7 points, while the lowest bracket's fell from +10.4 to +4.0
Contrast	Election	Difference (points)	95% CI low (Kish)	95% CI high (Kish)	95% CI low (clustered by state)	95% CI high (clustered by state)	n voters
Highest bracket minus the middle three	2008	+1.3	−1.7	4.3	−1.3	3.9	18,311
Highest bracket minus the middle three	2012	+3.8	0.6	7.0	0.2	7.4	26,151
Highest bracket minus the middle three	2016	+8.2	5.8	10.5	5.6	10.7	29,547
Highest bracket minus the middle three	2020	+7.7	5.7	9.6	5.2	10.2	31,787
Highest bracket minus the middle three	2024	+9.7	7.5	11.8	6.9	12.4	31,144
Lowest bracket minus the middle three	2008	+10.4	8.3	12.6	8.4	12.5	20,577
Lowest bracket minus the middle three	2012	+7.2	5.2	9.2	4.7	9.7	30,817
Lowest bracket minus the middle three	2016	+4.8	2.9	6.7	2.6	6.9	33,833
Lowest bracket minus the middle three	2020	+5.1	3.3	6.8	3.1	7.1	35,746
Lowest bracket minus the middle three	2024	+4.0	1.7	6.3	0.9	7.0	33,761

In 2024 both ends of the income range are above the middle three brackets, the highest by +9.7 points and the lowest by +4.0. The highest bracket's edge rose from +1.3 points in 2008, an interval that includes zero, while the lowest bracket's edge fell and stayed above zero. Each panel is one contrast, the Democratic share of the first group minus that of the pooled middle three, and a value above the dashed line at zero means the first group is more Democratic. Bands are confidence intervals.Note: Brackets are in the nominal dollars of each survey year; the lowest and highest brackets hold a different share of voters in different years.Note: Intervals are model based (Kish effective sample size, normal approximation) and ignore clustering in the sample design.Note: The bands use the Kish effective sample size. The last two columns of the table give intervals from standard errors clustered by state, which are wider.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

Dollar brackets are one way to describe income. A voter's rank in the year's income distribution gives a summary that does not depend on dollar cutoffs. The slope of the Democratic share on that rank, from the lowest rank to the highest, was −15.6 (95% CI −19.1 to −12.0) points in 2008 and +3.4 (95% CI −1.2 to 7.9) in 2024. The 2008 slope is clearly negative, and the 2024 slope has an interval that includes zero. A straight line summarizes a U poorly, so a single slope hides the shape that the brackets show.

The 2024 U shape within education groups

Degree holders make up a larger part of the highest bracket than of the lowest, and the difference has grown. In 2008 61.0% (95% CI 58.2% to 63.8%) of voters in the highest bracket held a four-year degree, against 12.0% (95% CI 10.7% to 13.4%) in the lowest bracket. In 2024 the shares were 74.5% (95% CI 72.8% to 76.2%) and 16.2% (95% CI 14.6% to 17.7%). Degree holders also vote more Democratic than other voters, by +4.2 (95% CI 2.7 to 5.8) points in 2008 and +12.6 (95% CI 11.0 to 14.1) in 2024 among all voters. A comparison of the highest and lowest brackets therefore mixes a difference in income with a difference in education.

Figure 3 draws the profile inside each education group. Among degree holders in 2024 the lowest and highest brackets stand at 60.4% (95% CI 56.1% to 64.8%) and 60.8% (95% CI 58.7% to 62.9%), against 54.2% (95% CI 52.2% to 56.3%) in the third bracket. Among voters without a degree they stand at 49.7% (95% CI 47.4% to 52.1%) and 46.5% (95% CI 42.2% to 50.7%), against 41.2% (95% CI 39.1% to 43.4%) in the third. Against the pooled middle three brackets, the lowest bracket differs by +5.4 (95% CI 0.8 to 9.9) points and the highest by +5.7 (95% CI 3.3 to 8.2) among degree holders, and both intervals are above zero. Among voters without a degree the lowest bracket differs from the pooled middle by +7.7 (95% CI 5.0 to 10.4) points, an interval above zero, and the highest by +4.4 (95% CI 0.0 to 8.9), an interval that reaches zero. The two highest-bracket differences are close and their intervals overlap, so the data do not show that the right side of the U is weaker among voters without a degree. The shape appears inside each education group, so mixing two groups with different average votes does not produce it.

Figure 3

Among all voters in 2024 the lowest bracket is more Democratic than the middle three in both education groups, and the highest is too among degree holders

Democratic share of the two-party presidential vote by household-income bracket and education, 2024 by default; two-party voters who reported income; 95% intervals from the Kish effective sample size

Four-year degree or more

No four-year degree

Democratic share of the two-party vote (%)

Show data

Data for Among all voters in 2024 the lowest bracket is more Democratic than the middle three in both education groups, and the highest is too among degree holders
Sample	Election	Education	Income bracket	Democratic share (%)	95% CI low	95% CI high	n voters
All voters	2024	Four-year degree or more	$150k or more	60.8	58.7	62.9	3,569
All voters	2024	Four-year degree or more	$100k to $150k	54.8	52.7	57.0	3,769
All voters	2024	Four-year degree or more	$60k to $100k	54.2	52.2	56.3	4,160
All voters	2024	Four-year degree or more	$30k to $60k	56.8	54.1	59.5	2,755
All voters	2024	Four-year degree or more	Under $30k	60.4	56.1	64.8	1,275
All voters	2024	No four-year degree	$150k or more	46.5	42.2	50.7	1,263
All voters	2024	No four-year degree	$100k to $150k	40.9	38.0	43.9	2,647
All voters	2024	No four-year degree	$60k to $100k	41.2	39.1	43.4	5,486
All voters	2024	No four-year degree	$30k to $60k	43.1	41.2	45.0	7,495
All voters	2024	No four-year degree	Under $30k	49.7	47.4	52.1	6,174
All voters	2020	Four-year degree or more	$150k or more	63.9	62.0	65.9	2,965
All voters	2020	Four-year degree or more	$100k to $150k	61.1	59.3	62.9	3,883
All voters	2020	Four-year degree or more	$60k to $100k	61.6	60.0	63.1	5,066
All voters	2020	Four-year degree or more	$30k to $60k	63.0	61.0	65.0	3,567
All voters	2020	Four-year degree or more	Under $30k	67.7	64.8	70.5	1,526
All voters	2020	No four-year degree	$150k or more	46.4	42.4	50.3	943
All voters	2020	No four-year degree	$100k to $150k	42.6	40.0	45.2	2,179
All voters	2020	No four-year degree	$60k to $100k	41.2	39.5	42.9	5,468
All voters	2020	No four-year degree	$30k to $60k	47.1	45.6	48.6	7,716
All voters	2020	No four-year degree	Under $30k	54.5	52.7	56.4	6,341
All voters	2016	Four-year degree or more	$150k or more	62.4	60.0	64.8	2,003
All voters	2016	Four-year degree or more	$100k to $150k	59.5	57.5	61.4	3,342
All voters	2016	Four-year degree or more	$60k to $100k	60.6	59.0	62.3	4,723
All voters	2016	Four-year degree or more	$30k to $60k	60.6	58.6	62.5	3,378
All voters	2016	Four-year degree or more	Under $30k	62.6	59.4	65.9	1,313
All voters	2016	No four-year degree	$150k or more	51.0	46.4	55.6	736
All voters	2016	No four-year degree	$100k to $150k	45.4	42.6	48.1	2,086
All voters	2016	No four-year degree	$60k to $100k	44.1	42.3	45.8	5,366
All voters	2016	No four-year degree	$30k to $60k	46.0	44.5	47.5	7,913
All voters	2016	No four-year degree	Under $30k	54.1	52.2	55.9	5,712
All voters	2012	Four-year degree or more	$150k or more	59.8	56.4	63.3	1,607
All voters	2012	Four-year degree or more	$100k to $150k	56.1	53.3	59.0	2,663
All voters	2012	Four-year degree or more	$60k to $100k	55.3	53.1	57.6	4,054
All voters	2012	Four-year degree or more	$30k to $60k	59.3	56.8	61.8	3,359
All voters	2012	Four-year degree or more	Under $30k	63.7	59.8	67.5	1,401
All voters	2012	No four-year degree	$150k or more	49.7	43.7	55.7	599
All voters	2012	No four-year degree	$100k to $150k	47.3	43.9	50.7	1,637
All voters	2012	No four-year degree	$60k to $100k	47.9	45.8	50.0	4,632
All voters	2012	No four-year degree	$30k to $60k	52.1	50.4	53.7	7,600
All voters	2012	No four-year degree	Under $30k	58.9	57.0	60.8	5,471
All voters	2008	Four-year degree or more	$150k or more	55.7	52.6	58.8	1,094
All voters	2008	Four-year degree or more	$100k to $150k	54.0	51.5	56.5	1,709
All voters	2008	Four-year degree or more	$60k to $100k	56.6	54.5	58.7	2,425
All voters	2008	Four-year degree or more	$30k to $60k	58.0	55.5	60.4	1,768
All voters	2008	Four-year degree or more	Under $30k	68.2	64.2	72.2	595
All voters	2008	No four-year degree	$150k or more	49.4	43.9	54.9	540
All voters	2008	No four-year degree	$100k to $150k	46.1	43.1	49.1	1,535
All voters	2008	No four-year degree	$60k to $100k	46.9	45.1	48.8	3,836
All voters	2008	No four-year degree	$30k to $60k	52.9	51.3	54.6	5,404
All voters	2008	No four-year degree	Under $30k	61.6	59.5	63.8	3,305
White voters	2024	Four-year degree or more	$150k or more	58.1	55.8	60.5	2,774
White voters	2024	Four-year degree or more	$100k to $150k	51.3	48.9	53.7	2,979
White voters	2024	Four-year degree or more	$60k to $100k	51.0	48.7	53.3	3,143
White voters	2024	Four-year degree or more	$30k to $60k	53.1	50.0	56.2	1,985
White voters	2024	Four-year degree or more	Under $30k	55.1	49.8	60.5	854
White voters	2024	No four-year degree	$150k or more	43.4	38.7	48.1	1,004
White voters	2024	No four-year degree	$100k to $150k	35.4	32.4	38.4	2,134
White voters	2024	No four-year degree	$60k to $100k	35.1	32.9	37.3	4,345
White voters	2024	No four-year degree	$30k to $60k	34.9	33.0	36.9	5,709
White voters	2024	No four-year degree	Under $30k	39.5	36.7	42.3	4,344
White voters	2020	Four-year degree or more	$150k or more	61.3	59.2	63.5	2,291
White voters	2020	Four-year degree or more	$100k to $150k	58.0	56.0	60.0	2,975
White voters	2020	Four-year degree or more	$60k to $100k	57.7	55.9	59.5	3,790
White voters	2020	Four-year degree or more	$30k to $60k	57.3	55.1	59.6	2,570
White voters	2020	Four-year degree or more	Under $30k	63.4	60.0	66.7	1,057
White voters	2020	No four-year degree	$150k or more	39.6	35.4	43.7	727
White voters	2020	No four-year degree	$100k to $150k	36.5	33.8	39.1	1,724
White voters	2020	No four-year degree	$60k to $100k	33.0	31.3	34.6	4,293
White voters	2020	No four-year degree	$30k to $60k	35.0	33.5	36.5	5,780
White voters	2020	No four-year degree	Under $30k	38.9	37.0	40.8	4,432
White voters	2016	Four-year degree or more	$150k or more	59.7	56.9	62.5	1,466
White voters	2016	Four-year degree or more	$100k to $150k	56.7	54.6	58.9	2,524
White voters	2016	Four-year degree or more	$60k to $100k	57.1	55.2	59.0	3,463
White voters	2016	Four-year degree or more	$30k to $60k	56.2	53.9	58.4	2,402
White voters	2016	Four-year degree or more	Under $30k	58.1	54.2	62.0	867
White voters	2016	No four-year degree	$150k or more	45.6	40.5	50.8	569
White voters	2016	No four-year degree	$100k to $150k	38.0	35.1	41.0	1,625
White voters	2016	No four-year degree	$60k to $100k	37.3	35.4	39.1	4,236
White voters	2016	No four-year degree	$30k to $60k	36.6	35.0	38.1	6,080
White voters	2016	No four-year degree	Under $30k	39.7	37.7	41.8	4,128
White voters	2012	Four-year degree or more	$150k or more	57.1	53.4	60.9	1,357
White voters	2012	Four-year degree or more	$100k to $150k	53.4	50.2	56.5	2,172
White voters	2012	Four-year degree or more	$60k to $100k	52.3	49.8	54.8	3,310
White voters	2012	Four-year degree or more	$30k to $60k	54.7	51.8	57.6	2,625
White voters	2012	Four-year degree or more	Under $30k	60.6	56.1	65.1	1,061
White voters	2012	No four-year degree	$150k or more	47.0	40.2	53.8	492
White voters	2012	No four-year degree	$100k to $150k	42.7	38.9	46.4	1,354
White voters	2012	No four-year degree	$60k to $100k	41.6	39.4	43.8	3,854
White voters	2012	No four-year degree	$30k to $60k	44.5	42.7	46.3	6,074
White voters	2012	No four-year degree	Under $30k	47.9	45.6	50.1	4,077
White voters	2008	Four-year degree or more	$150k or more	53.2	50.0	56.5	978
White voters	2008	Four-year degree or more	$100k to $150k	50.4	47.8	53.0	1,514
White voters	2008	Four-year degree or more	$60k to $100k	52.9	50.7	55.1	2,116
White voters	2008	Four-year degree or more	$30k to $60k	52.9	50.3	55.5	1,504
White voters	2008	Four-year degree or more	Under $30k	64.5	60.2	68.9	514
White voters	2008	No four-year degree	$150k or more	43.9	38.1	49.8	462
White voters	2008	No four-year degree	$100k to $150k	39.6	36.4	42.8	1,307
White voters	2008	No four-year degree	$60k to $100k	40.3	38.3	42.2	3,242
White voters	2008	No four-year degree	$30k to $60k	42.5	40.8	44.2	4,446
White voters	2008	No four-year degree	Under $30k	50.3	47.9	52.7	2,643

In both education groups the lowest bracket stands to the right of the middle brackets. The highest bracket does too among degree holders. Among voters without a degree its difference from the pooled middle brackets is +4.4 points and its interval reaches zero, close to the +5.7 points among degree holders. Every bracket is more Democratic among degree holders than among other voters. A dot further right is more Democratic, and whiskers are confidence intervals. The buttons switch the election and restrict the sample to white voters, and these statements are about the default view only.Note: Brackets are in the nominal dollars of each survey year; the lowest and highest brackets hold a different share of voters in different years.Note: Intervals are model based (Kish effective sample size, normal approximation) and ignore clustering in the sample design.Note: Cells with fewer than 100 respondents or a Kish effective sample below 100 are not shown; none met that condition.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

The button labeled 2024, white voters in Figure 3 restricts the sample to white voters. Among white voters without a degree the highest bracket voted 43.4% (95% CI 38.7% to 48.1%) Democratic in 2024, against 35.1% (95% CI 32.9% to 37.3%) in the third bracket, so income still separates these voters at the top of the range. In this view the highest bracket differs from the pooled middle brackets by +6.5 (95% CI 3.8 to 9.3) points among degree holders and +8.3 (95% CI 3.4 to 13.2) among the others, and both intervals are above zero. The lowest bracket differs by +4.4 (95% CI 1.4 to 7.5) points among white voters without a degree and by +3.5 (95% CI −2.0 to 9.1) among white degree holders, an interval that includes zero, so the left side of the U is clear in one white group and not in the other.

Net of education, the 2024 gap includes zero

Figure 4 gives the highest-minus-lowest gap for all voters, within each education group, and net of education, which compares voters with the same education and comes from a linear probability model. The model uses four education levels, so its coefficient is not an average of the two within-group rows. Among voters without a degree the gap moved from −12.2 (95% CI −17.4 to −7.0) in 2008 to −3.3 (95% CI −9.3 to 2.7) in 2024, and among degree holders from −12.5 (95% CI −19.8 to −5.1) to +0.4 (95% CI −5.2 to 6.0). Both 2024 intervals include zero. Net of education the gap is −13.7 (95% CI −17.6 to −9.8) in 2008 and −4.5 (95% CI −9.1 to 0.1) in 2024, and net of education together with race and ethnicity the 2024 gap is −1.9 (95% CI −6.2 to 2.3).

Figure 4

Net of education the 2024 gap is −4.5 points against +5.7 for all voters, and its interval includes zero

Democratic share in the highest bracket minus the lowest, percentage points, 2008 and 2024, for all voters, within education groups, and net of education and race and ethnicity; 95% intervals clustered by state

2008
2024

Show data

Data for Net of education the 2024 gap is −4.5 points against +5.7 for all voters, and its interval includes zero
Comparison	Gap 2008 (points)	95% CI low 2008	95% CI high 2008	n 2008	Gap 2024 (points)	95% CI low 2024	95% CI high 2024	n 2024
All voters	−9.1	−12.9	−5.4	5,534	+5.7	1.1	10.3	12,281
No degree	−12.2	−17.4	−7.0	3,845	−3.3	−9.3	2.7	7,437
Degree	−12.5	−19.8	−5.1	1,689	+0.4	−5.2	6.0	4,844
Net of education	−13.7	−17.6	−9.8	5,534	−4.5	−9.1	0.1	12,281
Net of education and race	−9.7	−13.2	−6.2	5,534	−1.9	−6.2	2.3	12,281

Within education groups and net of education the 2024 gap is nearer zero than for all voters, and each of those four 2024 intervals includes zero, while the gap still rises in every row. The ring is 2008 and the dot is 2024, and a value to the right of zero means the highest bracket is more Democratic than the lowest. Degree means a four-year degree or more. The last two rows are coefficients from a weighted linear probability model that adds education (four levels), or education and race and ethnicity (five groups), to the two end brackets, so they are not averages of the rows above them.Note: Model rows use only the highest and lowest brackets. Education has four levels: high school or less, some college, a four-year degree, and postgraduate study.Note: Intervals use standard errors clustered by state (the states and the District of Columbia) and a normal approximation; they are model based.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

The gap still rises. The change is +8.9 (95% CI 2.2 to 15.6) points among voters without a degree, +12.9 (95% CI 3.9 to 21.8) among degree holders, +9.2 (95% CI 4.9 to 13.4) net of education and +7.8 (95% CI 3.7 to 11.9) net of education with race and ethnicity, against +14.8 (95% CI 10.8 to 18.9) for all voters. Holding education and race fixed removes part of the change in the gap and leaves the rest. The positive 2024 gap for all voters combines a difference in income with a difference in education. This is a description of shared variation and not a causal decomposition, since income, education and race are each tied to where people live, what work they do and whom they know.

A study of Western democracies from 1948 to 2020 found education and income separating: by the 2010s higher-educated voters voted for the left while high-income voters continued to vote for the right [3]. In the 2024 CES, net of education, the highest bracket is still more Democratic than the middle three brackets, by +4.1 (95% CI 1.6 to 6.6) points, against +8.6 (95% CI 5.6 to 11.6) for the lowest bracket. Net of education the profile remains U-shaped, with both end brackets above the middle. The comparison does not test the study's claims, which cover other countries, measures and years. A book-length account of the education divide in American politics is [4], and Chapter 3.2, The diploma divide takes up the issues on which graduates and non-graduates differ. Chapter 7.3, Class, church and the two axes (planned) sets income and education against church attendance on the economic and cultural dimensions.

Among high-income voters, the education gap widened through 2020

The highest bracket is now a larger part of the electorate, so its composition matters. Figure 5 splits it by education. Among voters in the highest bracket with a four-year degree, the Democratic share was 55.7% (95% CI 52.6% to 58.8%) in 2008, 63.9% (95% CI 62.0% to 65.9%) in 2020 and 60.8% (95% CI 58.7% to 62.9%) in 2024, and it changed by −3.1 (95% CI −6.0 to −0.3) points from 2020 to 2024. Among voters in the highest bracket without a degree the shares were 49.4% (95% CI 43.9% to 54.9%), 46.4% (95% CI 42.4% to 50.3%) and 46.5% (95% CI 42.2% to 50.7%), with intervals that overlap. The gap between the two groups was +6.3 (95% CI 0.0 to 12.6) points in 2008, +17.6 (95% CI 13.2 to 22.0) in 2020 and +14.4 (95% CI 9.6 to 19.1) in 2024. It changed by +11.3 (95% CI 3.6 to 19.0) points from 2008 to 2020 and by −3.2 (95% CI −9.7 to 3.3) from 2020 to 2024, an interval that includes zero.

Figure 5

In the highest bracket, degree holders were 55.7% Democratic in 2008 and 63.9% in 2020, and voters without a degree were 49.4% and 46.4%

Democratic share of the two-party presidential vote among voters in the highest household-income bracket, by education, 2008 to 2024; 95% intervals from the Kish effective sample size

Democratic share of the two-party vote (%)

Show data

Data for In the highest bracket, degree holders were 55.7% Democratic in 2008 and 63.9% in 2020, and voters without a degree were 49.4% and 46.4%
Education	Election	Democratic share (%)	95% CI low	95% CI high	Gap, degree minus no degree (points)	Gap 95% CI low	Gap 95% CI high	n voters
Degree	2008	55.7	52.6	58.8	+6.3	0.0	12.6	1,094
No degree	2008	49.4	43.9	54.9	+6.3	0.0	12.6	540
Degree	2012	59.8	56.4	63.3	+10.2	3.2	17.1	1,607
No degree	2012	49.7	43.7	55.7	+10.2	3.2	17.1	599
Degree	2016	62.4	60.0	64.8	+11.3	6.1	16.5	2,003
No degree	2016	51.0	46.4	55.6	+11.3	6.1	16.5	736
Degree	2020	63.9	62.0	65.9	+17.6	13.2	22.0	2,965
No degree	2020	46.4	42.4	50.3	+17.6	13.2	22.0	943
Degree	2024	60.8	58.7	62.9	+14.4	9.6	19.1	3,569
No degree	2024	46.5	42.2	50.7	+14.4	9.6	19.1	1,263

The two bands overlap in 2008 and are apart from 2012 on. The share among degree holders rises to 2020 and then changes by −3.1 points to 2024; the share among voters without a degree stays within sampling error of its 2008 level in every later election. The gap between the two groups is +6.3 points in 2008, +17.6 in 2020 and +14.4 in 2024; the 2008 interval reaches zero, and the change from 2020 to 2024, −3.2 points, is within sampling error. Degree means a four-year degree or more, and the highest bracket is the top of the five used throughout the chapter. Bands are confidence intervals.Note: Brackets are in the nominal dollars of each survey year; the lowest and highest brackets hold a different share of voters in different years.Note: Intervals are model based (Kish effective sample size, normal approximation) and ignore clustering in the sample design.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

The gap is lower in 2024 than in 2020 by an amount within sampling error. The share among degree holders was lower in 2024, and the share among voters without a degree changed by +0.1 (95% CI −5.7 to 5.9) points over the same years, so the decline among degree holders accounts for the difference in the point estimates. Voters without a degree in the highest bracket stayed within sampling error of their 2008 level in every later election, including 2024, when the change from 2008 was −3.0 (95% CI −9.9 to 4.0) points. One election cannot say whether the 2024 decline among degree holders lasts.

The timing of the crossing depends on the weight

The crossing of the highest-minus-lowest gap cannot be assigned to one election. The vote question is asked after the election, and the CES file carries several weights. The cumulative-file weight used in the rest of the chapter is defined for all five elections. A post-election weight, built for the respondents who completed the post-election interview, exists for 2012, 2016 and 2020, and a validated-vote weight, for respondents whose turnout was checked against voter records, exists for 2020 and 2024. Neither exists for 2008. Figure 6 plots the gap by election under each weight, with intervals clustered by state.

With the cumulative-file weight the gap is −3.4 (95% CI −7.6 to 0.8) in 2012, +3.4 (95% CI −0.1 to 6.9) in 2016 and +2.6 (95% CI −1.0 to 6.2) in 2020, and each interval includes zero, so only 2008 and 2024 differ clearly from zero. With the post-election weight the gap is −6.9 (95% CI −10.9 to −3.0) in 2012, an interval below zero, and +2.0 (95% CI −1.7 to 5.6) in 2016 and +0.7 (95% CI −3.1 to 4.6) in 2020, whose intervals include zero. The 2024 gap is above zero under both weights defined for it, at +5.7 (95% CI 1.1 to 10.3) with the cumulative-file weight and +4.2 (95% CI 0.2 to 8.2) with the validated-vote weight.

Figure 6

Under every weight the 2024 gap is above zero and the 2016 and 2020 gaps include zero

Democratic share of the two-party presidential vote in the highest bracket minus the lowest, percentage points, by election and survey weight; two-party voters who reported income; 95% intervals clustered by state

Highest minus lowest bracket (percentage points)

Show data

Data for Under every weight the 2024 gap is above zero and the 2016 and 2020 gaps include zero
Election	Weight	Gap (points)	95% CI low (clustered by state)	95% CI high (clustered by state)	95% CI low (Kish)	95% CI high (Kish)	n voters
2008	Cumulative-file weight	−9.1	−12.9	−5.4	−12.6	−5.7	5,534
2012	Cumulative-file weight	−3.4	−7.6	0.8	−7.0	0.1	9,078
2016	Cumulative-file weight	+3.4	−0.1	6.9	0.5	6.2	9,764
2020	Cumulative-file weight	+2.6	−1.0	6.2	0.2	5.0	11,775
2024	Cumulative-file weight	+5.7	1.1	10.3	2.8	8.5	12,281
2012	Post-election weight	−6.9	−10.9	−3.0	−10.5	−3.4	9,078
2016	Post-election weight	+2.0	−1.7	5.6	−1.1	5.1	9,764
2020	Post-election weight	+0.7	−3.1	4.6	−1.9	3.4	11,775
2020	Validated-vote weight	+2.8	−1.0	6.7	0.2	5.4	9,585
2024	Validated-vote weight	+4.2	0.2	8.2	1.3	7.1	9,405

The 2024 gap is above zero under both weights defined for it, and the 2016 and 2020 intervals include zero under every weight. The post-election weight puts the 2012 gap below zero, at −6.9 points, where the cumulative-file interval includes zero, and moves the 2016 and 2020 estimates nearer zero. The 2020 point estimate is above zero under all three weights, from +0.7 to +2.8 points. The cumulative-file weight covers all five elections, the post-election weight 2012, 2016 and 2020, and the validated-vote weight 2020 and 2024. Bands are confidence intervals.Note: The post-election weight is not available for 2008 or 2024, and the validated-vote weight only for 2020 and 2024. Intervals use standard errors clustered by state (the states and the District of Columbia) and a normal approximation; they are model based.Note: The table also gives the interval from the Kish effective sample size, which is narrower; under the cumulative-file weight it is above zero for both 2016 and 2020.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative, weight_post or vvweight, as labelled); two-party presidential voters who reported household income

The change in the gap since 2008 is above zero by 2016 under the cumulative-file weight, at +12.5 (95% CI 8.0 to 17.0) points, and under the post-election weight, at +11.1 (95% CI 6.4 to 15.8). By 2012 it is above zero only under the cumulative-file weight, at +5.7 (95% CI 1.0 to 10.4) points against +2.2 (95% CI −2.5 to 6.9) with the post-election weight. Under the validated-vote weight the change from 2008 to 2024 is +13.3 (95% CI 8.9 to 17.8) points. The data therefore do not place the start of the movement before 2016. The 2008 and 2024 gaps keep their signs under every weight available for them.

The change holds under five other definitions

The change in the highest-minus-lowest gap between 2008 and 2024, a difference in differences (the change in the gap between two elections), is +14.8 (95% CI 10.8 to 18.9) points, with a state-clustered standard error of 2.1 points, and it is +11.4 (95% CI 8.5 to 14.3) without the survey weights. Dollar brackets raise a further question. The highest bracket held 6.7% (95% CI 6.3% to 7.1%) of the voters who reported income in 2008 and 13.3% (95% CI 12.8% to 13.9%) in 2024, so the highest bracket in 2024 is a broader group than the one in 2008.

Figure 7 repeats the estimate under five other definitions, which its notes describe. All six intervals are above zero, and the change is largest, at +20.6 (95% CI 16.4 to 24.8) points, with both cutoffs held constant in real dollars, so nominal dollars do not account for it. The two real-dollar groups are not equal in size across years: the top group holds 19.8% of voters in 2008 and 13.3% in 2024, and the bottom group holds 10.8% and 17.8%.

Figure 7

The gap between higher and lower incomes rose under all six definitions, by +10.2 to +20.6 points in the five comparisons of two groups

Change from 2008 to 2024 in the Democratic share of the higher-income group minus that of the lower-income group, percentage points; two-party voters who reported income; 95% intervals clustered by state

Two income groups
Different scale

Show data

Data for The gap between higher and lower incomes rose under all six definitions, by +10.2 to +20.6 points in the five comparisons of two groups
Estimate	Change (points)	95% CI low	95% CI high	n observations
Main: highest vs lowest	+14.8	10.8	18.9	17,815
Unweighted	+11.4	8.5	14.3	17,815
Top 2 vs bottom 2	+10.2	7.9	12.4	44,897
Top group, fixed share	+16.3	12.6	20.0	19,225
Real-dollar cutoffs	+20.6	16.4	24.8	18,933
Rank slope (other scale)	+18.9	15.3	22.6	60,804

All six estimates are above zero. The five comparisons of two income groups range from +10.2 to +20.6 points, and the rank slope, on a different scale, is +18.9. The rows differ in how the income groups are defined, as the notes below explain. Whiskers are confidence intervals.Note: Each row is a difference in differences, the change in the gap between two elections, estimated by weighted least squares with standard errors clustered by state (the states and the District of Columbia).Note: Unweighted: the same estimate without the survey weights. Top 2 vs bottom 2: each group is two brackets.Note: Top group, fixed share: in each election the top group's income cutoff is set so that the group holds about the same share of voters as the highest bracket does in 2024; the comparison group is the lowest bracket.Note: Real-dollar cutoffs: both cutoffs are held constant in real dollars with the consumer price index, at the income band edges nearest the 2024 cutoffs deflated to each election.Note: Rank slope: the vote is regressed on each voter's income rank within the election, so no dollar cutoffs are used. The estimate is the change in the difference between the lowest and the highest rank, and it does not compare two groups of voters.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

The highest bracket made up 6.6% (95% CI 6.0% to 7.2%) of Democratic voters in 2008 and 15.4% (95% CI 14.6% to 16.2%) in 2024, and 6.9% (95% CI 6.3% to 7.4%) and 11.3% (95% CI 10.6% to 12.0%) of Republican voters, so the increase is larger among Democratic voters. Part of the increase comes from the growth of the bracket among all voters, so the comparison of the two parties is more informative than either series alone. Chapter 3.1, The two coalitions breaks each party's voters down by education, income, race, religion, age and gender, and Chapter 3.4, Hispanic, Black and Asian voters (planned) follows the vote of three racial and ethnic groups.

Figure 8

The highest bracket's share of Democratic voters rose from 6.6% in 2008 to 15.4% in 2024, and of Republican voters from 6.9% to 11.3%

Share of each party's two-party presidential voters who live in a household in the highest income bracket, percent, 2008 and 2024; voters who reported income; 95% intervals from the Kish effective sample size

2008
2024

Voters in the highest income bracket (% of the party's voters)

Show data

Data for The highest bracket's share of Democratic voters rose from 6.6% in 2008 to 15.4% in 2024, and of Republican voters from 6.9% to 11.3%
Party's voters	Share 2008 (%)	95% CI low 2008	95% CI high 2008	n 2008	Share 2024 (%)	95% CI low 2024	95% CI high 2024	n 2024
Democratic voters	6.6	6.0	7.2	10,942	15.4	14.6	16.2	21,814
Republican voters	6.9	6.3	7.4	11,269	11.3	10.6	12.0	16,779

The highest bracket's share rose for both parties, by more among Democratic voters. The ring is 2008 and the dot is 2024, and each end is the share of a party's voters in the highest bracket, with a confidence interval.Note: Brackets are in the nominal dollars of each survey year; the lowest and highest brackets hold a different share of voters in different years.Note: Intervals are model based (Kish effective sample size, normal approximation) and ignore clustering in the sample design.Source: Cooperative Election Study cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative); two-party presidential voters who reported household income

What these data cannot show

The CES interviews different respondents in each election, so the chapter compares groups and not people. It cannot say whether any voter changed their vote as income or education changed, or how much of the change comes from migration, aging and the entry of new voters. The vote is self-reported, and an online survey could lean toward one party by an amount that varies with income. The check available here is coarse. Figure 9 compares the CES national Democratic share with the certified count in each election. The largest difference is 1.1 (95% CI 0.3 to 1.8) points, in 2012 with the survey above the count, and in that election the certified share lies outside the survey interval. The difference is −0.3 (95% CI −1.0 to 0.5) points in 2024. With the post-election weight the 2012 difference is +0.3 (95% CI −0.4 to 1.1) points, so the 2012 discrepancy depends on the weight. A lean of constant size would leave the differences between brackets intact, but agreement at the national level cannot show that the survey is accurate inside income brackets.

Figure 9

In every election the survey's Democratic share is within 1.1 points of the certified national result

Democratic share of the two-party presidential vote, CES respondents minus the certified national count, percentage points, by election; two-party voters whether or not they reported income; 95% intervals from the Kish effective sample size

CES minus certified count, Democratic share (percentage points)

Show data

Data for In every election the survey's Democratic share is within 1.1 points of the certified national result
Election	CES (%)	Certified (%)	Difference (points)	95% CI low	95% CI high	n voters
2008	53.9	53.8	+0.1	−0.7	+0.9	23,585
2012	53.1	52.0	+1.1	+0.3	+1.8	37,789
2016	51.0	51.2	−0.2	−0.8	+0.5	40,891
2020	52.5	52.3	+0.3	−0.4	+0.9	43,872
2024	49.0	49.2	−0.3	−1.0	+0.5	42,008

The largest difference is 1.1 points, in 2012, where the survey is above the count and the count lies outside the survey interval; in the other four elections the count lies inside it. Each dot is the CES Democratic share minus the certified national two-party share, with the survey's confidence interval. The CES share covers all two-party voters, whether or not they reported income.Note: The certified share is a full count and has no sampling interval; the interval shown is the survey's.Source: Cooperative Election Study cumulative file, 2008-2024; weighted (weight_cumulative). MIT Election Data and Science Lab state presidential returns, 1976-2024; full counts (unweighted)

Income is household income in bands and is missing for 6.3% of two-party voters in 2008 and 11.2% in 2012, and for 8.2% in 2024. The Democratic share among voters who skipped the income question minus the share among those who answered was −6.5 (95% CI −9.3 to −3.6) points in 2024, so the ladder leaves out a group whose vote differs. Brackets are nominal dollars for households of any size. The chapter has no data on wealth, occupation or individual earnings.

The design cannot separate income from the things that go with it. Education and race are the only two adjustments, and place of residence, occupation, homeownership and age are not held fixed. The chapter describes where the Democratic share is higher and how that changed, and it says nothing about why, since the file holds no measures of candidates, campaigns or policy positions. Five elections are a short series, and the 2024 result for degree holders in the highest bracket is one observation.

Income relates to the vote differently in different places. An analysis of survey and state data found that the association between income and Republican voting was stronger in poorer states than in richer ones [5], and this chapter pools voters across states without estimating the ladder for each. A pattern across places also need not match a pattern across households, which is the problem of ecological inference. Chapter 12.3, Income and the county vote (planned) compares counties with different incomes once education is held constant, and Chapter 12.4, Wages, capital and the ZIP code vote looks at wage and capital income in ZIP codes.

Data sources

Cooperative Election StudyCumulative Common Content, presidential years 2008, 2012, 2016, 2020 and 2024Years: 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.
State presidential returnsState presidential returns 1976-2024, used only for the certified national two-party shareYears: 1976-2024 (13 elections) · Weights: none (full counts) · License: See the provider's terms of use.

All data sources

Methods

Sample: respondents in the Cooperative Election Study cumulative file for the five presidential years 2008, 2012, 2016, 2020 and 2024 who reported voting for the Democratic or the Republican presidential candidate (188,145 respondents), of whom 170,053 also reported household income. Outcome: the Democratic share of the two-party vote as reported by the respondent in the post-election interview; third-party and other votes are excluded. Weights: weight_cumulative, the cumulative-file weight for pooled years, in every estimate except the row marked unweighted and the weight sensitivity figure. Within each election it is a constant multiple of the single-wave weight, and it is the only weight defined for all five elections. A post-election weight (weight_post), built for the respondents who completed the post-election interview, exists for 2012, 2016 and 2020, and a validated-vote weight (vvweight), for respondents whose turnout was checked against voter records, exists for 2020 and 2024; neither exists for 2008. The weight sensitivity figure re-estimates the highest-minus-lowest gap with each of them and gives state-clustered intervals, and the chapter quotes the change in the gap under them where a conclusion depends on the weight. Income: the file's five brackets, built from twelve reported income bands, in nominal dollars (under $30,000; $30,000 to $60,000; $60,000 to $100,000; $100,000 to $150,000; $150,000 or more). High income means the highest bracket everywhere in the chapter. Education: a four-year degree or more against all other levels. Estimators: a share is a weighted mean of the 0/1 Democratic vote; a gap is the difference of two weighted means; a change in a gap between two elections is a difference in differences (the change in the gap between two elections) fitted by weighted least squares; the net-of-education rows, which compare voters with the same education, are coefficients from a weighted linear probability model on the two end brackets, and the contrasts against the middle three brackets use the same model on all five brackets. Intervals: 95%. Shares, contrasts between groups within one election (the differences against the pooled middle three brackets, the education differences among voters in the highest bracket and the contrasts inside education groups) and changes in one share between two elections use the Kish effective sample size and a normal approximation, and combine the standard errors of independent groups in quadrature. The highest-minus-lowest gap in each election, a change in a gap between two elections and every regression coefficient use standard errors clustered by state (51 clusters, the states and the District of Columbia) with normal critical values; a t critical value with 50 degrees of freedom (2.01 in place of 1.96) would widen each clustered interval by about 2.5% and would not move zero in or out of any of the 84 clustered intervals in the chapter. State-clustered intervals for the 25 bracket-by-election shares are 1.3 to 3.0 times as wide as the Kish intervals (median 2.1), because between-state differences in the vote count as sampling error, so the Kish intervals on levels leave out that variation. The tables under the figure of the two contrasts against the middle three and under the figure of the gap by weight give the state-clustered interval beside the Kish interval for every one-election estimate they show. All intervals are model based: the CES is an opt-in panel, so an interval treats the weighted sample as a random sample and ignores how respondents came to be in the panel. Suppression: a cell needs at least 100 respondents and a Kish effective sample of at least 100, and no cell was suppressed. Rounding: displays round half up from the unrounded estimate, and differences are computed before rounding, so a printed difference can differ by 0.1 from the difference of two printed shares. Income rank: the weighted midpoint percentile of the respondent's income band among all respondents who reported income in that election. Top group held near a constant share: in each election the income cutoff among the twelve bands whose group share of voters is closest to the 2024 share of the highest bracket (income of $120,000 or more in 2008); the comparison group is the lowest bracket. Constant real dollars: the 2024 cutoffs ($150,000 and $30,000) deflated to each election with the BLS CPI-U annual average (series CUUR0000SA0; the 2024 price level is 1.457 times the 2008 level) and moved to the nearest band edge ($100,000 and $20,000 in 2008); the top group is the bands at or above the top edge and the bottom group the bands at or below the bottom edge. Benchmark: the certified national two-party share from the state presidential returns of the MIT Election Data and Science Lab, full counts. Survey mode: the CES is an online panel in every year, so there is no mode break in the series.

References

Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
Zacher, Sam (2024). Polarization of the Rich: The New Democratic Allegiance of Affluent Americans and the Politics of Redistribution. Perspectives on Politics 22(2): 338-356. doi:10.1017/S1537592722003310
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.
Gelman, Andrew; Shor, Boris; Bafumi, Joseph; Park, David (2008). Rich State, Poor State, Red State, Blue State: What's the Matter with Connecticut?. Quarterly Journal of Political Science 2(4): 345-367. doi:10.1561/100.00006026

How to cite

@misc{singh2026theincomeladder,
  author = {Singh, Vishal},
  title = {The income ladder},
  year = {2026},
  url = {https://vishalsingh.org/politics/the-income-ladder},
  note = {In The Public Mind. Published 30 September 2026.}
}

Same question at another grain

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

Income
- Chapter 12.3, Income and the county vote (planned) Places
- Chapter 12.4, Wages, capital and the ZIP code vote Places
- Chapter 8.7, System justification (planned) Countries
Donors and voters
- Chapter 14.6, Stated and funded ideology (planned) Places
- Chapter 14.2, The geography of donations (planned) Places

The most Democratic bracket went from lowest to highest

By 2024 both ends voted more Democratic than the middle

The 2024 U shape within education groups

Net of education, the 2024 gap includes zero

Among high-income voters, the education gap widened through 2020

The timing of the crossing depends on the weight

The change holds under five other definitions

The highest bracket's share of each party's voters

What these data cannot show

Data sources

Methods

References

How to cite

Same question at another grain

Reading paths