The income ladder

In the 2008 Cooperative Election Study (CES), 62.4%62.4% percent of the two-party presidential voteEstimandweighted Democratic share of the two-party voteInterval95% CI 60.5% to 64.4%n3,900Weightsweight_cumulativePopulationthe lowest bracket (under $30,000) among two-party presidential voters who reported household incomePeriod2008SourceCooperative Election Study of voters with a household income under $30,000$30,000 dollars of annual household incomeEstimanddefinition of a bracket edgeWeightsnone (definition)Populationedge of an income bracketPeriod2008, 2012, 2016, 2020 and 2024SourceCooperative Election Study reported a Democratic rather than a Republican presidential vote, against 53.3%53.3% percent of the two-party presidential voteEstimandweighted Democratic share of the two-party voteInterval95% CI 50.4% to 56.1%n1,634Weightsweight_cumulativePopulationthe highest bracket ($150,000 or more) among two-party presidential voters who reported household incomePeriod2008SourceCooperative Election Study of voters with a household income of $150,000$150,000 dollars of annual household incomeEstimanddefinition of a bracket edgeWeightsnone (definition)Populationedge of an income bracketPeriod2008, 2012, 2016, 2020 and 2024SourceCooperative Election Study or more. The lowest of the survey's five income brackets was the most Democratic of the five.Every share in this essay is of the two-party vote, so voters who reported a third-party vote are left out, and every estimate uses the file's survey weightSurvey weightsNumbers that scale each respondent so the sample matches the population on known traits such as age, race, education and region. A weighted average counts a respondent with a large weight for more than one with a small weight.. Brackets are in the nominal dollars of each survey year. By 2024 the highest bracket was the most Democratic, and the ladder, on which the Democratic share fell as income rose, had become a U.

The two end brackets traded places

The chapter's measure is the gap between the ends: the Democratic share of the two-party vote in the highest bracket minus the share in the lowest, in pointsPercentage pointThe unit for the difference between two percentages. A change from one share to another is stated in points, not in percent, so that absolute and relative change are not confused.. In 2024 the Democratic share was 51.5%51.5% percent of the two-party presidential voteEstimandweighted Democratic share of the two-party voteInterval95% CI 49.4% to 53.6%n7,449Weightsweight_cumulativePopulationthe lowest bracket (under $30,000) among two-party presidential voters who reported household incomePeriod2024SourceCooperative Election Study in the lowest bracket and 57.2%57.2% percent of the two-party presidential voteEstimandweighted Democratic share of the two-party voteInterval95% CI 55.2% to 59.1%n4,832Weightsweight_cumulativePopulationthe highest bracket ($150,000 or more) among two-party presidential voters who reported household incomePeriod2024SourceCooperative Election Study in the highest. The gap was −9.1−9.1 percentage points of Democratic two-party vote shareEstimandDemocratic share in the highest bracket minus that in the lowest; weighted regression on the two brackets, state-clustered standard errorsInterval95% CI −12.9 to −5.4n5,534Weightsweight_cumulativePopulationhighest minus lowest bracket, two-party presidential voters who reported household incomePeriod2008SourceCooperative Election Study points in 2008, negative because the lowest bracket was the more Democratic, and +5.7+5.7 percentage points of Democratic two-party vote shareEstimandDemocratic share in the highest bracket minus that in the lowest; weighted regression on the two brackets, state-clustered standard errorsInterval95% CI 1.1 to 10.3n12,281Weightsweight_cumulativePopulationhighest minus lowest bracket, two-party presidential voters who reported household incomePeriod2024SourceCooperative Election Study in 2024. That is a change of +14.8+14.8 percentage points of Democratic two-party vote shareEstimandchange in the highest-minus-lowest gap (difference in differences), weighted, state-clustered standard errorsInterval95% CI 10.8 to 18.9n17,815Weightsweight_cumulativePopulationhighest and lowest brackets, two-party presidential voters who reported household incomePeriod2008 to 2024SourceCooperative Election Study points, larger than sampling error.

Figure 1 compares the two elections bracket by bracket. In 2008 the Democratic share fell with income through the first four brackets and then rose in the highest. The rise at the top was small, and the highest bracket's edge over the three middle brackets pooled had an interval that includes zero. Between 2008 and 2024 the three lowest brackets moved toward the Republican candidate, the lowest furthest. The fourth bracket did not move by more than sampling error, and the highest bracket moved toward the Democratic candidate. Among all two-party voters in the survey, whether or not they reported income, the Democratic share changed by −4.9−4.9 percentage points of Democratic two-party vote shareEstimandCES weighted Democratic share in 2024 minus that in 2008Interval95% CI −6.0 to −3.9Weightsweight_cumulativePopulationall two-party presidential voters (income reported or not)Period2008 to 2024SourceCooperative Election Study points over the same years. The highest bracket therefore moved toward the Democratic candidate while the survey's voters as a whole moved away.

The 2024 profile is U-shaped rather than a ladder turned upside down. A ladder turned upside down would have the Democratic share rise steadily with income, and in 2024 it does not, because the lowest bracket is more Democratic than the three middle brackets pooled. The highest bracket is more Democratic than the pooled middle as well, and both of those differences have intervals above zero. The U rests on the pooled contrasts and not on every bracket, since the fourth bracket taken alone is not distinguishable from the lowest. The reversal applies only to the comparison of the two end brackets.

One online survey, five elections

The CES is an online panel survey fielded by YouGov around each election, and the vote question is asked after the election. The chapter uses the presidential years 2008, 2012, 2016, 2020 and 2024. Across the five elections about 170,000170,053 respondentsEstimandcount of respondentsWeightsnone (count)Populationtwo-party presidential voters who reported household incomePeriod2008, 2012, 2016, 2020 and 2024SourceCooperative Election Study respondents reported both a vote for one of the two major-party candidates and their household income, and every share by bracket is computed among them. The file groups the twelve income bands of the questionnaire into five brackets in nominal dollars, so the highest bracket in 2024 is a broader group than the one in 2008. The survey's national Democratic share differs from the certified count by at most 1.11.1 percentage points of Democratic two-party vote shareEstimandlargest absolute difference between the CES weighted Democratic share and the certified national two-party share in the five electionsInterval95% CI 0.3 to 1.8n37,789Weightsweight_cumulativePopulationall two-party presidential votersPeriod2012SourceCooperative Election StudyNoteThe largest difference is in 2012. points, in 2012 with the survey above the count, and that difference depends on the weight. Agreement at the national level cannot show that the survey is accurate inside income brackets. The file carries several weights, and the 2008 and 2024 gaps keep their signs under every weight available for them, while the timing of the crossing between those years depends on the weight. The report gives every estimate with its interval and sample size.

The U shape appears inside each education group

Degree holders make up a larger part of the highest bracket than of the lowest, the difference has grown since 2008, and degree holders vote more Democratic than other voters. A comparison of the highest and lowest brackets therefore mixes a difference in income with a difference in education. Within education groups the 2024 gap is −3.3 (95% CI −9.3 to 2.7)−3.3 percentage points of Democratic two-party vote shareEstimandweighted difference in Democratic share, state-clustered standard errorsInterval95% CI −9.3 to 2.7n7,437Weightsweight_cumulativePopulationhighest minus lowest bracket, voters without a four-year degree in those bracketsPeriod2024SourceCooperative Election Study points among voters without a four-year degree and +0.4 (95% CI −5.2 to 6.0)+0.4 percentage points of Democratic two-party vote shareEstimandweighted difference in Democratic share, state-clustered standard errorsInterval95% CI −5.2 to 6.0n4,844Weightsweight_cumulativePopulationhighest minus lowest bracket, voters with a four-year degree or more in those bracketsPeriod2024SourceCooperative Election Study among degree holders. Both intervals include zero. An interval that includes zero means the data cannot tell the within-group gap from zero. It does not mean the gap is zero. The reversal between the ends therefore combines income and education.

Figure 2 draws the whole profile inside each education group rather than its two ends alone. Among degree holders in 2024 the lowest and highest brackets both stand to the right of the three middle brackets, and both differences from the pooled middle have intervals above zero. Among voters without a degree the lowest bracket also differs from the pooled middle by an interval above zero, while the highest bracket's difference is +4.4 (95% CI 0.0 to 8.9)+4.4 percentage points of Democratic two-party vote shareEstimandDemocratic share in the first group minus that in the pooled second group; independent groups, standard errors combinedInterval95% CI 0.0 to 8.9n16,891Weightsweight_cumulativePopulationhighest bracket minus the pooled middle three brackets, voters without a four-year degreePeriod2024SourceCooperative Election Study points, an interval that reaches zero. That difference is close to the one among degree holders and the two 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. Every bracket is more Democratic among degree holders than among other voters.

Holding education fixed brings the 2024 gap near zero

Figure 3 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 modelLinear probability modelA regression of a yes-or-no outcome on predictors, fitted by least squares so that each coefficient is a change in the probability of yes. The book weights it and reports robust standard errors.. Net of education the 2024 gap is −4.5 (95% CI −9.1 to 0.1)−4.5 percentage points of Democratic two-party vote shareEstimandweighted linear probability model coefficient for the highest bracket against the lowest, net of education (four levels); state-clustered standard errorsInterval95% CI −9.1 to 0.1n12,281Weightsweight_cumulativePopulationhighest minus lowest bracket, two-party presidential voters who reported household incomePeriod2024SourceCooperative Election Study points, and with race and ethnicity held fixed as well the interval also includes zero. This compares the two end brackets within each level of education, so the highest bracket's larger share of degree holders no longer counts toward the gap. In every row the 2024 gap is to the right of the 2008 gap, so the gap still rises when education is held fixed. Holding education and race fixed removes part of the change in the gap and leaves the rest, and net of education the profile remains U-shaped, with both end brackets above the middle. 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. Work that covers a longer period found that Democratic candidates began winning a growing share of affluent voters in the 1990s [1]. That trend was concentrated among, but not limited to, college-educated voters, and it was visible only in larger metropolitan areas. The CES covers the recent end of that history. Chapter 3.4, The diploma divide takes up the issues on which graduates and non-graduates differ.

What these data cannot tell you

The CES interviews different respondents in each election, so the chapter compares groups and not people, and it cannot say whether any voter changed their vote as income or education changed. Income is household income in bands, in nominal dollars, for households of any size. Voters who skipped the income question are left out of the ladder, and their vote differs from that of voters who answered. The design cannot separate income from the things that go with it, since place of residence, occupation, homeownership and age are not held fixed. The file holds no measures of candidates, campaigns or policy positions, so the chapter says nothing about why. Income relates to the vote differently in different places [2], and this chapter pools voters across states; Planned, Income and the county vote (planned) compares counties with different incomes once education is held constant.

Go deeper

• The report: every estimate with its interval and sample size.
• Tables: every figure of the report as a table.

Same question at another grain:

• Income
  • Planned, Income and the county vote (planned) Places
  • Companion 8.1b, Wages, capital and the ZIP code vote Places
  • Planned, System justification (planned) Countries
• Donors and voters
  • Planned, Stated and funded ideology (planned) Places
  • Planned, The geography of donations (planned) Places

Reading paths that include this chapter:

• The core argument in eight chapters
• A tour of the book, one chapter per part
• Income and class

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

References

1. 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
2. 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.}
}