Skip to content
Contents
The Public Mind

    Type to search. Planned chapters match by title only.

    Chapter 2.2

    Feelings toward the other party

    Among ANES partisans, the average rating of the other party on the 0 to 100 feeling thermometer fell from 48.0 in 1978 to 20.6 in 2024.

    Affiliation
    NYU Stern
    Published
    DraftFigures and numbers may change as the analysis is rebuilt.what is still being checked

    Still to be done: intervals that use the ANES sampling design for years before 2024, the year-by-year distribution of the ratings, and decompositions by race, religion and age.

    Among Democrats and Republicans in the American National Election Studies (ANES) who rated both parties on the feeling thermometerFeeling thermometerA 0 to 100 rating of a group or person, where 0 is the coldest feeling and 100 the warmest. The American National Election Studies ask respondents to rate parties, candidates and social groups this way., the average rating of the other party fell from 48.048.0 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 47.1 to 48.9n1,723Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978SourceANES Cumulative Data File in 1978 to 20.620.6 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 19.8 to 21.5n5,080Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2024SourceANES Cumulative Data File in 2024. Their average rating of their own party went from 71.071.0 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 70.2 to 71.8n1,723Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978SourceANES Cumulative Data File to 70.570.5 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 69.7 to 71.3n5,080Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2024SourceANES Cumulative Data File over the same years and ended near where it started. The thermometer asks a respondent to rate a party from cold to warm, with a neutral midpoint between the two ends. Partisans' feelings toward the two parties moved apart mainly on the other-party side. Part of the fall after 2008 goes with the change in how the ANES interviewed people, as it added web and panel respondents to its in-person samples, and the fall continues inside each interview mode.

    The fall appears even among partisans born in 1942 or earlier, the oldest of the four birth cohortsCohortPeople born in the same period, such as the same decade. Comparing cohorts shows whether a difference comes from when people were born, as distinct from their age or the year they were surveyed. this chapter follows. Their average rating of the other party was 48.148.1 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 46.9 to 49.3n1,022Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both parties, born 1942 or earlierPeriod1978SourceANES Cumulative Data File in 1978 and 22.022.0 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 17.9 to 26.2n243Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both parties, born 1942 or earlierPeriod2024SourceANES Cumulative Data File in 2024. Everyone in this cohort was an adult in 1978, and the reading they gave the other party then sat below the midpoint, lukewarm. In 2024 the reading from the same birth years was cold. The fall from lukewarm to cold took place within one adult lifetime, among different people interviewed decades apart. It is a change in a group average and says nothing about how any one person changed.

    The other party's rating fell more than the own party's

    The partisans in these averages include leanersPartisan leanerA respondent who first says they are independent but, when asked, says they lean toward one party. Most figures group leaners with the party they lean toward, and each chapter says which grouping it uses., independents who say they lean toward a party, and every one of them rated both parties. The difference between the two ratings, own party minus other party, is a separate quantity, averaged over the same respondents. It rose from 23.023.0 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 21.8 to 24.2n1,723Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978SourceANES Cumulative Data File points in 1978 to 49.949.9 points on the 0 to 100 feeling thermometerEstimandweighted mean rating (95% interval from the Kish effective sample size)Interval95% CI 48.6 to 51.2n5,080Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2024SourceANES Cumulative Data File in 2024, with its peak in 2020. Researchers call this gap affective polarizationAffective polarizationThe gap between how warmly people feel toward their own party and how coldly they feel toward the other party. It measures feeling, not disagreement on issues. [1] [2]. In this record the gap widened as the other-party rating fell, while the own-party rating moved within a narrower range.

    Most of the fall came after 1990. Figure 1 plots the two averages and shades the gap between them, and its buttons repeat the trend for face-to-face and web respondents separately.

    Figure 1

    Partisans rated the other party 20.6 in 2024, down from 48.0 in 1978, and rated their own party 70.5

    Mean feeling thermometer rating of the respondent's own party and of the other party, partisans including leaners, 1978 to 2024, 0 to 100

    Show data
    Data for Partisans rated the other party 20.6 in 2024, down from 48.0 in 1978, and rated their own party 70.5
    conf_highconf_lowestimatemodenseriesyear
    48.8747.1248.00All interviews1,723Other party1978
    47.7245.4346.57All interviews1,310Other party1980
    45.4343.1244.27All interviews1,181Other party1982
    46.8644.9245.89All interviews1,881Other party1984
    47.2545.2546.25All interviews1,796Other party1986
    46.6044.5245.56All interviews1,717Other party1988
    47.8345.7446.78All interviews1,656Other party1990
    42.7440.9541.84All interviews2,098Other party1992
    43.6341.4042.51All interviews1,567Other party1994
    42.3539.9941.17All interviews1,528Other party1996
    42.3439.5940.96All interviews1,104Other party1998
    42.7840.4341.60All interviews1,525Other party2000
    41.0538.0839.57All interviews1,055Other party2004
    35.8833.5034.69All interviews1,979Other party2008
    28.6527.0727.86All interviews5,079Other party2012
    26.9025.1726.04All interviews3,610Other party2016
    19.9118.5819.24All interviews7,174Other party2020
    21.4619.8020.63All interviews5,080Other party2024
    71.8470.1971.01All interviews1,723Own party1978
    73.4671.4872.47All interviews1,310Own party1980
    75.2373.3174.27All interviews1,181Own party1982
    75.6174.0274.81All interviews1,881Own party1984
    75.2773.6674.46All interviews1,796Own party1986
    77.1275.4276.27All interviews1,717Own party1988
    72.0770.3071.19All interviews1,656Own party1990
    70.7569.2069.98All interviews2,098Own party1992
    71.1269.2570.19All interviews1,567Own party1994
    72.7670.7871.77All interviews1,528Own party1996
    71.6369.2270.42All interviews1,104Own party1998
    73.6271.5872.60All interviews1,525Own party2000
    74.8072.5073.65All interviews1,055Own party2004
    71.7369.6970.71All interviews1,979Own party2008
    70.5069.0569.78All interviews5,079Own party2012
    67.9866.2767.13All interviews3,610Own party2016
    72.1770.8971.53All interviews7,174Own party2020
    71.3469.7070.52All interviews5,080Own party2024
    35.8833.5034.69Face-to-face1,979Other party2008
    34.1631.4532.81Face-to-face1,798Other party2012
    32.2229.0130.61Face-to-face1,029Other party2016
    27.3523.6825.52Face-to-face960Other party2024
    71.7369.6970.71Face-to-face1,979Own party2008
    71.0668.6269.84Face-to-face1,798Own party2012
    66.5363.3564.94Face-to-face1,029Own party2016
    70.3266.9468.63Face-to-face960Own party2024
    26.0224.1225.07Web3,281Other party2012
    25.3423.3224.33Web2,581Other party2016
    19.1916.7317.96Web1,850Other party2024
    70.6468.8469.74Web3,281Own party2012
    68.9666.9467.95Web2,581Own party2016
    72.7270.1871.45Web1,850Own party2024

    The lower line is the rating of the other party: it starts below the neutral midpoint in 1978 and falls across the record, with most of the fall after 1990. The upper line, the rating of the own party, ends near where it began, and the shaded wedge between the two lines is the gap, widest in 2020. Each line is the weighted average among Democrats and Republicans, leaners included, with its 95% interval as a band. Values in Table 1.

    Note: Bands are confidence intervals from the Kish effective sample size; they treat the weighted sample as a simple random sample and ignore clustering.Source: ANES Cumulative Data File, 1978-2024; partisans including leaners; weighted (VCF0009z; mode weights VCF0009x and VCF0009y for 2008, 2012, 2016 and 2024)

    More partisans rate the other party at zero

    The average hides a change in how people answer. Zero is the coldest answer the thermometer allows. In 1978, 5.4%5.4% percent of partisans rating the other party at zeroEstimandweighted share of partisans (95% Wilson interval on the Kish effective sample size)Interval95% CI 4.4% to 6.6%n1,723Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978SourceANES Cumulative Data File of partisans gave the other party a zero; in 2024, 33.2%33.2% percent of partisans rating the other party at zeroEstimandweighted share of partisans (95% Wilson interval on the Kish effective sample size)Interval95% CI 31.4% to 35.1%n5,080Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2024SourceANES Cumulative Data File did, after a peak in 2020. The share who rated the other party above the neutral midpoint moved the other way over the same years. The share at zero is lower in face-to-face interviews than on the web in every year with both modes, and it rose inside face-to-face interviews as well, from 2008 to 2024.

    Figure 2

    The share of partisans rating the other party at zero rose from 5.4% in 1978 to 33.2% in 2024

    Percent of partisans including leaners who rate the other party at the coldest possible answer of zero, 1978 to 2024

    Share rating the other party at zero (percent of partisans)

    Show data
    Data for The share of partisans rating the other party at zero rose from 5.4% in 1978 to 33.2% in 2024
    conf_highconf_lowestimatemodenseriesyear
    6.574.435.40All interviews1,723Other party1978
    8.375.626.87All interviews1,310Other party1980
    8.815.867.20All interviews1,181Other party1982
    9.236.787.92All interviews1,881Other party1984
    9.136.647.80All interviews1,796Other party1986
    9.546.958.15All interviews1,717Other party1988
    8.926.377.55All interviews1,656Other party1990
    9.577.208.31All interviews2,098Other party1992
    8.245.496.74All interviews1,567Other party1994
    10.447.268.72All interviews1,528Other party1996
    11.507.759.46All interviews1,104Other party1998
    10.006.918.32All interviews1,525Other party2000
    14.6910.3312.35All interviews1,055Other party2004
    16.8212.9014.75All interviews1,979Other party2008
    24.3621.2522.77All interviews5,079Other party2012
    23.9020.6422.23All interviews3,610Other party2016
    41.5738.5140.03All interviews7,174Other party2020
    35.1231.3633.22All interviews5,080Other party2024
    16.8212.9014.75Face-to-face1,979Other party2008
    19.8615.0217.31Face-to-face1,798Other party2012
    20.1914.5117.16Face-to-face1,029Other party2016
    26.2818.9022.38Face-to-face960Other party2024
    27.8923.9125.85Web3,281Other party2012
    26.1322.2024.11Web2,581Other party2016
    41.7635.8838.78Web1,850Other party2024

    The line climbs from 1978 to its peak in 2020 and ends lower in 2024. It is the weighted share of partisans, leaners included, who rate the other party at zero, with a Wilson band, an interval for a share that stays inside its possible range. The buttons restrict the sample to face-to-face or web respondents, and the web line runs above the face-to-face line in every year with both. Values in Table 3.

    Note: Bands are Wilson intervals on the Kish effective sample size.Source: ANES Cumulative Data File, 1978-2024; partisans including leaners; weighted (VCF0009z; mode weights VCF0009x and VCF0009y for 2008, 2012, 2016 and 2024)

    Who was asked, and what changed in the asking

    The ratings come from the ANES Cumulative Data File, 1818 surveysEstimandcount of survey yearsWeightsnone (count)PopulationANES time-series studies with both party thermometersPeriod1978 to 2024SourceANES Cumulative Data File surveys from 1978 to 2024 that asked both party thermometers. The sample is 43,06343,063 respondentsEstimandcount of respondents in the sampleWeightsnone (count)Populationpartisans, leaners included, with a rating of both partiesPeriod18 surveys, 1978 to 2024SourceANES Cumulative Data File respondents who identified with a party or leaned toward one and rated both parties, so the own-party, other-party and gap averages in each year rest on the same people. Pure independents are left out because they have no own party. Each average uses the ANES full-sample 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., which scales respondents so the sample matches the population on known traits.The bands and intervals treat the weighted sample as a simple random sample. For the 2024 all-interview average the report also gives the standard error computed from the survey's strata and sampling units, and the printed interval is the wider of the two. Each survey is a separate cross-section of people, so the chapter describes how averages changed and says nothing about how any one person's feelings changed. The qualification that matters most is survey modeSurvey modeHow respondents were interviewed: in person, by phone, on paper or online. Mode can change answers, so a change in mode can look like a change in opinion.. Through 2008 the ANES interviewed mainly in person, and in 2012 it added a separate web sample. In 2020 it interviewed no one in person [3], and in 2024 it combined a new face-to-face sample, a new web sample and respondents from an earlier panel [4]. The report gives every estimate with its interval and sample size.

    Part of the fall after 2008 goes with survey mode

    Web respondents rate the other party lower than face-to-face respondents in every year with both kinds of interview, 2012, 2016 and 2024. The ANES web and face-to-face samples are separate samples, so the difference combines how people answer with who answers. The comparison describes the two samples, and it does not identify what interviewing the same people in the other mode would have done to their answers.

    The mix of modes changes what the all-interview line shows. Between 2008, a year with face-to-face interviews only, and 2024, the all-interview rating of the other party fell by 14.114.1 points on the 0 to 100 feeling thermometerEstimandfall in the all-interview mean (95% percentile interval, bootstrap within year)Interval95% CI 12.6 to 15.5n7,059Weightsweight (VCF0009z); weight_ftf (VCF0009x)Populationpartisans, leaners included, with a rating of both partiesPeriod2008 to 2024SourceANES Cumulative Data File points. Among face-to-face respondents alone it fell by 9.29.2 points on the 0 to 100 feeling thermometerEstimandfall in the face-to-face mean (95% percentile interval, bootstrap within year)Interval95% CI 6.9 to 11.4n7,059Weightsweight (VCF0009z); weight_ftf (VCF0009x)Populationpartisans, leaners included, with a rating of both partiesPeriod2008 to 2024SourceANES Cumulative Data File points. The remaining 4.94.9 points on the 0 to 100 feeling thermometerEstimandgap between the all-interview and the face-to-face mean in 2024 (95% percentile interval, bootstrap within year)Interval95% CI 3.5 to 6.3n7,059Weightsweight (VCF0009z); weight_ftf (VCF0009x)Populationpartisans, leaners included, with a rating of both partiesPeriod2008 to 2024SourceANES Cumulative Data File points are the distance between the 2024 all-interview average and the 2024 face-to-face average. Figure 3 splits them between web respondents and the panel and paper respondents who belong to neither mode group. The fall continues inside each mode, so the direction of the trend holds however the interviews are counted, and its size depends on which interviews are counted. The change from 2016 to 2020 cannot be split by mode at all, because the 2020 study has no mode weights.

    Figure 3

    Face-to-face respondents rated the other party 4.9 points warmer than the all-interview average in 2024, and web respondents colder

    Rating of the other party by interview mode and the split of the all-interview fall, partisans including leaners, 2008 to 2024, 0 to 100

    Of the 14.1 points the all-interview rating fell from 2008 to 2024, 4.9 are the distance between the 2024 all-interview and face-to-face averages

    Split of the fall in the all-interview mean rating of the other party, partisans including leaners, 2008 to 2024, points on the 0 to 100 thermometer

    Mean rating of the other party (0 to 100)

    The first step is the fall within face-to-face interviews; the next two are the parts of the 2024 gap that come from web respondents and from panel and paper respondents. Each of those two is the group's share of the 2024 all-interview weight times the distance between the face-to-face average and the group's average.

    Show data
    Data for Face-to-face respondents rated the other party 4.9 points warmer than the all-interview average in 2024, and web respondents colder, Split of the fall
    conf_highconf_lowkindnstepvalue
    35.8833.50start1,9792008 mean34.69
    −6.93−11.35delta7,059Face-to-face fall−9.18
    −1.49−2.92delta7,059Web in the mix−2.20
    −1.97−3.43delta7,059Panel and paper−2.70
    21.4619.80end5,0802024 mean20.63

    Read the columns left to right: the first is the 2008 average rating of the other party, and the next three are the steps that take it down to the 2024 average. The first step is the fall within face-to-face interviews, and the other two are the parts of the 2024 gap that come from web respondents and from panel and paper respondents. The first tab plots the rating by mode, with the face-to-face line above the web line in 2012, 2016 and 2024 and the all-interview line between them. Values in Table 4.

    Note: The 2008 study was interviewed in person only, so its face-to-face and all-interview estimates coincide and there is no web estimate. The 2024 all-interview sample also contains re-interviewed panel and paper respondents who belong to neither mode group. The steps of the second tab add to the total within rounding.Source: ANES Cumulative Data File, 2008-2024; partisans including leaners; weighted (VCF0009z; mode weights VCF0009x and VCF0009y)

    What the fall means, and what it does not

    The fall is not the work of one kind of partisan. Strong partisans rate the other party lowest, and their share of partisans rose from 27.2%27.2% percent of partisansEstimandweighted share who are strong partisans (95% Wilson interval)Interval95% CI 25.1% to 29.3%n1,723Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978SourceANES Cumulative Data File in 1978 to 46.2%46.2% percent of partisansEstimandweighted share who are strong partisans (95% Wilson interval)Interval95% CI 44.2% to 48.2%n5,080Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2024SourceANES Cumulative Data File in 2024. A change in the mix alone could therefore have lowered the average without any group changing. A decomposition that holds the 1978 mix fixed finds that change within groups accounts for 90%89.7% percent of the total changeEstimandshift-share of the other-party rating over strong, weak and leaning partisans: within as a share of the total change (shift-share term (95% percentile interval, bootstrap within year))Interval95% CI 87.8% to 91.7%n6,803Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978 to 2024SourceANES Cumulative Data File of the fall, and the same result holds when the groups are education levels. Every strength, party, education and birth cohort group the chapter follows ends its record with a lower other-party rating than it began with. Leaners in 2024 rate the other party lower than strong partisans did in 1978. What these ratings go with in voting belongs to Companion 2.2a, Negative partisanship, and how far apart partisans stand on issues to Planned, Affective and issue polarization (planned). Accounts that start from partisanship as a social identity [5], from the alignment of party with other traits or from negative partisanship [6] are candidate explanations, and the ratings alone cannot tell them apart.

    What these ratings cannot tell you

    The ratings are repeated cross-sections, so they cannot show that any partisan grew colder, and the cohort lines cannot separate age, cohort and the year of the survey. The mode results rest on mode weights that exist for 2008, 2012, 2016 and 2024 only. The gap between the all-interview and face-to-face averages is a decomposition of the record, not an estimate of what a mode does to an answer. A thermometer rating of a party may reflect feelings about party elites more than about the fellow citizens who support it [7]. A rating of zero is the coldest answer the scale allows and not a measure of behavior. The change from 2020 to 2024 depends on which 2024 sample is compared with 2020, and because 2020 has no face-to-face interviews these files cannot settle it.

    Go deeper

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

    Reading paths that include this chapter:

    Data sources

    • ANES Cumulative Data FileCumulative Data File 1948-2024: party thermometers 1978-2024, presidential nominee thermometers 1968-2024Years: 1948-2024 (33 study years; even years only after 2004; no 2006, 2010 or 2014 study) · Weights: weight (VCF0009z); weight_ftf and weight_web (VCF0009x, VCF0009y) for mode-specific estimates in 2008, 2012, 2016 and 2024 · License: Free download after accepting the ANES terms of use; cite the release.
    • ANES 2024 Time Series2024 Time Series: party thermometers V241166 and V241167, party identification V241227x, variance strata V240107d and primary sampling units V240107c, used for one design-based check of the 2024 all-interview other-party meanYears: 2024 (pre-election and post-election interviews) · Weights: pre-election full-sample weight (V240107a), equal to the Cumulative File weight · License: Free download after accepting the ANES terms of use; cite the release.

    All data sources

    References

    1. Iyengar, Shanto; Sood, Gaurav; Lelkes, Yphtach (2012). Affect, Not Ideology: A Social Identity Perspective on Polarization. Public Opinion Quarterly 76(3): 405-431. academic.oup.com/poq/article-abstract/76/3/405/1894274
    2. Iyengar, Shanto; Lelkes, Yphtach; Levendusky, Matthew; Malhotra, Neil; Westwood, Sean J. (2019). The Origins and Consequences of Affective Polarization in the United States. Annual Review of Political Science 22: 129-146. doi:10.1146/annurev-polisci-051117-073034
    3. American National Election Studies (2022). Methodology Report for the ANES 2020 Time Series Study. Stanford University and the University of Michigan. electionstudies.org/wp-content/uploads/2022/08/anes_timeseries_2020_methodology_report.pdf
    4. American National Election Studies (2025). ANES 2024 Time Series Study Full Release: User Guide and Codebook. Stanford University and the University of Michigan. electionstudies.org/wp-content/uploads/2025/08/anes_timeseries_2024_userguidecodebook_20250808.pdf
    5. Mason, Lilliana (2018). Uncivil Agreement: How Politics Became Our Identity. University of Chicago Press. press.uchicago.edu/ucp/books/book/chicago/U/bo27527354.html
    6. Abramowitz, Alan I.; Webster, Steven W. (2016). The Rise of Negative Partisanship and the Nationalization of U.S. Elections in the 21st Century. Electoral Studies 41: 12-22. www.sciencedirect.com/science/article/abs/pii/S0261379415001857
    7. Druckman, James N.; Levendusky, Matthew S. (2019). What Do We Measure When We Measure Affective Polarization?. Public Opinion Quarterly 83(1): 114-122. academic.oup.com/poq/article-abstract/83/1/114/5486527

    How to cite

    @misc{singh2026feelingstowardtheotherparty,
  author = {Singh, Vishal},
  title = {Feelings toward the other party},
  year = {2026},
  url = {https://vishalsingh.org/politics/feelings-toward-the-other-party},
  note = {In The Public Mind. Published 30 September 2026.}
}