Skip to content
Contents
The Public Mind

    Type to search. Planned chapters match by title only.

    Chapter 2.2

    Negative partisanship

    The association of reported turnout with coldness toward the other party, beyond warmth toward the own party, grew by 2.6 points per 10 rating points between 1978 to 1990 and 2012 to 2024 and grew under every measure, election type and interview mode checked, while House voting and campaign activity show a change only under some measures and comparisons.

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

    Still to be checked: intervals that use the ANES sampling design (the Cumulative File has no strata or primary sampling units), the 2012 to 2024 results against the ANES 2012, 2016 and 2020 time-series files, and a panel check with the 2016, 2020 and 2024 panel.

    In the American National Election Studies (ANES) of 2012 to 2024, partisans who rated the other party colder were more likely to vote for their own party's House candidate, and the coefficient on coldness toward the other party was larger than the coefficient on warmth toward the own party. Among Democrats and Republicans (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. included), a rating of the other party 1010 points on the feeling thermometerEstimandsize of the change in a rating that a coefficient describes (a definition, not an estimate)Weightsnone (constant)Populationthe unit in which coefficients are reportedPeriod1978 to 2024SourceANES Cumulative Data File points colder 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. went with 3.5 (95% CI 3.1 to 3.9)3.5 percentage points of the outcome per 10 rating pointsEstimandcoldness toward the other party on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 3.1 to 3.9n13,912Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File 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. more own-party House voting, with the rating of the own party held fixed, and a rating of the own party 1010 points on the feeling thermometerEstimandsize of the change in a rating that a coefficient describes (a definition, not an estimate)Weightsnone (constant)Populationthe unit in which coefficients are reportedPeriod1978 to 2024SourceANES Cumulative Data File points warmer went with 1.9 (95% CI 1.5 to 2.3)1.9 percentage points of the outcome per 10 rating pointsEstimandwarmth toward the own party on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.5 to 2.3n13,912Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File points, with the rating of the other party held fixed. This chapter calls coldness minus warmth the per-point difference. It was 1.6 (95% CI 1.0 to 2.1)1.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.0 to 2.1n13,912Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File points for House voting, 1.6 (95% CI 1.0 to 2.2)1.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.0 to 2.2n18,976Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024SourceANES Cumulative Data File for reported turnout and 1.5 (95% CI 0.8 to 2.3)1.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.8 to 2.3n18,697Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024SourceANES Cumulative Data File for campaign activity in 2012 to 2024, and 0.7 (95% CI 0.3 to 1.2)0.7 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.3 to 1.2n15,612Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File for party-line presidential voting.

    Whether the difference is new depends on the comparison. In 1978 to 1990 the per-point difference was 0.0 (95% CI −0.8 to 0.9)0.0 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.8 to 0.9n5,930Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File for House voting and 0.0 (95% CI −0.7 to 0.7)0.0 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.7 to 0.7n10,606Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 1990SourceANES Cumulative Data File for campaign activity. The change since then in reported turnout is above zero under every measure the chapter uses, in presidential years alone and within each interview mode (2.6 (95% CI 1.7 to 3.6)2.6 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 1.7 to 3.6Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File points in all interviews, 2.0 (95% CI 0.8 to 3.3)2.0 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.8 to 3.3Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File in presidential years alone, 3.5 (95% CI 2.0 to 4.9)3.5 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 2.0 to 4.9Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, face-to-face interviews minus 1978 to 1990SourceANES Cumulative Data File face to face and 2.3 (95% CI 1.2 to 3.5)2.3 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 1.2 to 3.5Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, web interviews minus 1978 to 1990SourceANES Cumulative Data File on the web). The changes for House voting and campaign activity are above zero in all interviews and include zero in presidential years alone and within each mode, and the cut-off measures disagree with the per-point measure about presidential voting.

    These are comparisons of how feelings and behavior go together, so the chapter says "went with". Both are measured in the same survey, so either may shape the other, and both may follow from traits that the survey measures only in part, such as interest in politics or the strength of party identification.

    The ratings and the behavior come from the ANES Cumulative Data File: 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 in 1818 surveysEstimandcount of survey yearsWeightsnone (count)PopulationANES time-series studies with both party thermometersPeriod1978 to 2024SourceANES Cumulative Data File surveys from 1978 to 2024 who identify with a party or lean toward one and rated both parties on a scale from 00 points on the feeling thermometerEstimandlowest answer the feeling thermometer allows (a definition of the scale, not an estimate)Weightsnone (constant)Populationanswers to the feeling thermometerPeriod1978 to 2024SourceANES Cumulative Data File to 100100 points on the feeling thermometerEstimandhighest answer the feeling thermometer allows (a definition of the scale, not an estimate)Weightsnone (constant)Populationanswers to the feeling thermometerPeriod1978 to 2024SourceANES Cumulative Data File. Pure independents are left out because they have no own party. Leaners count as partisans throughout, as in Chapter 2.1, Feelings toward the other party, and they are 27%27% percent of partisansEstimandweighted share who lean toward a party and do not identify with itWeightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2012 to 2024SourceANES Cumulative Data File of partisans in 2012 to 2024 and 27%27% percent of partisansEstimandweighted share who lean toward a party and do not identify with itWeightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990; a check below drops them. Warmth is the rating of the own party. Coldness is 100100 points on the feeling thermometerEstimandhighest answer the feeling thermometer allows (a definition of the scale, not an estimate)Weightsnone (constant)Populationanswers to the feeling thermometerPeriod1978 to 2024SourceANES Cumulative Data File minus the rating of the other party, so that a higher value is colder and the two coefficients can be set side by side. Five behaviors follow. Party-line presidential voting is a vote for the own party's candidate among partisans who reported voting for a major-party candidate, and defection is its complement. Party-line House voting is the same for the House candidate. A straight-ticket vote is a vote for one party for president and the House, the complement of ticket splitting. Reported turnout is the report of having voted. Campaign activity is a yes to at least one of these: tried to persuade others how to vote, attended a meeting or rally, worked for a party or candidate, displayed a button or sticker, or gave money. All of these behaviors are self-reports. Presidential voting and the straight ticket exist only in presidential election years. The other three also exist in midterm years, which are 56%56% percent of respondentsEstimandshare of the sample interviewed in a midterm election year (respondent count)n11,264Weightsnone (count)Populationpartisans, leaners included, with a rating of both partiesPeriod1978 to 1990SourceANES Cumulative Data File of the respondents in 1978 to 1990, 25%25% percent of respondentsEstimandshare of the sample interviewed in a midterm election year (respondent count)n10,856Weightsnone (count)Populationpartisans, leaners included, with a rating of both partiesPeriod1992 to 2008SourceANES Cumulative Data File in 1992 to 2008 and none in 2012 to 2024, so a comparison of those three across periods mixes the period with the election type.Estimates use 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.. Coefficients come from weighted linear probability models with year fixed effects, and their intervals use HC3HC3 robust standard errorsA way of computing standard errors that stays valid when the spread of the outcome differs across respondents, with a correction for small samples. The book uses it for regressions that have no clusters. robust standard errors [1]. Shares of groups use the Wilson interval [2] on the Kish effective sample sizeEffective sample sizeThe number of equally weighted respondents that would give the same precision as the weighted sample. It is smaller than the raw count when the weights vary a lot, and it sets how wide the confidence interval is. [3]. All intervals treat the weighted sample as a simple random sample and ignore the ANES strata and clusters.

    Abramowitz and Webster use ANES ratings of parties and candidates to argue that increasingly negative feelings among party supporters, leaners included, about the opposing party have led to more party loyalty and more straight-ticket voting [4] [5]. A study of Canada reports that negative partisan attitudes influence turnout and a range of political activities as well as vote choice [6], a study of elections in Australia, Canada, New Zealand and the United States reports an autonomous influence of negative party identification [7], and a study of the 2016 American election examines negative and positive partisanship together [8]. This chapter asks the question of the ANES behavior items from 1978 to 2024: holding the rating of the own party fixed, is behavior more closely tied to the rating of the other party, and when did that start? The trend in the ratings themselves is in Chapter 2.1, Feelings toward the other party, the comparison of emotional distance with distance on issues is in Chapter 2.3, Affective and issue polarization (planned), and feelings toward social groups are in Chapter 2.4, Party and feelings toward social groups (planned).

    Cold-only partisans now outnumber warm-only partisans

    The group comparisons sort partisans by two cut-offs. A partisan is very cold toward the other party when the rating of that party is at or below 1515 points on the feeling thermometerEstimandcut-off chosen by the author (a definition, not an estimate)Weightsnone (constant)Populationcut-off between a rating that counts as very cold toward the other party and one that does notPeriod1978 to 2024SourceANES Cumulative Data File, and very warm toward the own party when the rating of the own party is at or above 8585 points on the feeling thermometerEstimandcut-off chosen by the author (a definition, not an estimate)Weightsnone (constant)Populationcut-off between a rating that counts as very warm toward the own party and one that does notPeriod1978 to 2024SourceANES Cumulative Data File. A looser pair, 3030 points on the feeling thermometerEstimandcut-off chosen by the author (a definition, not an estimate)Weightsnone (constant)Populationcut-off between a rating that counts as very cold toward the other party and one that does notPeriod1978 to 2024SourceANES Cumulative Data File and 7070 points on the feeling thermometerEstimandcut-off chosen by the author (a definition, not an estimate)Weightsnone (constant)Populationcut-off between a rating that counts as very warm toward the own party and one that does notPeriod1978 to 2024SourceANES Cumulative Data File, serves as a check. Four groups follow: neither, warm only, cold only and both.

    Figure 1 gives the size of the four groups. In 1978 to 1990, 32.0%32.0% percent of partisansEstimandweighted share in the group warm only (95% Wilson interval)Interval95% CI 31.2% to 32.9%n11,264Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978 to 1990SourceANES Cumulative Data File of partisans were warm only and 4.4%4.4% percent of partisansEstimandweighted share in the group cold only (95% Wilson interval)Interval95% CI 4.0% to 4.8%n11,264Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978 to 1990SourceANES Cumulative Data File were cold only. In 2012 to 2024 the shares were 10.5%10.5% percent of partisansEstimandweighted share in the group warm only (95% Wilson interval)Interval95% CI 9.9% to 11.1%n20,943Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2012 to 2024SourceANES Cumulative Data File and 28.1%28.1% percent of partisansEstimandweighted share in the group cold only (95% Wilson interval)Interval95% CI 27.3% to 28.9%n20,943Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2012 to 2024SourceANES Cumulative Data File, and 25.2%25.2% percent of partisansEstimandweighted share in the group both (95% Wilson interval)Interval95% CI 24.4% to 26.0%n20,943Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2012 to 2024SourceANES Cumulative Data File were both. Cold only is also larger than warm only within each interview mode in 2012, 2016 and 2024 pooled: 19.5% (95% CI 17.8% to 21.2%)19.5% percent of partisansEstimandweighted share in the group cold only (95% Wilson interval)Interval95% CI 17.8% to 21.2%n3,787Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, with a rating of both partiesPeriod2012, 2016 and 2024, face-to-face interviewsSourceANES Cumulative Data File against 13.5% (95% CI 12.1% to 15.0%)13.5% percent of partisansEstimandweighted share in the group warm only (95% Wilson interval)Interval95% CI 12.1% to 15.0%n3,787Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, with a rating of both partiesPeriod2012, 2016 and 2024, face-to-face interviewsSourceANES Cumulative Data File face to face and 29.1% (95% CI 27.8% to 30.4%)29.1% percent of partisansEstimandweighted share in the group cold only (95% Wilson interval)Interval95% CI 27.8% to 30.4%n7,712Weightsweight_web (VCF0009y)Populationpartisans, leaners included, with a rating of both partiesPeriod2012, 2016 and 2024, web interviewsSourceANES Cumulative Data File against 11.0% (95% CI 10.1% to 11.9%)11.0% percent of partisansEstimandweighted share in the group warm only (95% Wilson interval)Interval95% CI 10.1% to 11.9%n7,712Weightsweight_web (VCF0009y)Populationpartisans, leaners included, with a rating of both partiesPeriod2012, 2016 and 2024, web interviewsSourceANES Cumulative Data File on the web. The average rating, the share rating the other party at zero and the mode checks on both belong to Chapter 2.1, Feelings toward the other party.

    Figure 1

    Partisans very cold toward the other party only went from 4.4% of partisans in 1978 to 1990 to 28.1% in 2012 to 2024, and partisans very warm toward their own party only went from 32.0% to 10.5%

    Share of partisans, leaners included, in each of four groups defined by rating the other party at or below 15 (cold) and their own party at or above 85 (warm) on the 0 to 100 feeling thermometer, three periods and two interview modes

    • 1978 to 1990
    • 1992 to 2008
    • 2012 to 2024
    • 2012, 2016 and 2024, face to face
    • 2012, 2016 and 2024, web
    Show data
    Data for Partisans very cold toward the other party only went from 4.4% of partisans in 1978 to 1990 to 28.1% in 2012 to 2024, and partisans very warm toward their own party only went from 32.0% to 10.5%
    conf_highconf_lowcutestimategroupnsample
    55.5753.73strict54.65Neither11,2641978 to 1990
    32.9031.18strict32.03Warm only11,2641978 to 1990
    4.814.05strict4.41Cold only11,2641978 to 1990
    9.448.39strict8.90Both11,2641978 to 1990
    58.1156.03strict57.07Neither10,8561992 to 2008
    23.8722.10strict22.98Warm only10,8561992 to 2008
    9.368.17strict8.75Cold only10,8561992 to 2008
    11.8910.56strict11.21Both10,8561992 to 2008
    37.0835.31strict36.19Neither20,9432012 to 2024
    11.079.95strict10.50Warm only20,9432012 to 2024
    28.9327.28strict28.10Cold only20,9432012 to 2024
    26.0224.43strict25.21Both20,9432012 to 2024
    28.6126.96loose27.78Neither11,2641978 to 1990
    49.6747.83loose48.75Warm only11,2641978 to 1990
    5.124.34loose4.71Cold only11,2641978 to 1990
    19.4918.05loose18.76Both11,2641978 to 1990
    28.2426.37loose27.30Neither10,8561992 to 2008
    38.7436.70loose37.71Warm only10,8561992 to 2008
    10.018.79loose9.38Cold only10,8561992 to 2008
    26.5324.70loose25.61Both10,8561992 to 2008
    16.9815.63loose16.29Neither20,9432012 to 2024
    15.0613.77loose14.41Warm only20,9432012 to 2024
    22.6121.10loose21.85Cold only20,9432012 to 2024
    48.3746.54loose47.46Both20,9432012 to 2024
    51.4347.13strict49.28Neither3,7872012, 2016 and 2024, face to face
    15.0412.10strict13.50Warm only3,7872012, 2016 and 2024, face to face
    21.2517.85strict19.49Cold only3,7872012, 2016 and 2024, face to face
    19.4316.14strict17.73Both3,7872012, 2016 and 2024, face to face
    38.7335.95strict37.33Neither7,7122012, 2016 and 2024, web
    11.9310.13strict11.00Warm only7,7122012, 2016 and 2024, web
    30.3927.79strict29.07Cold only7,7122012, 2016 and 2024, web
    23.8221.42strict22.60Both7,7122012, 2016 and 2024, web
    Cold only is the largest of the two single-condition groups in 2012 to 2024 and the smaller in 1978 to 1990, and it is larger than warm only within each interview mode in 2012, 2016 and 2024 pooled. The both group went from 8.9% to 25.2%. Warm only means rating the own party at or above 85 and the other party above 15; cold only means the reverse; both means meeting both conditions. The second view uses the looser cut-offs of 30 and 70. Each dot is a weighted share with a 95% interval (Wilson). The trend in the average rating and in the share rating the other party at zero is in the chapter on feelings toward the other party.Note: Face-to-face and web estimates use the mode weights weight_ftf and weight_web, which exist for 2008, 2012, 2016 and 2024. The 2020 study has no mode weights and the 2024 panel and paper respondents belong to neither mode group.Note: Respondents give the ratings 15, 30, 70 and 85 more often than the values next to them, so each cut-off sits on a common answer and includes it.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z; weight_ftf and weight_web for the mode estimates of 2008, 2012, 2016 and 2024)

    Cold-only partisans voted the party line more often than warm-only partisans

    Figure 2 compares the two groups that differ on one condition. Among partisans who voted for a major-party presidential candidate in 2012 to 2024, 98.1% (95% CI 97.5% to 98.5%)98.1% percent of the groupEstimandweighted share, party-line presidential vote, group: cold only (95% Wilson interval)Interval95% CI 97.5% to 98.5%n4,774Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File of those cold only voted for their own party's candidate, against 95.4% (95% CI 93.8% to 96.6%)95.4% percent of the groupEstimandweighted share, party-line presidential vote, group: warm only (95% Wilson interval)Interval95% CI 93.8% to 96.6%n1,651Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File of those warm only, 83.2%83.2% percent of the groupEstimandweighted share, party-line presidential vote, group: neither (95% Wilson interval)Interval95% CI 81.7% to 84.6%n4,794Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File of those in neither group and 99.2%99.2% percent of the groupEstimandweighted share, party-line presidential vote, group: both (95% Wilson interval)Interval95% CI 98.7% to 99.5%n4,393Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File of those in both. Cold only minus warm only was 2.6 (95% CI 1.2 to 4.1)2.6 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, party-line presidential vote (95% interval, independent groups)Interval95% CI 1.2 to 4.1n6,425Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File points in 2012 to 2024 and 6.0 (95% CI 2.3 to 9.6)6.0 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, party-line presidential vote (95% interval, independent groups)Interval95% CI 2.3 to 9.6n1,221Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990, so the very cold voted the party line more often in both periods, and the interval excludes zero in each. The gap is smaller in the later period, when party-line voting was higher everywhere: the share voting the party line among all partisans who voted for a major-party candidate was 85.9%85.9% percent of partisansEstimandweighted share acting (95% Wilson interval)Interval95% CI 84.6% to 87.0%n3,090Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990 and 93.3%93.3% percent of partisansEstimandweighted share acting (95% Wilson interval)Interval95% CI 92.7% to 93.8%n15,612Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File in 2012 to 2024.

    Figure 2

    Partisans very cold toward the other party only voted for their own party's presidential candidate more often than those very warm toward their own party only in each period, at 98.1% against 95.4% in 2012 to 2024

    Share of partisans, leaners included, who acted as the view names (first view: voted for their own party's presidential candidate), by whether they rated the other party at or below 15 (cold) and their own party at or above 85 (warm), three periods

    1978 to 1990

    1992 to 2008

    2012 to 2024

    Share voting for own party's presidential candidate (%)

    Show data
    Data for Partisans very cold toward the other party only voted for their own party's presidential candidate more often than those very warm toward their own party only in each period, at 98.1% against 95.4% in 2012 to 2024
    conf_highconf_lowestimategroupnoutcomeperiod
    80.3576.2078.34Neither1,510pres1978 to 1990
    92.3088.8790.73Warm only1,100pres1978 to 1990
    98.7191.8196.69Cold only121pres1978 to 1990
    99.5797.1798.89Both359pres1978 to 1990
    85.9682.9584.51Neither2,753pres1992 to 2008
    95.7293.0194.52Warm only1,361pres1992 to 2008
    98.2794.7996.98Cold only496pres1992 to 2008
    99.7498.2799.32Both790pres1992 to 2008
    84.5681.7383.19Neither4,794pres2012 to 2024
    96.6193.8495.42Warm only1,651pres2012 to 2024
    98.5297.4798.06Cold only4,774pres2012 to 2024
    99.4898.7299.18Both4,393pres2012 to 2024
    73.6470.4972.09Neither3,117house1978 to 1990
    81.5077.9879.80Warm only1,995house1978 to 1990
    93.0585.4989.88Cold only247house1978 to 1990
    91.4486.3289.14Both571house1978 to 1990
    76.7373.4975.14Neither3,336house1992 to 2008
    87.0383.0485.14Warm only1,501house1992 to 2008
    93.1387.9790.87Cold only590house1992 to 2008
    96.3492.9094.89Both844house1992 to 2008
    82.1779.0380.65Neither4,286house2012 to 2024
    91.8987.5489.92Warm only1,376house2012 to 2024
    95.7794.0494.97Cold only4,397house2012 to 2024
    96.2594.4195.42Both3,853house2012 to 2024
    72.7167.7470.28Neither1,299straight1978 to 1990
    79.4674.1176.89Warm only952straight1978 to 1990
    93.3481.0888.57Cold only105straight1978 to 1990
    90.5383.0587.25Both306straight1978 to 1990
    77.7073.7975.80Neither2,282straight1992 to 2008
    87.6183.0785.49Warm only1,140straight1992 to 2008
    94.0288.0691.50Cold only421straight1992 to 2008
    97.1793.6095.73Both685straight1992 to 2008
    84.2781.1782.78Neither4,028straight2012 to 2024
    92.2687.9490.32Warm only1,351straight2012 to 2024
    95.7193.9294.89Cold only4,232straight2012 to 2024
    96.3494.5195.52Both3,824straight2012 to 2024
    63.5261.0562.29Neither5,885turnout1978 to 1990
    70.1467.0368.61Warm only3,405turnout1978 to 1990
    67.3758.6963.14Cold only472turnout1978 to 1990
    74.5268.7471.72Both930turnout1978 to 1990
    71.6969.0570.39Neither5,620turnout1992 to 2008
    75.5071.5573.57Warm only2,368turnout1992 to 2008
    82.0576.0179.19Cold only864turnout1992 to 2008
    85.3480.4083.01Both1,158turnout1992 to 2008
    79.5376.8378.21Neither6,432turnout2012 to 2024
    84.3479.8282.19Warm only2,043turnout2012 to 2024
    89.0386.7487.93Cold only5,582turnout2012 to 2024
    89.6387.1588.45Both4,919turnout2012 to 2024
    34.9932.5633.76Neither5,835act_any1978 to 1990
    41.7038.4040.04Warm only3,384act_any1978 to 1990
    48.0839.0943.53Cold only464act_any1978 to 1990
    47.7541.3544.53Both923act_any1978 to 1990
    40.4937.6639.07Neither5,614act_any1992 to 2008
    47.1342.6744.89Warm only2,362act_any1992 to 2008
    60.8353.4757.19Cold only864act_any1992 to 2008
    64.4158.0161.26Both1,157act_any1992 to 2008
    44.1440.9042.51Neither6,364act_any2012 to 2024
    57.5451.6154.59Warm only2,011act_any2012 to 2024
    61.7558.2860.03Cold only5,510act_any2012 to 2024
    64.0760.2662.18Both4,812act_any2012 to 2024
    In each panel the shares rise from neither to warm only, cold only and both, and cold only sits to the right of warm only in all three periods. Cold-only partisans were 4.4% of partisans in 1978 to 1990 and 28.1% in 2012 to 2024, and warm-only partisans were 32.0% and 10.5%. Warm only means rating the own party at or above 85 and the other party above 15; cold only means the reverse; both means meeting both conditions. Each dot is a weighted share with a 95% interval (Wilson). The other views change the behavior.Note: Cells below 100 respondents are left out.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z)

    Cold only is above warm only in all three periods for House voting as well, where cold only minus warm only was 10.1 (95% CI 5.9 to 14.2)10.1 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, party-line House vote (95% interval, independent groups)Interval95% CI 5.9 to 14.2n2,242Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File points in 1978 to 1990 and 5.1 (95% CI 2.7 to 7.4)5.1 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, party-line House vote (95% interval, independent groups)Interval95% CI 2.7 to 7.4n5,773Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File in 2012 to 2024, and for the straight ticket, at 11.7 (95% CI 5.0 to 18.4)11.7 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, straight-ticket vote (95% interval, independent groups)Interval95% CI 5.0 to 18.4n1,057Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 1990SourceANES Cumulative Data File and 4.6 (95% CI 2.2 to 6.9)4.6 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, straight-ticket vote (95% interval, independent groups)Interval95% CI 2.2 to 6.9n5,583Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024SourceANES Cumulative Data File. The House gap fell by −5.0 (95% CI −9.8 to −0.3)−5.0 percentage pointsEstimandchange in the difference between the share among partisans cold only and the share among partisans warm only, party-line House vote (95% interval, independent samples)Interval95% CI −9.8 to −0.3Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File points. Unlike presidential voting, in 1978 to 1990 the both group is not above cold only for either: 89.1%89.1% percent of the groupEstimandweighted share, party-line House vote, group: both (95% Wilson interval)Interval95% CI 86.3% to 91.4%n571Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File against 89.9%89.9% percent of the groupEstimandweighted share, party-line House vote, group: cold only (95% Wilson interval)Interval95% CI 85.5% to 93.0%n247Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File for House voting and 87.3%87.3% percent of the groupEstimandweighted share, straight-ticket vote, group: both (95% Wilson interval)Interval95% CI 83.1% to 90.5%n306Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 1990SourceANES Cumulative Data File against 88.6%88.6% percent of the groupEstimandweighted share, straight-ticket vote, group: cold only (95% Wilson interval)Interval95% CI 81.1% to 93.3%n105Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 1990SourceANES Cumulative Data File for the straight ticket. Reported turnout and campaign activity differ. In 1978 to 1990 warm-only partisans reported voting more often than cold-only partisans, with a difference of −5.5 (95% CI −10.1 to −0.8)−5.5 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, turnout (95% interval, independent groups)Interval95% CI −10.1 to −0.8n3,877Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 1990SourceANES Cumulative Data File points, and in 2012 to 2024 cold-only partisans reported it more often, by 5.7 (95% CI 3.2 to 8.3)5.7 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, turnout (95% interval, independent groups)Interval95% CI 3.2 to 8.3n7,625Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024SourceANES Cumulative Data File. For campaign activity the difference was 3.5 (95% CI −1.3 to 8.3)3.5 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, any campaign activity (95% interval, independent groups)Interval95% CI −1.3 to 8.3n3,848Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990, an interval that includes zero, and 5.4 (95% CI 2.0 to 8.9)5.4 percentage pointsEstimanddifference between the share among partisans very cold toward the other party only and the share among those very warm toward their own party only, any campaign activity (95% interval, independent groups)Interval95% CI 2.0 to 8.9n7,521Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024SourceANES Cumulative Data File in 2012 to 2024. For House voting, reported turnout and campaign activity the 1978 to 1990 cells include the midterm studies.

    The cold-only group was a small part of the partisans in 1978 to 1990, so its shares carry wide intervals, and the groups say nothing about partisans between the cut-offs, which the next section uses.

    In 2012 to 2024, per rating point, coldness goes with more House voting, reported turnout and campaign activity than warmth does

    A group comparison discards the variation inside each group. For each behavior and period, Figure 3 uses the whole rating. It reports the coefficients on coldness and warmth from one 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., in points of the behavior per 1010 points on the feeling thermometerEstimandsize of the change in a rating that a coefficient describes (a definition, not an estimate)Weightsnone (constant)Populationthe unit in which coefficients are reportedPeriod1978 to 2024SourceANES Cumulative Data File points of rating, with each rating held fixed when the other is estimated. The correlation between the own-party rating and the other-party rating, each centered on its survey year, was −0.06 (95% CI −0.08 to −0.04)−0.06 correlation coefficientEstimandweighted correlation between the rating of the own party and the rating of the other party, each centered on its survey-year mean (Fisher z interval on the Kish effective sample size)Interval95% CI −0.08 to −0.04n11,264Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990 and −0.23 (95% CI −0.25 to −0.21)−0.23 correlation coefficientEstimandweighted correlation between the rating of the own party and the rating of the other party, each centered on its survey-year mean (Fisher z interval on the Kish effective sample size)Interval95% CI −0.25 to −0.21n20,943Weightsweight (VCF0009z)Populationpartisans, leaners included, with a rating of both partiesPeriod2012 to 2024SourceANES Cumulative Data File in 2012 to 2024, so the two feelings are now more closely tied.

    Figure 3

    Per rating point, coldness toward the other party went with more House voting, reported turnout and campaign activity than warmth toward the own party did in 2012 to 2024, while in 1978 to 1990 the two went with similar amounts of House voting and campaign activity

    Percentage points of each behavior per 10 points on the 0 to 100 thermometer, from one linear probability model per behavior and period with both ratings and year fixed effects, partisans including leaners

    • Warmth toward own party
    • Coldness toward other party
    Show data
    Data for Per rating point, coldness toward the other party went with more House voting, reported turnout and campaign activity than warmth toward the own party did in 2012 to 2024, while in 1978 to 1990 the two went with similar amounts of House voting and campaign activity
    BehaviorSampleWarmth coefficientWarmth, lower endWarmth, upper endColdness coefficientColdness, lower endColdness, upper endColdness minus warmthDifference, lower endDifference, upper endRespondents
    Party-line presidential vote1978 to 19904.273.545.014.173.644.71−0.10−1.06+0.873,090
    Party-line House vote1978 to 19902.922.293.542.952.473.43+0.03−0.79+0.865,930
    Straight ticket1978 to 19902.051.113.002.832.103.56+0.78−0.50+2.062,662
    Reported turnout1978 to 19902.151.642.661.080.661.51−1.07−1.76−0.3810,692
    Campaign activity1978 to 19901.941.432.451.961.522.40+0.02−0.67+0.7010,606
    Party-line House vote1978 to 1990, midterm years3.112.263.952.982.333.63−0.12−1.23+0.993,138
    Reported turnout1978 to 1990, midterm years2.331.643.020.880.291.46−1.45−2.37−0.546,348
    Campaign activity1978 to 1990, midterm years1.681.032.331.520.952.08−0.16−1.02+0.696,321
    Party-line presidential vote1978 to 1990, presidential years4.273.545.014.173.644.71−0.10−1.06+0.873,090
    Party-line House vote1978 to 1990, presidential years2.711.793.642.932.233.64+0.22−1.03+1.462,792
    Straight ticket1978 to 1990, presidential years2.051.113.002.832.103.56+0.78−0.50+2.062,662
    Reported turnout1978 to 1990, presidential years1.861.112.621.390.791.99−0.47−1.51+0.574,344
    Campaign activity1978 to 1990, presidential years2.261.433.082.541.843.23+0.28−0.85+1.414,285
    Party-line presidential vote1992 to 20083.673.094.263.292.853.74−0.38−1.13+0.375,400
    Party-line House vote1992 to 20083.112.483.733.342.863.81+0.23−0.61+1.076,271
    Straight ticket1992 to 20082.842.103.593.312.753.87+0.46−0.56+1.484,528
    Reported turnout1992 to 20081.340.781.912.411.952.88+1.07+0.29+1.8610,010
    Campaign activity1992 to 20081.861.272.463.553.054.06+1.69+0.88+2.509,997
    Party-line presidential vote2012 to 20242.882.553.203.623.273.96+0.74+0.25+1.2315,612
    Party-line House vote2012 to 20241.921.552.293.483.073.88+1.56+0.99+2.1313,912
    Straight ticket2012 to 20241.210.851.572.552.162.93+1.34+0.77+1.9013,435
    Reported turnout2012 to 20240.680.301.062.251.872.64+1.58+0.98+2.1718,976
    Campaign activity2012 to 20242.021.542.503.563.094.04+1.54+0.79+2.3018,697
    In the first view the solid dot (coldness toward the other party) lies to the right of the ring (warmth toward the own party) in all five rows, and the difference is largest for House voting, reported turnout and campaign activity and smallest for presidential voting. In the 1978 to 1990 view the difference has an interval that includes zero in every row except turnout, where the ring lies to the right. The last two views split 1978 to 1990 by election type: the turnout ring lies to the right in the midterm studies and the difference includes zero in the presidential studies. Presidential voting and the straight ticket exist only in presidential years. A row's right-hand column is coldness minus warmth. Coldness is 100 minus the rating of the other party.Note: Intervals are 95% intervals from HC3 robust standard errors (a linear probability model with year fixed effects) and treat the weighted sample as a simple random sample.Note: Presidential voting is among partisans who voted for a major-party candidate. Turnout and campaign activity are self-reports. The 1978 to 1990 period has midterm studies in 1978, 1982, 1986 and 1990; 2012 to 2024 has none.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z)

    In 2012 to 2024 the coefficient on coldness exceeds the coefficient on warmth for all five behaviors. The per-point difference is 1.6 (95% CI 1.0 to 2.1)1.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.0 to 2.1n13,912Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File points for House voting, 1.6 (95% CI 1.0 to 2.2)1.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.0 to 2.2n18,976Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024SourceANES Cumulative Data File for reported turnout, 1.5 (95% CI 0.8 to 2.3)1.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.8 to 2.3n18,697Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024SourceANES Cumulative Data File for campaign activity, 1.3 (95% CI 0.8 to 1.9)1.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on straight-ticket vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.8 to 1.9n13,435Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024SourceANES Cumulative Data File for the straight ticket and 0.7 (95% CI 0.3 to 1.2)0.7 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.3 to 1.2n15,612Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File for presidential voting, and every interval is above zero. That period includes 2020, which has no mode weights. In 2012, 2016 and 2024 pooled the presidential difference is 0.54 (95% CI −0.05 to 1.12)0.54 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.05 to 1.12n10,032Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012, 2016 and 2024, all interviewsSourceANES Cumulative Data File, an interval that includes zero, and the other four stay above zero, with campaign activity at 1.14 (95% CI 0.23 to 2.05)1.14 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.23 to 2.05n12,248Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012, 2016 and 2024, all interviewsSourceANES Cumulative Data File.

    In 1978 to 1990 the difference had an interval that included zero for four behaviors: 0.0 (95% CI −0.8 to 0.9)0.0 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.8 to 0.9n5,930Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File for House voting, 0.0 (95% CI −0.7 to 0.7)0.0 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.7 to 0.7n10,606Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 1990SourceANES Cumulative Data File for campaign activity, 0.8 (95% CI −0.5 to 2.1)0.8 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on straight-ticket vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.5 to 2.1n2,662Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 1990SourceANES Cumulative Data File for the straight ticket and −0.1 (95% CI −1.1 to 0.9)−0.1 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −1.1 to 0.9n3,090Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 1990SourceANES Cumulative Data File for presidential voting. Reported turnout is the exception, at −1.1 (95% CI −1.8 to −0.4)−1.1 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −1.8 to −0.4n10,692Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 1990SourceANES Cumulative Data File points, so warmth toward the own party went with more reported turnout than coldness toward the other party did. That result comes from the midterm studies: the difference was −1.5 (95% CI −2.4 to −0.5)−1.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −2.4 to −0.5n6,348Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 1990, midterm years onlySourceANES Cumulative Data File in 1978, 1982, 1986 and 1990 and −0.5 (95% CI −1.5 to 0.6)−0.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −1.5 to 0.6n4,344Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 1990, presidential election years onlySourceANES Cumulative Data File in 1980, 1984 and 1988, an interval that includes zero. For House voting and campaign activity the presidential years alone also include zero, at 0.2 (95% CI −1.0 to 1.5)0.2 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −1.0 to 1.5n2,792Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990, presidential election years onlySourceANES Cumulative Data File and 0.3 (95% CI −0.8 to 1.4)0.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.8 to 1.4n4,285Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 1990, presidential election years onlySourceANES Cumulative Data File.

    The middle period separates the behaviors. In 1992 to 2008 the difference for reported turnout was 1.1 (95% CI 0.3 to 1.9)1.1 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.3 to 1.9n10,010Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1992 to 2008SourceANES Cumulative Data File and for campaign activity 1.7 (95% CI 0.9 to 2.5)1.7 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.9 to 2.5n9,997Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1992 to 2008SourceANES Cumulative Data File, both above zero, while the differences for House voting (0.2 (95% CI −0.6 to 1.1)0.2 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.6 to 1.1n6,271Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1992 to 2008SourceANES Cumulative Data File) and presidential voting (−0.4 (95% CI −1.1 to 0.4)−0.4 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −1.1 to 0.4n5,400Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1992 to 2008SourceANES Cumulative Data File) still included zero. The edge for coldness reached participation first and party-line voting later.

    The yearly differences are noisy, and 2020 stands out

    Figure 4 fits the same model to one survey at a time. The yearly differences scatter around zero before the 1990s. The interval lies above zero in 1414 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File of 7878 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File yearly estimates, all from 1992 on, and below zero in 33 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File, all before 1998. In 2012, 2016, 2020 and 2024 the point estimate is above zero in every panel.

    Figure 4

    The edge of coldness over warmth rose by 0.3 to 0.5 points per decade for presidential voting, House voting and reported turnout, with intervals above zero, and the yearly estimates are too noisy to date the rise

    Coldness coefficient minus warmth coefficient, percentage points of each behavior per 10 rating points, one model per survey year, partisans including leaners, 1978 to 2024

    Coldness minus warmth (points)

    Party-line presidential voteParty-line House voteStraight ticketReported turnoutCampaign activity
    Show data
    Data for The edge of coldness over warmth rose by 0.3 to 0.5 points per decade for presidential voting, House voting and reported turnout, with intervals above zero, and the yearly estimates are too noisy to date the rise
    behaviordiff_conf_highdiff_conf_lowdiff_estimatenyear
    Party-line presidential vote1.44−2.55−0.5557731980
    Party-line presidential vote2.26−0.6600.7991,2301984
    Party-line presidential vote0.953−2.26−0.6551,0871988
    Party-line presidential vote1.46−1.030.2141,2371992
    Party-line presidential vote−0.690−4.24−2.479791996
    Party-line presidential vote1.30−2.58−0.6391,0262000
    Party-line presidential vote1.37−3.15−0.8897542004
    Party-line presidential vote2.14−0.8290.6551,4042008
    Party-line presidential vote1.37−0.7580.3073,8382012
    Party-line presidential vote1.91−0.0530.9302,3812016
    Party-line presidential vote2.040.2591.155,5802020
    Party-line presidential vote1.44−0.4860.4753,8132024
    Party-line House vote1.87−2.95−0.5408561978
    Party-line House vote1.36−3.25−0.9477651980
    Party-line House vote2.70−2.080.3146541982
    Party-line House vote2.93−1.240.8431,0691984
    Party-line House vote1.31−2.95−0.8218951986
    Party-line House vote2.85−1.380.7329581988
    Party-line House vote2.62−1.350.6377331990
    Party-line House vote0.847−2.63−0.8911,2481992
    Party-line House vote1.43−3.16−0.8648781994
    Party-line House vote2.40−2.040.1809641996
    Party-line House vote6.000.5583.285161998
    Party-line House vote3.03−1.830.6048362000
    Party-line House vote3.47−1.640.9146642004
    Party-line House vote2.30−1.810.2441,1652008
    Party-line House vote2.580.2621.423,3212012
    Party-line House vote2.830.0771.452,1472016
    Party-line House vote3.291.262.284,9292020
    Party-line House vote1.69−0.4650.6133,5152024
    Straight ticket2.13−2.58−0.2276831980
    Straight ticket3.65−0.7021.471,0351984
    Straight ticket3.11−1.260.9259441988
    Straight ticket2.33−1.320.5051,0431992
    Straight ticket0.642−4.11−1.748781996
    Straight ticket4.85−0.1512.358152000
    Straight ticket3.03−1.950.5376542004
    Straight ticket2.91−1.540.6861,1382008
    Straight ticket2.710.4941.603,2392012
    Straight ticket2.55−0.2781.131,9872016
    Straight ticket2.770.6771.734,7932020
    Straight ticket1.53−0.4580.5393,4162024
    Reported turnout2.98−0.6831.151,7221978
    Reported turnout3.20−0.4941.361,1561980
    Reported turnout−1.15−5.58−3.361,1741982
    Reported turnout0.446−2.99−1.271,6731984
    Reported turnout−1.15−4.68−2.911,7961986
    Reported turnout0.635−3.08−1.221,5151988
    Reported turnout0.269−3.03−1.381,6561990
    Reported turnout4.331.602.961,9111992
    Reported turnout3.07−1.370.8521,5671994
    Reported turnout1.16−3.21−1.021,3731996
    Reported turnout4.05−0.8341.611,1021998
    Reported turnout2.27−2.85−0.2931,3252000
    Reported turnout7.191.954.579342004
    Reported turnout0.865−2.48−0.8081,7982008
    Reported turnout3.581.282.434,7472012
    Reported turnout2.43−0.1431.143,1302016
    Reported turnout2.580.5031.546,4982020
    Reported turnout2.46−0.2681.094,6012024
    Campaign activity3.00−0.5381.231,7191978
    Campaign activity3.27−0.8931.191,1491980
    Campaign activity1.11−3.21−1.051,1711982
    Campaign activity2.40−1.500.4491,6271984
    Campaign activity0.826−2.50−0.8391,7771986
    Campaign activity1.18−2.56−0.6921,5091988
    Campaign activity1.20−1.70−0.2511,6541990
    Campaign activity3.950.6352.291,9091992
    Campaign activity3.03−1.140.9491,5581994
    Campaign activity3.87−0.4211.721,3721996
    Campaign activity4.19−0.0662.061,1031998
    Campaign activity3.98−0.8501.561,3252000
    Campaign activity4.87−1.021.929342004
    Campaign activity3.44−0.9181.261,7962008
    Campaign activity3.240.2281.734,7312012
    Campaign activity2.66−0.3241.173,1062016
    Campaign activity3.711.042.386,4492020
    Campaign activity2.11−1.340.3864,4112024
    The yearly differences scatter around zero before the 1990s and are positive in every panel in 2012, 2016, 2020 and 2024. The interval lies above zero in 14 of 78 yearly estimates, all from 1992 on, and below zero in 3, all before 1998. Each point is the difference between the coldness and warmth coefficients of a linear probability model fitted to one year, with a 95% interval. Behaviors that the ANES did not ask in a year, and years with fewer than 150 respondents, have no point. Panels share the vertical scale. The per-decade rises in the title come from a model that lets both coefficients change linearly with time; the intervals for the straight ticket and for campaign activity are given in the text.Note: Intervals are 95% intervals from HC3 robust standard errors (a linear probability model fitted to one year at a time) and treat the weighted sample as a simple random sample.Note: Presidential vote and straight-ticket voting exist only in presidential election years; House vote, turnout and campaign activity also exist in midterm years.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z)

    A model that lets each coefficient change linearly with time summarizes the movement. The coldness-minus-warmth difference rose by 0.5 (95% CI 0.3 to 0.8)0.5 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.3 to 0.8n39,678Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File points per decade for reported turnout, 0.5 (95% CI 0.2 to 0.7)0.5 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.2 to 0.7n26,113Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File for House voting and 0.3 (95% CI 0.1 to 0.6)0.3 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.1 to 0.6n24,102Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File for presidential voting, with intervals above zero. For the straight ticket the rise was 0.2 (95% CI −0.1 to 0.6)0.2 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.1 to 0.6n20,625Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File, an interval that includes zero, and for campaign activity 0.3 (95% CI 0.0 to 0.6)0.3 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.0 to 0.6n39,300Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File, an interval whose lower end reaches zero at one decimal. Fitted to presidential years alone, the rise is 0.5 (95% CI 0.1 to 0.8)0.5 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.1 to 0.8n21,581Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1980 to 2024, presidential election years onlySourceANES Cumulative Data File for House voting, an interval above zero, and 0.30.3 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness-minus-warmth difference, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.0 to 0.6n30,661Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1980 to 2024, presidential election years onlySourceANES Cumulative Data File for reported turnout, an interval whose lower end is just below zero. The yearly estimates are too wide to say whether the rise was gradual or sudden.

    The 2020 survey is the only one in which the interval is above zero in all five panels. In 2024 the point estimate is lower than in 2020 in every panel and every interval includes zero: 2.3 (95% CI 1.3 to 3.3)2.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, one linear probability model for the survey year (95% interval, HC3)Interval95% CI 1.3 to 3.3n4,929Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2020SourceANES Cumulative Data File in 2020 and 0.6 (95% CI −0.5 to 1.7)0.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, one linear probability model for the survey year (95% interval, HC3)Interval95% CI −0.5 to 1.7n3,515Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2024SourceANES Cumulative Data File in 2024 for House voting, and 2.4 (95% CI 1.0 to 3.7)2.4 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, one linear probability model for the survey year (95% interval, HC3)Interval95% CI 1.0 to 3.7n6,449Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2020SourceANES Cumulative Data File and 0.4 (95% CI −1.3 to 2.1)0.4 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, one linear probability model for the survey year (95% interval, HC3)Interval95% CI −1.3 to 2.1n4,411Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2024SourceANES Cumulative Data File for campaign activity. The 2020 study has no mode weights and the 2024 sample mixes new face-to-face, new web and panel respondents, so the figure cannot say whether the lower 2024 estimates reflect a change in feelings or in who was interviewed.

    Face-to-face and web respondents show the same sign

    The 2012 to 2024 sample mixes interview modes, and web respondents are more often cold only than face-to-face respondents (Figure 1). Figure 5 restricts the sample to 2012, 2016 and 2024, the years with mode weights, and estimates each mode alone. In both modes the per-point difference is positive for all five behaviors. For House voting it is 1.5 (95% CI 0.2 to 2.9)1.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.2 to 2.9n2,283Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012, 2016 and 2024, face-to-face interviewsSourceANES Cumulative Data File points face to face and 1.2 (95% CI 0.4 to 2.1)1.2 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.4 to 2.1n4,990Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012, 2016 and 2024, web interviewsSourceANES Cumulative Data File on the web. For reported turnout it is 2.4 (95% CI 1.1 to 3.7)2.4 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.1 to 3.7n3,476Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, face-to-face interviewsSourceANES Cumulative Data File and 1.3 (95% CI 0.3 to 2.2)1.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.3 to 2.2n6,861Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, web interviewsSourceANES Cumulative Data File, with intervals above zero in both modes. For campaign activity it is 1.6 (95% CI −0.1 to 3.2)1.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.1 to 3.2n3,465Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012, 2016 and 2024, face-to-face interviewsSourceANES Cumulative Data File and 1.1 (95% CI −0.1 to 2.2)1.1 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.1 to 2.2n6,822Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012, 2016 and 2024, web interviewsSourceANES Cumulative Data File, and for presidential voting 0.9 (95% CI −0.4 to 2.3)0.9 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.4 to 2.3n2,620Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012, 2016 and 2024, face-to-face interviewsSourceANES Cumulative Data File and 0.5 (95% CI −0.2 to 1.1)0.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.2 to 1.1n5,586Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012, 2016 and 2024, web interviewsSourceANES Cumulative Data File, with intervals that include zero in both modes for both behaviors.

    Figure 5

    In 2012, 2016 and 2024 pooled, the coldness-minus-warmth difference has the same sign in face-to-face and web interviews for all five behaviors, and the two intervals overlap in every row

    Coldness coefficient minus warmth coefficient, percentage points of each behavior per 10 rating points, face-to-face and web respondents separately, partisans including leaners

    • Face to face
    • Web
    Show data
    Data for In 2012, 2016 and 2024 pooled, the coldness-minus-warmth difference has the same sign in face-to-face and web interviews for all five behaviors, and the two intervals overlap in every row
    behaviorftfftf_hiftf_lon_ftfn_webperiodwebweb_hiweb_loyear
    Party-line presidential vote1.173.61−1.281,2502,5882012−0.2160.705−1.142012
    Party-line House vote2.544.740.3461,0362,28520120.9572.24−0.3302012
    Straight ticket0.3062.79−2.181,0122,22720122.363.421.292012
    Reported turnout3.094.961.231,6853,06220121.763.230.2832012
    Campaign activity2.505.04−0.0371,6843,04720121.443.36−0.4902012
    Party-line presidential vote0.6532.31−1.016961,68520161.112.32−0.0892016
    Party-line House vote0.6632.65−1.336531,49420162.033.820.2362016
    Straight ticket0.7232.87−1.425881,39920161.513.31−0.2942016
    Reported turnout1.974.00−0.0639312,19920160.8872.48−0.7072016
    Campaign activity1.664.47−1.159262,18020160.9192.69−0.8552016
    Party-line presidential vote0.9343.17−1.306741,31320240.6832.02−0.6512024
    Party-line House vote0.8673.80−2.065941,21120240.3821.82−1.052024
    Straight ticket1.504.12−1.115711,18520240.1961.62−1.232024
    Reported turnout1.324.60−1.978601,60020241.122.79−0.5452024
    Campaign activity−0.8452.49−4.188551,59520240.4763.23−2.282024
    Party-line presidential vote0.9362.31−0.4372,6205,5862012, 2016 and 2024 pooled0.4651.12−0.184pooled
    Party-line House vote1.532.910.1512,2834,9902012, 2016 and 2024 pooled1.252.140.359pooled
    Straight ticket0.5242.01−0.9642,1714,8112012, 2016 and 2024 pooled1.582.420.740pooled
    Reported turnout2.413.681.133,4766,8612012, 2016 and 2024 pooled1.252.170.329pooled
    Campaign activity1.563.21−0.0863,4656,8222012, 2016 and 2024 pooled1.062.24−0.122pooled
    Each row joins the face-to-face difference (ring) to the web difference (solid dot). The pooled view has positive differences in both modes for every behavior, with intervals above zero for House voting and turnout in both modes and for straight-ticket voting on the web. Within a single year the intervals are wide and a few differences are negative, and the year views are descriptive. The 2020 study has no mode weights and is left out. Each dot has a 95% interval.Note: Mode groups are the fresh face-to-face and fresh web samples with their own weights. In 2024 the panel and paper respondents belong to neither group and are not in this figure. Pooled estimates include year fixed effects.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z; weight_ftf and weight_web for the mode estimates of 2008, 2012, 2016 and 2024)

    Only the change in reported turnout holds in presidential years alone and within each interview mode

    Figure 6 sets the 2012 to 2024 per-point difference against the 1978 to 1990 difference. In all interviews the change was 2.6 (95% CI 1.7 to 3.6)2.6 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 1.7 to 3.6Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File points for reported turnout, 1.5 (95% CI 0.5 to 2.5)1.5 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.5 to 2.5Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File for House voting and 1.5 (95% CI 0.5 to 2.5)1.5 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.5 to 2.5Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File for campaign activity, with intervals above zero, and 0.8 (95% CI −0.2 to 1.9)0.8 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.2 to 1.9Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File for presidential voting and 0.6 (95% CI −0.8 to 2.0)0.6 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.8 to 2.0Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File for the straight ticket, with intervals that include zero. The interval for the presidential change reaches 1.91.9 percentage points of the outcome per 10 rating pointsEstimandupper end of the 95% interval for the change in the coldness-minus-warmth difference between the two periods (independent samples)Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File points, more than the House point estimate, so the data cannot rule out a presidential change of that size.

    Two comparisons narrow the list. The first holds election type fixed by setting 2012 to 2024 against 1980, 1984 and 1988 alone. Reported turnout changed by 2.0 (95% CI 0.8 to 3.3)2.0 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.8 to 3.3Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File points, still above zero, while House voting changed by 1.31.3 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.0 to 2.7Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File and campaign activity by 1.3 (95% CI −0.1 to 2.6)1.3 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.1 to 2.6Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024 minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File, intervals that include zero. The second holds interview mode fixed. The 1978 to 1990 studies were interviewed mainly in person, so the face-to-face estimate is the closer like-for-like comparison. Among face-to-face respondents in 2012, 2016 and 2024 reported turnout changed by 3.5 (95% CI 2.0 to 4.9)3.5 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 2.0 to 4.9Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, face-to-face interviews minus 1978 to 1990SourceANES Cumulative Data File points and among web respondents by 2.3 (95% CI 1.2 to 3.5)2.3 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 1.2 to 3.5Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, web interviews minus 1978 to 1990SourceANES Cumulative Data File, both above zero, and both stay above zero against the presidential years alone (2.9 (95% CI 1.2 to 4.5)2.9 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 1.2 to 4.5Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, face-to-face interviews minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File and 1.7 (95% CI 0.3 to 3.1)1.7 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.3 to 3.1Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who answered the turnout questionPeriod2012, 2016 and 2024, web interviews minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File). House voting changed by 1.5 (95% CI −0.1 to 3.1)1.5 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.1 to 3.1Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012, 2016 and 2024, face-to-face interviews minus 1978 to 1990SourceANES Cumulative Data File face to face and 1.21.2 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI 0.0 to 2.4Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012, 2016 and 2024, web interviews minus 1978 to 1990SourceANES Cumulative Data File on the web, and campaign activity by 1.5 (95% CI −0.2 to 3.3)1.5 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.2 to 3.3Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012, 2016 and 2024, face-to-face interviews minus 1978 to 1990SourceANES Cumulative Data File and 1.0 (95% CI −0.3 to 2.4)1.0 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.3 to 2.4Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012, 2016 and 2024, web interviews minus 1978 to 1990SourceANES Cumulative Data File, and those intervals include zero or reach it.

    Figure 6

    Per rating point, the change since 1978 to 1990 in the difference between coldness and warmth is above zero for reported turnout in all interviews, in presidential years alone and within each interview mode, and for House voting and campaign activity only in all interviews

    Coldness minus warmth coefficient in a later sample minus the same difference in 1978 to 1990, percentage points of each behavior per 10 rating points, partisans including leaners

    • 2012 to 2024, all interviews
    • 2012, 2016 and 2024, face to face
    • 2012, 2016 and 2024, web
    • 1992 to 2008, all interviews
    Show data
    Data for Per rating point, the change since 1978 to 1990 in the difference between coldness and warmth is above zero for reported turnout in all interviews, in presidential years alone and within each interview mode, and for House voting and campaign activity only in all interviews
    behaviorconf_highconf_lowestimatesample_labelview
    Party-line presidential vote0.087−0.111−0.0122012 to 2024, all interviewslogodds_all
    Party-line presidential vote0.285−0.1140.0862012, 2016 and 2024, face to facelogodds_all
    Party-line presidential vote0.069−0.149−0.0402012, 2016 and 2024, weblogodds_all
    Party-line presidential vote0.080−0.151−0.0351992 to 2008, all interviewslogodds_all
    Party-line House vote0.1820.0440.1132012 to 2024, all interviewslogodds_all
    Party-line House vote0.3120.0030.1572012, 2016 and 2024, face to facelogodds_all
    Party-line House vote0.175−0.0060.0852012, 2016 and 2024, weblogodds_all
    Party-line House vote0.090−0.0530.0181992 to 2008, all interviewslogodds_all
    Straight ticket0.161−0.0160.0722012 to 2024, all interviewslogodds_all
    Straight ticket0.179−0.1540.0122012, 2016 and 2024, face to facelogodds_all
    Straight ticket0.199−0.0130.0932012, 2016 and 2024, weblogodds_all
    Straight ticket0.097−0.0960.0011992 to 2008, all interviewslogodds_all
    Reported turnout0.2050.1010.1532012 to 2024, all interviewslogodds_all
    Reported turnout0.2800.1100.1952012, 2016 and 2024, face to facelogodds_all
    Reported turnout0.1960.0590.1282012, 2016 and 2024, weblogodds_all
    Reported turnout0.1590.0550.1071992 to 2008, all interviewslogodds_all
    Campaign activity0.1070.0200.0632012 to 2024, all interviewslogodds_all
    Campaign activity0.141−0.0100.0662012, 2016 and 2024, face to facelogodds_all
    Campaign activity0.101−0.0140.0432012, 2016 and 2024, weblogodds_all
    Campaign activity0.1200.0260.0731992 to 2008, all interviewslogodds_all
    Party-line presidential vote0.087−0.111−0.0122012 to 2024, all interviewslogodds_pres
    Party-line presidential vote0.285−0.1140.0862012, 2016 and 2024, face to facelogodds_pres
    Party-line presidential vote0.069−0.149−0.0402012, 2016 and 2024, weblogodds_pres
    Party-line presidential vote0.080−0.151−0.0351992 to 2008, all interviewslogodds_pres
    Party-line House vote0.1880.0190.1042012 to 2024, all interviewslogodds_pres
    Party-line House vote0.310−0.0140.1482012, 2016 and 2024, face to facelogodds_pres
    Party-line House vote0.178−0.0280.0752012, 2016 and 2024, weblogodds_pres
    Party-line House vote0.088−0.097−0.0041992 to 2008, all interviewslogodds_pres
    Straight ticket0.161−0.0160.0722012 to 2024, all interviewslogodds_pres
    Straight ticket0.179−0.1540.0122012, 2016 and 2024, face to facelogodds_pres
    Straight ticket0.199−0.0130.0932012, 2016 and 2024, weblogodds_pres
    Straight ticket0.097−0.0960.0011992 to 2008, all interviewslogodds_pres
    Reported turnout0.1960.0580.1272012 to 2024, all interviewslogodds_pres
    Reported turnout0.2660.0730.1692012, 2016 and 2024, face to facelogodds_pres
    Reported turnout0.1840.0200.1022012, 2016 and 2024, weblogodds_pres
    Reported turnout0.1620.0100.0861992 to 2008, all interviewslogodds_pres
    Campaign activity0.109−0.0040.0522012 to 2024, all interviewslogodds_pres
    Campaign activity0.139−0.0290.0552012, 2016 and 2024, face to facelogodds_pres
    Campaign activity0.100−0.0360.0322012, 2016 and 2024, weblogodds_pres
    Campaign activity0.1260.0010.0631992 to 2008, all interviewslogodds_pres
    Party-line presidential vote1.92−0.2440.8372012 to 2024, all interviewspp_all
    Party-line presidential vote2.71−0.6451.032012, 2016 and 2024, face to facepp_all
    Party-line presidential vote1.72−0.6010.5622012, 2016 and 2024, webpp_all
    Party-line presidential vote0.938−1.50−0.2831992 to 2008, all interviewspp_all
    Party-line House vote2.530.5161.522012 to 2024, all interviewspp_all
    Party-line House vote3.10−0.1141.492012, 2016 and 2024, face to facepp_all
    Party-line House vote2.43−0.0021.212012, 2016 and 2024, webpp_all
    Party-line House vote1.37−0.9840.1951992 to 2008, all interviewspp_all
    Straight ticket1.96−0.8450.5572012 to 2024, all interviewspp_all
    Straight ticket1.71−2.22−0.2542012, 2016 and 2024, face to facepp_all
    Straight ticket2.34−0.7300.8042012, 2016 and 2024, webpp_all
    Straight ticket1.32−1.95−0.3141992 to 2008, all interviewspp_all
    Reported turnout3.561.732.642012 to 2024, all interviewspp_all
    Reported turnout4.932.023.472012, 2016 and 2024, face to facepp_all
    Reported turnout3.471.162.322012, 2016 and 2024, webpp_all
    Reported turnout3.181.092.141992 to 2008, all interviewspp_all
    Campaign activity2.540.5141.532012 to 2024, all interviewspp_all
    Campaign activity3.33−0.2381.552012, 2016 and 2024, face to facepp_all
    Campaign activity2.41−0.3211.042012, 2016 and 2024, webpp_all
    Campaign activity2.740.6141.671992 to 2008, all interviewspp_all
    Party-line presidential vote1.92−0.2440.8372012 to 2024, all interviewspp_pres
    Party-line presidential vote2.71−0.6451.032012, 2016 and 2024, face to facepp_pres
    Party-line presidential vote1.72−0.6010.5622012, 2016 and 2024, webpp_pres
    Party-line presidential vote0.938−1.50−0.2831992 to 2008, all interviewspp_pres
    Party-line House vote2.71−0.0311.342012 to 2024, all interviewspp_pres
    Party-line House vote3.17−0.5471.312012, 2016 and 2024, face to facepp_pres
    Party-line House vote2.56−0.5001.032012, 2016 and 2024, webpp_pres
    Party-line House vote1.37−1.77−0.2031992 to 2008, all interviewspp_pres
    Straight ticket1.96−0.8450.5572012 to 2024, all interviewspp_pres
    Straight ticket1.71−2.22−0.2542012, 2016 and 2024, face to facepp_pres
    Straight ticket2.34−0.7300.8042012, 2016 and 2024, webpp_pres
    Straight ticket1.32−1.95−0.3141992 to 2008, all interviewspp_pres
    Reported turnout3.250.8482.052012 to 2024, all interviewspp_pres
    Reported turnout4.531.232.882012, 2016 and 2024, face to facepp_pres
    Reported turnout3.120.3311.722012, 2016 and 2024, webpp_pres
    Reported turnout2.860.1211.491992 to 2008, all interviewspp_pres
    Campaign activity2.62−0.0921.262012 to 2024, all interviewspp_pres
    Campaign activity3.28−0.7161.282012, 2016 and 2024, face to facepp_pres
    Campaign activity2.41−0.8560.7792012, 2016 and 2024, webpp_pres
    Campaign activity2.980.0071.491992 to 2008, all interviewspp_pres
    The turnout row has every dot to the right of zero with an interval above zero, in the all-interview comparison, in the presidential-years-only comparison and within each interview mode. The House and campaign-activity rows have the all-interview dot above zero and, in the presidential-years-only view, an interval that includes zero; the face-to-face and web dots for those rows have intervals that include zero or reach it. Presidential voting and the straight ticket have intervals that include zero throughout. The presidential-years views compare 2012 to 2024 with 1980, 1984 and 1988 alone, because 1978 to 1990 includes midterm studies and 2012 to 2024 does not. The log-odds views put the same change on a scale that does not shrink as a behavior nears everyone. Each dot is the difference of two independent estimates with a 95% interval.Note: The 1978 to 1990 comparison sample has no mode weights and was interviewed mainly in person; mode estimates of the later period use weight_ftf and weight_web and cover 2012, 2016 and 2024 only. The 2020 study has no mode weights.Note: In the presidential-years views the earlier sample is 1980, 1984 and 1988. Presidential voting and the straight ticket exist only in presidential years, so their two baselines coincide.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z; weight_ftf and weight_web for the mode estimates of 2008, 2012, 2016 and 2024)

    Party-line voting also rose over these years, which compresses any coefficient on the probability scale. The share voting the party line was 77.1%77.1% percent of partisansEstimandweighted share acting (95% Wilson interval)Interval95% CI 76.0% to 78.1%n5,930Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File for the House in 1978 to 1990 and 90.0%90.0% percent of partisansEstimandweighted share acting (95% Wilson interval)Interval95% CI 89.3% to 90.7%n13,912Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File in 2012 to 2024, and 85.9%85.9% percent of partisansEstimandweighted share acting (95% Wilson interval)Interval95% CI 84.6% to 87.0%n3,090Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 1990SourceANES Cumulative Data File and 93.3%93.3% percent of partisansEstimandweighted share acting (95% Wilson interval)Interval95% CI 92.7% to 93.8%n15,612Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File for the presidency. The log-odds views of Figure 6 repeat the change on a scale where a constant proportional association gives a constant coefficient. On that scale the presidential difference was 0.09 (95% CI 0.01 to 0.17)0.09 log-odds of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, probability-weight sandwich)Interval95% CI 0.01 to 0.17n3,090Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990 and 0.08 (95% CI 0.02 to 0.14)0.08 log-odds of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, probability-weight sandwich)Interval95% CI 0.02 to 0.14n15,612Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File in 2012 to 2024, a change of −0.01 (95% CI −0.11 to 0.09)−0.01 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI −0.11 to 0.09Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File. Coldness went with more presidential voting than warmth in both periods, and the data show no change. The House change is 0.11 (95% CI 0.04 to 0.18)0.11 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI 0.04 to 0.18Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File in all interviews and 0.10 (95% CI 0.02 to 0.19)0.10 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI 0.02 to 0.19Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File in presidential years alone, both above zero, and 0.16 (95% CI 0.00 to 0.31)0.16 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI 0.00 to 0.31Weightsweight_ftf (VCF0009x)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012, 2016 and 2024, face-to-face interviews minus 1978 to 1990SourceANES Cumulative Data File face to face and 0.08 (95% CI −0.01 to 0.18)0.08 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI −0.01 to 0.18Weightsweight_web (VCF0009y)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012, 2016 and 2024, web interviews minus 1978 to 1990SourceANES Cumulative Data File on the web. For campaign activity it is 0.06 (95% CI 0.02 to 0.11)0.06 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI 0.02 to 0.11Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File in all interviews and 0.05 (95% CI 0.00 to 0.11)0.05 log-odds of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, log-odds coefficient in a weighted logistic regression with year fixed effects, per 10 rating points (95% interval, independent samples)Interval95% CI 0.00 to 0.11Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024 minus 1978 to 1990, presidential election years onlySourceANES Cumulative Data File in presidential years alone, an interval that reaches zero.

    The edge for coldness rests on different coefficients for votes and for participation

    The difference can rise because the coldness coefficient rises, because the warmth coefficient falls, or both. Interacting each coefficient with time separates them. For the three vote behaviors the warmth coefficient fell, by −0.4 (95% CI −0.6 to −0.2)−0.4 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the warmth coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.6 to −0.2n24,102Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File points per decade for presidential voting, −0.4 (95% CI −0.5 to −0.2)−0.4 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the warmth coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.5 to −0.2n26,113Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File for House voting and −0.4 (95% CI −0.6 to −0.2)−0.4 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the warmth coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.6 to −0.2n20,625Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File for the straight ticket, while the coldness coefficient changed by −0.1 (95% CI −0.3 to 0.1)−0.1 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.3 to 0.1n24,102Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File, 0.1 (95% CI −0.1 to 0.3)0.1 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.1 to 0.3n26,113Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File and −0.2 (95% CI −0.4 to 0.0)−0.2 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.4 to 0.0n20,625Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File, all with intervals that include zero. For reported turnout the coldness coefficient rose, by 0.2 (95% CI 0.1 to 0.4)0.2 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.1 to 0.4n39,678Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File points per decade, as the warmth coefficient fell, by −0.3 (95% CI −0.5 to −0.1)−0.3 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the warmth coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.5 to −0.1n39,678Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File. For campaign activity the coldness coefficient rose by 0.4 (95% CI 0.2 to 0.6)0.4 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the coldness coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI 0.2 to 0.6n39,300Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File and the warmth coefficient changed by 0.1 (95% CI −0.1 to 0.3)0.1 percentage points of the outcome per 10 rating points, per decadeEstimandchange per decade in the warmth coefficient, weighted linear probability model with year fixed effects and both feelings interacted with decades since 2001 (95% interval, HC3)Interval95% CI −0.1 to 0.3n39,300Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod1978 to 2024 (18 surveys)SourceANES Cumulative Data File, an interval that includes zero. In votes the association of warmth weakened and that of coldness held; in reported turnout and campaign activity the association with coldness grew.

    The sign of the recent difference holds under other measures, and the change does not

    The group comparison and the per-point comparison disagree about the 1980s, and Figure 7 shows that the measure matters. In 2012 to 2024 the coefficient on coldness exceeds the coefficient on warmth for all five behaviors under the four measures the figure defines, and all 2020 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File intervals are above zero. In 1978 to 1990 the measures disagree, and 1212 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File of the 2020 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File intervals include zero. The cut-off versions put the difference for House voting at 5.6 (95% CI 1.9 to 9.2)5.6 percentage points of the outcome between groupsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the strict cut-off, own party at or above it) (95% interval, HC3)Interval95% CI 1.9 to 9.2n5,930Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File points and for the straight ticket at 7.2 (95% CI 1.6 to 12.8)7.2 percentage points of the outcome between groupsEstimandcoldness coefficient minus warmth coefficient on straight-ticket vote, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the strict cut-off, own party at or above it) (95% interval, HC3)Interval95% CI 1.6 to 12.8n2,662Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 1990SourceANES Cumulative Data File in 1978 to 1990, above zero, as the group comparison above did, while the per-point difference is 0.0 (95% CI −0.8 to 0.9)0.0 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.8 to 0.9n5,930Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod1978 to 1990SourceANES Cumulative Data File and 0.8 (95% CI −0.5 to 2.1)0.8 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on straight-ticket vote, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.5 to 2.1n2,662Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod1978 to 1990SourceANES Cumulative Data File.

    The change since 1978 to 1990 also depends on the measure: 1212 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File of the 2020 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File change intervals are above zero, and which behaviors they cover differs by measure. Presidential voting changed by 6.1 (95% CI 2.5 to 9.7)6.1 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the strict cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI 2.5 to 9.7Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File points at the first pair of cut-offs and 8.4 (95% CI 4.0 to 12.8)8.4 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the loose cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI 4.0 to 12.8Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File at the looser pair, both above zero, against 0.8 (95% CI −0.2 to 1.9)0.8 percentage points of the outcome per 10 rating pointsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, independent samples)Interval95% CI −0.2 to 1.9Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File per point and 0.4 (95% CI −1.7 to 2.6)0.4 percentage points of the outcome per standard deviation of the ratingEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per within-year standard deviation (95% interval, independent samples)Interval95% CI −1.7 to 2.6Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File per standard deviation, both including zero. House voting changed by 2.7 (95% CI −1.4 to 6.8)2.7 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the strict cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI −1.4 to 6.8Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File at the first pair, an interval that includes zero, and 4.29 (95% CI 0.03 to 8.55)4.29 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the loose cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI 0.03 to 8.55Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File at the looser pair. The straight ticket changed by −0.3 (95% CI −6.2 to 5.7)−0.3 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the strict cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI −6.2 to 5.7Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File and −2.5 (95% CI −8.5 to 3.5)−2.5 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the loose cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI −8.5 to 3.5Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File, both including zero. Reported turnout is above zero under all four measures, at 5.0 (95% CI 3.2 to 6.8)5.0 percentage points of the outcome per standard deviation of the ratingEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings per within-year standard deviation (95% interval, independent samples)Interval95% CI 3.2 to 6.8Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File per standard deviation, 10.1 (95% CI 5.9 to 14.3)10.1 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the strict cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI 5.9 to 14.3Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File at the first pair and 9.1 (95% CI 5.2 to 13.1)9.1 percentage points of the outcome between groupsEstimandchange in the coldness-minus-warmth difference between the two periods, coefficient in a weighted linear probability model with year fixed effects, both feelings as indicators (other party rated at or below the loose cut-off, own party at or above it) (95% interval, independent samples)Interval95% CI 5.2 to 13.1Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File at the looser pair. The group comparison sides with the measures that show no growth for votes, because the gap between cold only and warm only fell for House voting and for the straight ticket (−5.0 (95% CI −9.8 to −0.3)−5.0 percentage pointsEstimandchange in the difference between the share among partisans cold only and the share among partisans warm only, party-line House vote (95% interval, independent samples)Interval95% CI −9.8 to −0.3Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File and −7.1 (95% CI −14.2 to 0.0)−7.1 percentage pointsEstimandchange in the difference between the share among partisans cold only and the share among partisans warm only, straight-ticket vote (95% interval, independent samples)Interval95% CI −14.2 to 0.0Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File points) and rose for reported turnout (11.2 (95% CI 5.9 to 16.5)11.2 percentage pointsEstimandchange in the difference between the share among partisans cold only and the share among partisans warm only, turnout (95% interval, independent samples)Interval95% CI 5.9 to 16.5Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024 minus 1978 to 1990SourceANES Cumulative Data File).

    Figure 7

    In 2012 to 2024 the coldness coefficient exceeds the warmth coefficient for all five behaviors under each of four measures, every one of the 20 intervals lies above zero, and the measures disagree about the change since 1978 to 1990

    Coldness coefficient minus warmth coefficient, partisans including leaners, by measure of the ratings, 2012 to 2024; units are percentage points per 10 rating points, per standard deviation, or between groups

    Per 10 rating points

    Per standard deviation

    Cut-offs 15 and 85

    Cut-offs 30 and 70

    Coldness minus warmth

    Show data
    Data for In 2012 to 2024 the coldness coefficient exceeds the warmth coefficient for all five behaviors under each of four measures, every one of the 20 intervals lies above zero, and the measures disagree about the change since 1978 to 1990
    behaviordiff_conf_highdiff_conf_lowdiff_estimatenperiodspec_label
    Party-line presidential vote0.867−1.06−0.0963,0901978 to 1990Per 10 rating points
    Party-line House vote0.864−0.7940.0355,9301978 to 1990Per 10 rating points
    Straight ticket2.06−0.5030.7792,6621978 to 1990Per 10 rating points
    Reported turnout−0.375−1.76−1.0710,6921978 to 1990Per 10 rating points
    Campaign activity0.699−0.6660.01610,6061978 to 1990Per 10 rating points
    Party-line presidential vote3.22−0.5071.363,0901978 to 1990Per standard deviation
    Party-line House vote2.57−0.5920.9885,9301978 to 1990Per standard deviation
    Straight ticket4.91−0.0712.422,6621978 to 1990Per standard deviation
    Reported turnout−0.213−2.86−1.5410,6921978 to 1990Per standard deviation
    Campaign activity1.96−0.6590.65010,6061978 to 1990Per standard deviation
    Party-line presidential vote3.67−3.040.3143,0901978 to 1990Cut-offs 15 and 85
    Party-line House vote9.191.955.575,9301978 to 1990Cut-offs 15 and 85
    Straight ticket12.841.627.232,6621978 to 1990Cut-offs 15 and 85
    Reported turnout−0.132−7.10−3.6110,6921978 to 1990Cut-offs 15 and 85
    Campaign activity4.87−2.601.1410,6061978 to 1990Cut-offs 15 and 85
    Party-line presidential vote1.60−6.40−2.403,0901978 to 1990Cut-offs 30 and 70
    Party-line House vote7.600.4904.055,9301978 to 1990Cut-offs 30 and 70
    Straight ticket14.443.458.952,6621978 to 1990Cut-offs 30 and 70
    Reported turnout0.437−5.65−2.6010,6921978 to 1990Cut-offs 30 and 70
    Campaign activity7.240.9214.0810,6061978 to 1990Cut-offs 30 and 70
    Party-line presidential vote0.370−1.13−0.3795,4001992 to 2008Per 10 rating points
    Party-line House vote1.07−0.6090.2296,2711992 to 2008Per 10 rating points
    Straight ticket1.48−0.5560.4644,5281992 to 2008Per 10 rating points
    Reported turnout1.860.2871.0710,0101992 to 2008Per 10 rating points
    Campaign activity2.500.8791.699,9971992 to 2008Per 10 rating points
    Party-line presidential vote1.81−1.080.3685,4001992 to 2008Per standard deviation
    Party-line House vote3.06−0.1581.456,2711992 to 2008Per standard deviation
    Straight ticket3.86−0.0801.894,5281992 to 2008Per standard deviation
    Reported turnout4.171.142.6510,0101992 to 2008Per standard deviation
    Campaign activity5.702.554.139,9971992 to 2008Per standard deviation
    Party-line presidential vote2.82−1.990.4125,4001992 to 2008Cut-offs 15 and 85
    Party-line House vote7.220.7583.996,2711992 to 2008Cut-offs 15 and 85
    Straight ticket8.320.6994.514,5281992 to 2008Cut-offs 15 and 85
    Reported turnout8.241.354.8010,0101992 to 2008Cut-offs 15 and 85
    Campaign activity14.156.3010.239,9971992 to 2008Cut-offs 15 and 85
    Party-line presidential vote2.57−3.01−0.2165,4001992 to 2008Cut-offs 30 and 70
    Party-line House vote4.73−2.091.326,2711992 to 2008Cut-offs 30 and 70
    Straight ticket6.96−1.352.814,5281992 to 2008Cut-offs 30 and 70
    Reported turnout8.162.035.1010,0101992 to 2008Cut-offs 30 and 70
    Campaign activity13.606.8810.249,9971992 to 2008Cut-offs 30 and 70
    Party-line presidential vote1.230.2510.74115,6122012 to 2024Per 10 rating points
    Party-line House vote2.130.9861.5613,9122012 to 2024Per 10 rating points
    Straight ticket1.900.7681.3413,4352012 to 2024Per 10 rating points
    Reported turnout2.170.9821.5818,9762012 to 2024Per 10 rating points
    Campaign activity2.300.7941.5418,6972012 to 2024Per 10 rating points
    Party-line presidential vote2.830.7671.8015,6122012 to 2024Per standard deviation
    Party-line House vote4.642.253.4513,9122012 to 2024Per standard deviation
    Straight ticket4.111.742.9313,4352012 to 2024Per standard deviation
    Reported turnout4.692.203.4418,9762012 to 2024Per standard deviation
    Campaign activity5.011.873.4418,6972012 to 2024Per standard deviation
    Party-line presidential vote7.795.016.4015,6122012 to 2024Cut-offs 15 and 85
    Party-line House vote10.236.408.3113,9122012 to 2024Cut-offs 15 and 85
    Straight ticket8.895.026.9613,4352012 to 2024Cut-offs 15 and 85
    Reported turnout8.874.156.5118,9762012 to 2024Cut-offs 15 and 85
    Campaign activity12.125.728.9218,6972012 to 2024Cut-offs 15 and 85
    Party-line presidential vote7.874.095.9815,6122012 to 2024Cut-offs 30 and 70
    Party-line House vote10.685.998.3413,9122012 to 2024Cut-offs 30 and 70
    Straight ticket8.864.106.4813,4352012 to 2024Cut-offs 30 and 70
    Reported turnout9.074.006.5418,9762012 to 2024Cut-offs 30 and 70
    Campaign activity9.482.936.2018,6972012 to 2024Cut-offs 30 and 70
    Party-line presidential vote1.92−0.2440.83718,702Change since 1978 to 1990Per 10 rating points
    Party-line House vote2.530.5161.5219,842Change since 1978 to 1990Per 10 rating points
    Straight ticket1.96−0.8450.55716,097Change since 1978 to 1990Per 10 rating points
    Reported turnout3.561.732.6429,668Change since 1978 to 1990Per 10 rating points
    Campaign activity2.540.5141.5329,303Change since 1978 to 1990Per 10 rating points
    Party-line presidential vote2.57−1.690.44118,702Change since 1978 to 1990Per standard deviation
    Party-line House vote4.440.4762.4619,842Change since 1978 to 1990Per standard deviation
    Straight ticket3.27−2.250.51116,097Change since 1978 to 1990Per standard deviation
    Reported turnout6.803.164.9829,668Change since 1978 to 1990Per standard deviation
    Campaign activity4.840.7442.7929,303Change since 1978 to 1990Per standard deviation
    Party-line presidential vote9.712.456.0818,702Change since 1978 to 1990Cut-offs 15 and 85
    Party-line House vote6.84−1.352.7519,842Change since 1978 to 1990Cut-offs 15 and 85
    Straight ticket5.66−6.21−0.27416,097Change since 1978 to 1990Cut-offs 15 and 85
    Reported turnout14.335.9210.1329,668Change since 1978 to 1990Cut-offs 15 and 85
    Campaign activity12.702.867.7829,303Change since 1978 to 1990Cut-offs 15 and 85
    Party-line presidential vote12.813.968.3818,702Change since 1978 to 1990Cut-offs 30 and 70
    Party-line House vote8.550.0344.2919,842Change since 1978 to 1990Cut-offs 30 and 70
    Straight ticket3.52−8.45−2.4716,097Change since 1978 to 1990Cut-offs 30 and 70
    Reported turnout13.105.189.1429,668Change since 1978 to 1990Cut-offs 30 and 70
    Campaign activity6.67−2.432.1229,303Change since 1978 to 1990Cut-offs 30 and 70
    Each panel measures the two ratings differently and each dot is the difference between the coldness and warmth coefficients with a 95% interval. The per-point panel is the one in the previous figures, the second panel rescales both ratings to a standard deviation within the year, and the last two panels replace the ratings with indicators for being very cold toward the other party and very warm toward the own party at two pairs of cut-offs. In the 1978 to 1990 view the four measures disagree: the interval includes zero in 12 of the 20 cells, both cut-off panels put the House and straight-ticket differences above zero, and the looser pair does the same for campaign activity. In the change view the measures disagree again: 12 of the 20 change intervals lie above zero, presidential voting is above zero under both cut-off pairs and includes zero per point and per standard deviation, and only the turnout row is above zero under all four. Panels have their own horizontal scales.Note: Intervals are 95% intervals from HC3 robust standard errors (a linear probability model with year fixed effects) and treat the weighted sample as a simple random sample.Note: Panels are on different scales and cannot be compared in size. The cut-off panels are indicator coefficients in one model per behavior and period; each indicator is adjusted for the other. The change view is the 2012 to 2024 difference minus the 1978 to 1990 difference, with the two standard errors combined in quadrature; its respondent count is the two periods together. The 1978 to 1990 and 1992 to 2008 samples include midterm studies for House voting, reported turnout and campaign activity, and 2012 to 2024 has none.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z)

    Figure 8 adds five checks for 2012 to 2024. Counting leaners as partisans is a choice, and dropping them leaves the differences positive and narrower: 0.3 (95% CI −0.3 to 0.9)0.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, leaners excluded, per 10 rating points (95% interval, HC3)Interval95% CI −0.3 to 0.9n11,862Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File for presidential voting and 0.7 (95% CI −0.1 to 1.4)0.7 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, leaners excluded, per 10 rating points (95% interval, HC3)Interval95% CI −0.1 to 1.4n10,454Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File for House voting, both with intervals that include zero, and 1.5 (95% CI 0.8 to 2.2)1.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, leaners excluded, per 10 rating points (95% interval, HC3)Interval95% CI 0.8 to 2.2n13,947Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024SourceANES Cumulative Data File for reported turnout, 0.8 (95% CI 0.1 to 1.4)0.8 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on straight-ticket vote, coefficient in a weighted linear probability model with year fixed effects, leaners excluded, per 10 rating points (95% interval, HC3)Interval95% CI 0.1 to 1.4n10,207Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party candidate for both president and the HousePeriod2012 to 2024SourceANES Cumulative Data File for the straight ticket and 1.2 (95% CI 0.2 to 2.2)1.2 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on any campaign activity, coefficient in a weighted linear probability model with year fixed effects, leaners excluded, per 10 rating points (95% interval, HC3)Interval95% CI 0.2 to 2.2n13,739Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered all five campaign activity questionsPeriod2012 to 2024SourceANES Cumulative Data File for campaign activity, above zero. For the two vote items the edge for coldness depends partly on counting leaners. Holding party, strength of identification, education, age and interest in the campaign fixed leaves the presidential difference at 0.6 (95% CI 0.1 to 1.1)0.6 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, party, strength of identification, education, age and interest in the campaign held fixed, per 10 rating points (95% interval, HC3)Interval95% CI 0.1 to 1.1n14,882Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File and the House difference at 1.2 (95% CI 0.6 to 1.9)1.2 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, party, strength of identification, education, age and interest in the campaign held fixed, per 10 rating points (95% interval, HC3)Interval95% CI 0.6 to 1.9n13,254Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File. Fitting each rating in its own model gives 0.5 (95% CI 0.1 to 0.9)0.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, each feeling in its own model, per 10 rating points (95% interval, HC3)Interval95% CI 0.1 to 0.9n15,612Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File and 1.2 (95% CI 0.7 to 1.6)1.2 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on party-line House vote, coefficient in a weighted linear probability model with year fixed effects, each feeling in its own model, per 10 rating points (95% interval, HC3)Interval95% CI 0.7 to 1.6n13,912Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File. On the log-odds scale every difference is above zero. On the like-dislike scale for each party, asked in 77 surveysEstimandcount of survey yearsWeightsnone (count)PopulationANES time-series studies with the like-dislike scale for both partiesPeriod1996 to 2024SourceANES Cumulative Data File studies from 1996, the presidential difference is 0.3 (95% CI −0.2 to 0.8)0.3 percentage points of the outcome per scale point of the like-dislike scaleEstimandcoldness coefficient minus warmth coefficient on party-line presidential vote, coefficient in a weighted linear probability model with year fixed effects, the 0 to 10 like-dislike scale for each party, per scale point (95% interval, HC3)Interval95% CI −0.2 to 0.8n15,238Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party presidential candidatePeriod2012 to 2024SourceANES Cumulative Data File and the turnout difference 0.3 (95% CI −0.4 to 0.9)0.3 percentage points of the outcome per scale point of the like-dislike scaleEstimandcoldness coefficient minus warmth coefficient on turnout, coefficient in a weighted linear probability model with year fixed effects, the 0 to 10 like-dislike scale for each party, per scale point (95% interval, HC3)Interval95% CI −0.4 to 0.9n18,479Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the turnout questionPeriod2012 to 2024SourceANES Cumulative Data File, both with intervals that include zero, while House voting, the straight ticket and campaign activity stay above zero. Across the 2525 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File cells of Figure 8 every difference is positive and 2121 countEstimandcount of estimates with the stated property (see the figure caption)Weightsnone (count)Populationestimates shown in the chapter's figuresPeriod1978 to 2024SourceANES Cumulative Data File have an interval above zero.

    Figure 8

    In 2012 to 2024 the difference between coldness and warmth is positive in all 25 cells of five further checks, with an interval above zero in 21

    Coldness coefficient minus warmth coefficient, partisans, by check, 2012 to 2024; percentage points per 10 rating points (log-odds in one panel, per scale point of the like-dislike scale in another)

    Leaners excluded

    Party, strength, education, age and interest held fixed

    Each rating in its own model

    Log-odds scale

    Like-dislike scale

    Coldness minus warmth

    Show data
    Data for In 2012 to 2024 the difference between coldness and warmth is positive in all 25 cells of five further checks, with an interval above zero in 21
    behaviordiff_conf_highdiff_conf_lowdiff_estimatenperiodspec_label
    Party-line presidential vote0.457−1.78−0.6622,3451978 to 1990Leaners excluded
    Party-line House vote0.477−1.41−0.4684,5181978 to 1990Leaners excluded
    Straight ticket0.959−2.09−0.5642,0151978 to 1990Leaners excluded
    Reported turnout0.094−1.55−0.7277,8631978 to 1990Leaners excluded
    Campaign activity0.480−1.16−0.3397,8011978 to 1990Leaners excluded
    Party-line presidential vote0.446−1.56−0.5603,0211978 to 1990Party, strength, education, age and interest held fixed
    Party-line House vote1.760.0040.8805,8301978 to 1990Party, strength, education, age and interest held fixed
    Straight ticket2.830.1031.472,6021978 to 1990Party, strength, education, age and interest held fixed
    Reported turnout0.844−0.4590.19210,5251978 to 1990Party, strength, education, age and interest held fixed
    Campaign activity0.920−0.3840.26810,4421978 to 1990Party, strength, education, age and interest held fixed
    Party-line presidential vote0.466−1.21−0.3703,0901978 to 1990Each rating in its own model
    Party-line House vote0.659−0.841−0.0915,9301978 to 1990Each rating in its own model
    Straight ticket1.60−0.6010.4982,6621978 to 1990Each rating in its own model
    Reported turnout−0.384−1.69−1.0410,6921978 to 1990Each rating in its own model
    Campaign activity0.623−0.666−0.02110,6061978 to 1990Each rating in its own model
    Party-line presidential vote0.1690.0130.0913,0901978 to 1990Log-odds scale
    Party-line House vote0.062−0.0310.0165,9301978 to 1990Log-odds scale
    Straight ticket0.118−0.0170.0512,6621978 to 1990Log-odds scale
    Reported turnout−0.016−0.079−0.04710,6921978 to 1990Log-odds scale
    Campaign activity0.030−0.0310.00010,6061978 to 1990Log-odds scale
    Party-line presidential vote0.543−1.13−0.2964,0291992 to 2008Leaners excluded
    Party-line House vote1.18−0.8190.1824,6581992 to 2008Leaners excluded
    Straight ticket1.51−0.8830.3153,4091992 to 2008Leaners excluded
    Reported turnout2.160.2901.237,0681992 to 2008Leaners excluded
    Campaign activity2.790.7811.797,0571992 to 2008Leaners excluded
    Party-line presidential vote0.027−1.69−0.8334,6371992 to 2008Party, strength, education, age and interest held fixed
    Party-line House vote1.66−0.2400.7105,5931992 to 2008Party, strength, education, age and interest held fixed
    Straight ticket1.66−0.7260.4683,9091992 to 2008Party, strength, education, age and interest held fixed
    Reported turnout1.980.4321.218,9701992 to 2008Party, strength, education, age and interest held fixed
    Campaign activity1.910.2251.078,9581992 to 2008Party, strength, education, age and interest held fixed
    Party-line presidential vote0.057−1.18−0.5635,4001992 to 2008Each rating in its own model
    Party-line House vote0.719−0.7080.0066,2711992 to 2008Each rating in its own model
    Straight ticket0.956−0.6960.1304,5281992 to 2008Each rating in its own model
    Reported turnout1.560.1780.87010,0101992 to 2008Each rating in its own model
    Campaign activity2.100.6641.389,9971992 to 2008Each rating in its own model
    Party-line presidential vote0.141−0.0290.0565,4001992 to 2008Log-odds scale
    Party-line House vote0.088−0.0210.0346,2711992 to 2008Log-odds scale
    Straight ticket0.120−0.0180.0514,5281992 to 2008Log-odds scale
    Reported turnout0.1010.0180.06010,0101992 to 2008Log-odds scale
    Campaign activity0.1080.0370.0739,9971992 to 2008Log-odds scale
    Party-line presidential vote1.30−0.7590.2703,1271992 to 2008Like-dislike scale
    Party-line House vote3.040.4271.732,7831992 to 2008Like-dislike scale
    Straight ticket3.250.3931.822,6621992 to 2008Like-dislike scale
    Reported turnout−0.857−3.21−2.034,0751992 to 2008Like-dislike scale
    Campaign activity3.080.2771.684,0741992 to 2008Like-dislike scale
    Party-line presidential vote0.896−0.2790.30811,8622012 to 2024Leaners excluded
    Party-line House vote1.40−0.0540.67210,4542012 to 2024Leaners excluded
    Straight ticket1.450.1320.79110,2072012 to 2024Leaners excluded
    Reported turnout2.240.7881.5113,9472012 to 2024Leaners excluded
    Campaign activity2.160.2431.2013,7392012 to 2024Leaners excluded
    Party-line presidential vote1.090.0520.56914,8822012 to 2024Party, strength, education, age and interest held fixed
    Party-line House vote1.860.6111.2413,2542012 to 2024Party, strength, education, age and interest held fixed
    Straight ticket1.970.7341.3512,8042012 to 2024Party, strength, education, age and interest held fixed
    Reported turnout1.960.7751.3718,1052012 to 2024Party, strength, education, age and interest held fixed
    Campaign activity2.000.4881.2518,0072012 to 2024Party, strength, education, age and interest held fixed
    Party-line presidential vote0.8780.1290.50415,6122012 to 2024Each rating in its own model
    Party-line House vote1.600.7321.1713,9122012 to 2024Each rating in its own model
    Straight ticket1.390.5450.96613,4352012 to 2024Each rating in its own model
    Reported turnout1.650.7371.1918,9762012 to 2024Each rating in its own model
    Campaign activity1.730.5731.1518,6972012 to 2024Each rating in its own model
    Party-line presidential vote0.1410.0190.08015,6122012 to 2024Log-odds scale
    Party-line House vote0.1800.0770.12913,9122012 to 2024Log-odds scale
    Straight ticket0.1800.0660.12313,4352012 to 2024Log-odds scale
    Reported turnout0.1470.0640.10618,9762012 to 2024Log-odds scale
    Campaign activity0.0940.0320.06318,6972012 to 2024Log-odds scale
    Party-line presidential vote0.806−0.1640.32115,2382012 to 2024Like-dislike scale
    Party-line House vote1.340.2000.77113,5862012 to 2024Like-dislike scale
    Straight ticket2.100.9441.5213,1262012 to 2024Like-dislike scale
    Reported turnout0.916−0.3600.27818,4792012 to 2024Like-dislike scale
    Campaign activity2.380.8781.6318,4752012 to 2024Like-dislike scale
    The panels leave leaners out, hold party, strength of identification, education, age and interest in the campaign fixed, fit each rating in its own model, move to the log-odds scale, and replace the thermometer with the 0 to 10 like-dislike scale. The intervals include zero for presidential voting and House voting when leaners are left out, and for presidential voting and turnout on the like-dislike scale. The like-dislike scale was asked in seven studies from 1996, so it has no 1978 to 1990 view. Each dot is a difference with a 95% interval.Note: Intervals are 95% intervals from HC3 robust standard errors (a linear probability model with year fixed effects) and treat the weighted sample as a simple random sample. The log-odds panel uses a probability-weight sandwich covariance.Note: The like-dislike scale is the 0 to 10 scale for each party, asked in 1996, 2004, 2008, 2012, 2016, 2020 and 2024.Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z)

    In 2012 to 2024, coldness goes with persuading others and giving money

    The campaign activities differ. Figure 9 splits the activity index into its items. In 2012 to 2024 the coefficient on coldness exceeds the coefficient on warmth for trying to persuade others, by 1.9 (95% CI 1.1 to 2.6)1.9 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on act_persuade, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.1 to 2.6n18,886Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the question on whether they tried to persuade others how to votePeriod2012 to 2024SourceANES Cumulative Data File points, and for giving money, by 1.9 (95% CI 1.4 to 2.4)1.9 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on act_money, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 1.4 to 2.4n18,708Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the question on whether they gave money to a party or candidatePeriod2012 to 2024SourceANES Cumulative Data File. Attending a meeting (0.3 (95% CI −0.1 to 0.7)0.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on act_meeting, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.1 to 0.7n18,886Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the question on whether they attended a political meeting or rallyPeriod2012 to 2024SourceANES Cumulative Data File) and working for a party or candidate (0.3 (95% CI 0.0 to 0.7)0.3 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on act_work, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI 0.0 to 0.7n18,892Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the question on whether they worked for a party or candidatePeriod2012 to 2024SourceANES Cumulative Data File) show smaller differences, with the second interval close to zero. Displaying a button or sticker shows none, at −0.1 (95% CI −0.7 to 0.4)−0.1 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on act_button, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.7 to 0.4n18,889Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the question on whether they displayed a button or stickerPeriod2012 to 2024SourceANES Cumulative Data File, and in 1978 to 1990 warmth led for that item, at −0.5 (95% CI −0.9 to −0.1)−0.5 percentage points of the outcome per 10 rating pointsEstimandcoldness coefficient minus warmth coefficient on act_button, coefficient in a weighted linear probability model with year fixed effects, both feelings per 10 rating points (95% interval, HC3)Interval95% CI −0.9 to −0.1n10,656Weightsweight (VCF0009z)Populationpartisans, leaners included, who answered the question on whether they displayed a button or stickerPeriod1978 to 1990SourceANES Cumulative Data File.

    Figure 9

    In 2012 to 2024 coldness went with more persuading others and giving money than warmth did, and the two went with similar amounts of displaying a button

    Percentage points of each campaign activity per 10 points on the 0 to 100 thermometer, one linear probability model per activity and period with both ratings and year fixed effects, partisans including leaners

    • Warmth toward own party
    • Coldness toward other party
    Show data
    Data for In 2012 to 2024 coldness went with more persuading others and giving money than warmth did, and the two went with similar amounts of displaying a button
    activitycold_conf_highcold_conf_lowcold_estimatediff_conf_highdiff_conf_lowdiff_estimatenperiodwarm_conf_highwarm_conf_lowwarm_estimate
    Tried to persuade others how to vote2.371.551.961.11−0.1440.48510,6481978 to 19901.941.001.47
    Attended a meeting or rally0.8160.2880.5520.183−0.605−0.21110,6561978 to 19901.050.4750.763
    Worked for a party or candidate0.7010.3000.5010.420−0.1590.13010,6461978 to 19900.5880.1530.371
    Displayed a button or sticker0.9650.4200.692−0.116−0.898−0.50710,6561978 to 19901.500.9031.20
    Gave money to a party or candidate1.210.6560.9330.705−0.1160.29510,6341978 to 19900.9430.3340.638
    Tried to persuade others how to vote4.203.223.713.622.072.8410,0061992 to 20081.450.2930.869
    Attended a meeting or rally1.190.6450.9160.594−0.1580.21810,0061992 to 20080.9910.4040.698
    Worked for a party or candidate0.7750.3810.5780.525−0.0090.25810,0041992 to 20080.5230.1180.320
    Displayed a button or sticker1.881.131.510.452−0.675−0.11110,0071992 to 20082.031.201.62
    Gave money to a party or candidate1.951.281.611.450.4700.96210,0041992 to 20080.9990.3050.652
    Tried to persuade others how to vote3.542.623.082.611.131.8718,8862012 to 20241.680.7301.21
    Attended a meeting or rally0.8520.3880.6200.705−0.0760.31418,8862012 to 20240.5550.0570.306
    Worked for a party or candidate0.6620.3360.4990.6500.0350.34318,8922012 to 20240.358−0.0460.156
    Displayed a button or sticker2.231.551.890.430−0.686−0.12818,8892012 to 20242.391.652.02
    Gave money to a party or candidate3.022.402.712.451.421.9418,7082012 to 20241.120.4330.774
    The solid dot (coldness toward the other party) lies well to the right of the ring (warmth toward the own party) for persuading others and giving money in 2012 to 2024, and for displaying a button the two dots overlap. In 1978 to 1990 the intervals for the difference include zero for persuading and giving money, and warmth leads for displaying a button. The five items are VCF0717 to VCF0721. Coldness is 100 minus the rating of the other party.Note: Intervals are 95% intervals from HC3 robust standard errors (a linear probability model with year fixed effects) and treat the weighted sample as a simple random sample.Note: The 1978 to 1990 and 1992 to 2008 samples include midterm studies, and 2012 to 2024 has none. The activity questions are the cumulative-file items VCF0717 (tried to influence others), VCF0718 (attended a meeting or rally), VCF0719 (worked for a party or candidate), VCF0720 (displayed a button or sticker) and VCF0721 (gave money).Source: ANES Cumulative Data File, 1978-2024; weighted (VCF0009z)

    What these data cannot show

    The comparison is between two feelings measured in the same survey as the behavior. A respondent who voted for a party's candidate may rate that party warmer afterward, and a respondent who dislikes the other party may be more willing to say that they voted or took part, so the coefficients cannot say which way the association runs. Interest in politics, strength of party identification, ideology and age may each raise warmth, coldness and behavior together. The checks hold a few of them fixed and do not hold ideology or issue positions fixed. Ratings of the parties may also reflect feelings about party elites more than about other citizens [9], and the party rating is not the rating of the presidential nominee, which goes with the presidential vote and is studied in Chapter 2.1, Feelings toward the other party.

    The behaviors are reports. Validation against public records finds that survey respondents consistently overreport voting and that the nonvoters who say they voted tend to be politically engaged [10]. The file has no validated-vote variable, so the turnout results describe respondents who say they voted, and overreporting that is concentrated among engaged partisans could change the comparison of coldness and warmth. The self-reported vote also leans Democratic: among respondents who reported a major-party presidential vote, 56.0%56.0% percent of respondents who reported a major-party presidential voteEstimandweighted share reporting a Democratic vote (95% Wilson interval)Interval95% CI 54.3% to 57.7%n6,119Weightsweight (VCF0009z)PopulationANES respondents who reported voting for the Democratic or Republican presidential candidatePeriod2020SourceANES Cumulative Data File reported a Democratic vote in 2020 and 55.3%55.3% percent of respondents who reported a major-party presidential voteEstimandweighted share reporting a Democratic vote (95% Wilson interval)Interval95% CI 53.0% to 57.5%n3,948Weightsweight (VCF0009z)PopulationANES respondents who reported voting for the Democratic or Republican presidential candidatePeriod2024SourceANES Cumulative Data File in 2024, against 52.3%52.3% percent of the two-party presidential voteEstimandDemocratic share of the two-party vote, summed over countiesWeightsnone (full counts)Populationvotes cast for the Democratic or Republican candidate in all counties except AlaskaPeriod2020SourceCounty election returns and 49.3%49.3% percent of the two-party presidential voteEstimandDemocratic share of the two-party vote, summed over countiesWeightsnone (full counts)Populationvotes cast for the Democratic or Republican candidate in all counties except AlaskaPeriod2024SourceCounty election returns of the two-party vote in county returns, Alaska excluded. The party-line measures compare each partisan with the own party's candidate, so they are less exposed to that lean than a vote share is. The vote items cover only partisans who reported voting for a major-party candidate, so partisans who stayed home or chose a third-party candidate, among whom coldness toward both parties may be common, are outside them. Party-line voting is near its ceiling in recent years, which limits how much any coefficient can show, and the log-odds check addresses scale and not selection.

    The results come from repeated cross-sections, so they cannot show that any partisan grew colder or that any partisan's behavior changed as feelings changed. A panel that follows the same people is needed for that, and the ANES 2016, 2020 and 2024 panel is not in the files used here. The three periods were fixed before estimation, but their boundaries are a choice, and the per-decade slope and the yearly figure are there so that the result does not rest on them alone. The periods also differ in election type, and the presidential-years comparisons use three early studies. The 2012 to 2024 sample includes 2020, which has no mode weights, so the mode comparison covers 2012, 2016 and 2024 only. The intervals are model based and ignore the ANES sampling design. The chapter reports five behaviors, three periods and several measures without adjusting for multiple comparisons, so an interval that excludes zero in one cell is weaker evidence than it looks. What the ratings go with in views about political violence and democratic norms belongs to Chapter 6.8, Political violence and democratic norms, and who turns out to vote belongs to Chapter 3.7, Turnout.

    Data sources

    • ANES Cumulative Data FileCumulative Data File 1948-2024: party thermometers 1978-2024; presidential vote VCF0704, House vote VCF0707, ticket splitting VCF0709, turnout VCF0702 and campaign activity VCF0717 to VCF0721; the 0 to 10 like-dislike scale for each party in seven studies from 1996Years: 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 2012, 2016 and 2024 · License: Free download after accepting the ANES terms of use; cite the release.
    • County election returnsCounty presidential returns 2020 and 2024, Alaska excluded: the Democratic share of the two-party vote, used only to set beside the ANES self-reported voteYears: President 1868-2024 (40 elections); House and Senate 1990-2024 · Weights: none (full counts) · License: The Heideman file is CC BY 4.0 (doi:10.5281/zenodo.22976553), but the state records behind it keep their own terms, and the sites of Florida, Indiana, Colorado and Wisconsin restrict reuse.

    All data sources

    Methods

    Sample. Respondents to the American National Election Studies (ANES) time series, from the Cumulative Data File, who identify as Democrats or Republicans on the seven-point party identification scale with leaners included (codes 1 to 3 and 5 to 7) and who gave a valid rating of both parties on the 0 to 100 feeling thermometer: 43,063 respondents in 18 surveys from 1978 to 2024 (the surveys with party thermometers; the 2002 study has none). Pure independents are left out because they have no own party. Each survey is a separate cross-section of people, so every result describes how ratings and behavior go together among partisans interviewed in the same year and not how any one person changed. Feelings. Warmth is the rating of the respondent's own party. Coldness is 100 minus the rating of the other party, so that a higher value is colder and a coefficient on coldness can be set beside a coefficient on warmth. Coefficients are reported per 10 rating points, or per standard deviation (each rating standardized within its survey year over the whole paired sample, weighted), or between groups. The classification used for the group comparisons marks a partisan as very cold toward the other party when the other-party rating is 15 or lower and as very warm toward the own party when the own-party rating is 85 or higher (the strict cut-offs, each 35 points from the midpoint of 50), with 30 and 70 as the looser pair (20 points from the midpoint). Four groups follow: neither, warm only, cold only and both. Respondents give 15, 30, 70 and 85 more often than the values next to them, so each cut-off sits on a common answer and includes it. A second measure of feeling, the 0 to 10 like-dislike scale for each party (asked in 1996, 2004, 2008, 2012, 2016, 2020 and 2024), is put in the same direction, and one scale point corresponds to 10 rating points. Behavior. Party-line presidential vote is a vote for the own party's candidate among partisans who reported voting for a major-party presidential candidate (VCF0704); defection is one minus this share. Party-line House vote is the same for the House candidate (VCF0707). Straight ticket is a vote for the same party for president and the House among partisans who reported both, the complement of ticket splitting (VCF0709); it is rebuilt from VCF0704 and VCF0707 in every year because the file carries VCF0709 only through 2020, and the rebuilt variable equals VCF0709 for every respondent who has both. Reported turnout is the report of having voted (VCF0702). Campaign activity is a yes to at least one of five questions, whether the respondent tried to influence how others vote (VCF0717), attended a meeting or rally (VCF0718), worked for a party or candidate (VCF0719), displayed a button or sticker (VCF0720) or gave money (VCF0721), with all five answered. All behaviors are self-reports, and the vote items are asked only of respondents who reported voting. Estimators. For each behavior and period, a weighted linear probability model of the 0/1 behavior on coldness and warmth with survey-year fixed effects and HC3 heteroskedasticity-robust standard errors (MacKinnon and White, 1985); the difference between the two coefficients (coldness minus warmth) and its interval come from the joint covariance matrix. The three periods are 1978 to 1990 (7 surveys), 1992 to 2008 (7 surveys) and 2012 to 2024 (4 surveys), fixed before estimation. The change between two periods is the difference of two independent estimates and combines their standard errors in quadrature. Yearly estimates fit the same model to one survey year without fixed effects. The trend per decade interacts coldness and warmth with decades since 2001 in a model with year fixed effects. The logit check fits a weighted logistic regression and uses a probability-weight sandwich covariance with an n/(n-k) factor. The check with each rating alone stacks the two single-rating models and clusters on the respondent so that their difference has an interval. The controls check adds party, strength of identification (strong, weak, leaner), education (four categories), age and interest in the campaign (three categories; complete cases). Group shares carry Wilson intervals on the Kish effective sample size, and the difference between two groups combines the two normal standard errors. Correlations use Fisher z on the Kish effective sample size, with each rating centered on its survey-year mean. All intervals treat the weighted sample as a simple random sample and ignore the strata and clusters of the ANES design, which the Cumulative File does not carry. Election type. Presidential voting and the straight ticket exist only in presidential election years. House voting, reported turnout and campaign activity also exist in midterm years, which are 56% of the respondents in 1978 to 1990, 25% in 1992 to 2008 and none in 2012 to 2024. The presidential-years-only comparisons keep 1980, 1984 and 1988 from the earliest period (and every presidential year of the middle one), the midterm-only estimates use 1978, 1982, 1986 and 1990, and the trend per decade is also fitted on presidential years alone. Weights. The full-sample weight VCF0009z for all estimates. Mode: weight_ftf (VCF0009x) on the fresh face-to-face sample and weight_web (VCF0009y) on the fresh web sample, which exist for 2008 (face to face only), 2012, 2016 and 2024; the 2020 study has no mode weights, and the 2024 re-interviewed panel and paper respondents belong to neither mode group. Mode estimates pool 2012, 2016 and 2024 and are set against the 1978 to 1990 estimate (all years of that period, and its presidential years alone), which comes from studies interviewed mainly in person. A check with the post-election weight, available in the file for 7 surveys, changes a yearly coefficient or difference by at most 0.5 points. Lean of the self-reported vote. Among all ANES respondents who reported a vote for the Democratic or Republican presidential candidate, the weighted Democratic share is 56.0% in 2020 and 55.3% in 2024; the Democratic share of the two-party vote in county returns, Alaska excluded, is 52.3% and 49.3%. Suppression. Cells of group shares need at least 100 respondents and an effective sample of at least 100. A yearly model needs at least 150 respondents who answered the behavior question and an effective sample of 100 or more. Every pooled period estimate rests on at least 1,394 respondents.

    References

    1. MacKinnon, James G.; White, Halbert (1985). Some Heteroskedasticity-Consistent Covariance Matrix Estimators with Improved Finite Sample Properties. Journal of Econometrics 29(3): 305-325. doi:10.1016/0304-4076(85)90158-7
    2. Wilson, Edwin B. (1927). Probable Inference, the Law of Succession, and Statistical Inference. Journal of the American Statistical Association 22(158): 209-212. doi:10.1080/01621459.1927.10502953
    3. Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
    4. 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. doi:10.1016/j.electstud.2015.11.001
    5. Abramowitz, Alan I.; Webster, Steven W. (2018). Negative Partisanship: Why Americans Dislike Parties But Behave Like Rabid Partisans. Political Psychology 39(S1): 119-135. doi:10.1111/pops.12479
    6. Caruana, Nicholas J.; McGregor, R. Michael; Stephenson, Laura B. (2015). The Power of the Dark Side: Negative Partisanship and Political Behaviour in Canada. Canadian Journal of Political Science 48(4): 771-789. doi:10.1017/S0008423914000882
    7. Medeiros, Mike; Noël, Alain (2014). The Forgotten Side of Partisanship: Negative Party Identification in Four Anglo-American Democracies. Comparative Political Studies 47(7): 1022-1046. doi:10.1177/0010414013488560
    8. Bankert, Alexa (2021). Negative and Positive Partisanship in the 2016 U.S. Presidential Elections. Political Behavior 43(4): 1467-1485. doi:10.1007/s11109-020-09599-1
    9. Druckman, James N.; Levendusky, Matthew S. (2019). What Do We Measure When We Measure Affective Polarization?. Public Opinion Quarterly 83(1): 114-122. doi:10.1093/poq/nfz003
    10. Ansolabehere, Stephen; Hersh, Eitan (2012). Validation: What Big Data Reveal About Survey Misreporting and the Real Electorate. Political Analysis 20(4): 437-459. doi:10.1093/pan/mps023

    How to cite

    @misc{singh2026negativepartisanship,
  author = {Singh, Vishal},
  title = {Negative partisanship},
  year = {2026},
  url = {https://vishalsingh.org/politics/negative-partisanship},
  note = {In The Public Mind. Published 3 October 2026.}
}

    Reading paths