Negative partisanship

Among American National Election Studies (ANES) partisans who voted for a major-party House candidate in 2012 to 2024, 95.0%95.0% percent of the groupEstimandweighted share, party-line House vote, group: cold only (95% Wilson interval)Interval95% CI 94.0% to 95.8%n4,397Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File of those who were very cold toward the other party, without being very warm toward their own, voted for their own party's candidate. Among those who were very warm toward their own party, without being very cold toward the other, 89.9%89.9% percent of the groupEstimandweighted share, party-line House vote, group: warm only (95% Wilson interval)Interval95% CI 87.5% to 91.9%n1,376Weightsweight (VCF0009z)Populationpartisans, leaners included, who reported voting for a major-party House candidatePeriod2012 to 2024SourceANES Cumulative Data File did. The chapter calls these groups cold only and warm only, both defined by ratings on the ANES 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., with 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. counted as partisans. In the same years cold-only partisans also reported voting more often. Being very cold toward the other party went with the party line and with turning out more than being very warm toward one's own party did.

Cold-only partisans voted the party line more often

A partisan counts as very warm toward their own party at a rating of 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 or above out of 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, and as very cold toward the other party at 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 or below. Figure 1 adds the partisans who meet neither condition and those who meet both. It gives the share of each of the four groups that voted for its own party's House candidate in three periods from 1978 to 2024. The cold-only share sits above the warm-only share in all three, by 5.15.1 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 pointsPercentage pointThe unit for the difference between two percentages. A change from one share to another is stated in points, not in percent, so that absolute and relative change are not confused. in 2012 to 2024 and by more in 1978 to 1990, when the share voting the party line was lower. The same order holds in every period for the presidential vote and for the straight ticket, a vote for one party for president and for the House. Reported turnout differs: in 1978 to 1990 warm-only partisans reported voting more often than cold-only partisans, and in 2012 to 2024 the order was reversed.

Point for point, coldness goes with more party-line voting

A group comparison discards the variation inside each group, so the chapter's principal result uses the whole rating. Warmth is the rating of the own party. Coldness is the rating of the other party measured down from the top of the scale, so that a higher value is colder. For each behavior, a linear probability modelLinear probability modelA regression of a yes-or-no outcome on predictors, fitted by least squares so that each coefficient is a change in the probability of yes. The book weights it and reports robust standard errors. estimates the points of the behavior that go with each 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 rating points of coldness while warmth is held fixed. It does the same for a step of warmth while coldness is held fixed. Among partisans who voted for a major-party House candidate in 2012 to 2024, a step colder toward the other party went with 3.53.5 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 points more own-party House voting. A step warmer toward the own party went with 1.91.9 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. That comparison of recent levels is the chapter's principal result. The same order holds in 2012 to 2024 for reported turnout, campaign activity, the straight ticket and the presidential vote, where the difference is smallest. In 2012, 2016 and 2024 pooled, the years with mode weights, the presidential interval includes zero.

How the ANES measures feeling and behavior

The ratings and the behavior come from the ANES Cumulative Data File: about 43,00043,063 respondentsEstimandcount of respondents in the sampleWeightsnone (count)Populationpartisans, leaners included, with a rating of both partiesPeriod18 surveys, 1978 to 2024SourceANES Cumulative Data File partisans who rated both parties, in 1818 surveysEstimandcount of survey yearsWeightsnone (count)PopulationANES time-series studies with both party thermometersPeriod1978 to 2024SourceANES Cumulative Data File ANES studies from 1978 to 2024, every one that asked both party thermometers. Each study interviews a separate sample of Americans, weightedSurvey 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. to match the population on known traits. Every behavior is a self-report of voting, of the vote itself, or of campaign activity, a yes to at least one of five activities from trying to persuade others to giving money. The qualification that changes the meaning is that feelings and behavior are measured in the same interview, so the coefficients describe how the two go together and cannot say which way the association runs. The report gives every estimate with its interval and sample size.

Whether the edge is new is a second question

Figure 2 fits the same model in each of the three periods, and whether the edge for coldness is new depends on the comparison. In 1978 to 1990 the per-point difference, coldness minus warmth, 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 points for House voting, an interval that includes zero. It included zero for campaign activity, the straight ticket and the presidential vote as well. Reported turnout was 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: warmth toward the own party went with more reported turnout than coldness toward the other party did. That result comes from the midterm studies; in presidential years alone the interval includes zero. In 1992 to 2008 the differences for reported turnout and campaign activity were above zero while those for House and presidential voting still included zero. The edge for coldness reached participation first and party-line voting later.

Set 2012 to 2024 against 1978 to 1990 and the difference grew for House voting, reported turnout and campaign activity, with intervals above zero. The changes for the presidential vote and the straight ticket include zero, and the presidential interval is wide enough to include a change larger than the House point estimate. Party-line voting also rose over these years, which compresses coefficients on the probability scale; on the log-odds scale the House change stays above zero. Two narrower comparisons follow. The early period has midterm studies and the recent one has none, so one sets 2012 to 2024 against the presidential years 1980, 1984 and 1988 alone. The early studies were interviewed mainly in person, so the other estimates the face-to-face and web respondents of 2012, 2016 and 2024 separately. Under either, the changes for House voting and campaign activity have intervals that include zero or reach it. Only reported turnout changed by an amount above zero under every comparison, at 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. The group comparison of Figure 1 and the per-point comparison also disagree about 1978 to 1990. The cold-only group voted the party line more often than the warm-only group in those years, while per rating point the two feelings went with similar amounts of House voting.

More partisans are now cold only than warm only

The two groups have also changed size. In 1978 to 1990, 32%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% 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 11%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%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 cold only was the larger group within each interview mode as well. The trend in the ratings themselves belongs to Chapter 2.2, Feelings toward the other party.

These are comparisons of how feelings and behavior go together. Abramowitz and Webster read ANES ratings of the parties as evidence that negative feelings toward the opposing party have led to more party loyalty and more straight-ticket voting [1] [2]. These data show the association and not its direction. Either may shape the other, and both may follow from traits the survey measures only in part, such as interest in politics or strength of party identification. Counting leaners as partisans is a choice. Without them the differences stay positive and are narrower: the House and presidential intervals include zero, while those for reported turnout, the straight ticket and campaign activity stay above zero. For the two vote items the edge for coldness depends partly on counting leaners.

What these surveys cannot tell you

The results come from repeated cross-sections, so they cannot show that any partisan grew colder or changed behavior as feelings changed; a panel that follows the same people is needed for that. The behaviors are reports, and validation against public records finds that respondents overreport voting and that the nonvoters who say they voted tend to be politically engaged [3]. The turnout results therefore describe people who say they voted. The vote items cover only partisans who reported voting for a major-party candidate, so those who stayed home or chose a third-party candidate, among whom coldness toward both parties may be common, are outside them. 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.

Go deeper

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

Reading paths that include this chapter:

• Party feeling and its consequences

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

References

1. 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
2. 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
3. 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.}
}