Polarization: when disagreement turns into distrust of the other side

Political polarization in the United States is two related problems: party officeholders have moved far apart on policy, and ordinary Republicans and Democrats increasingly dislike, distrust, and misjudge each other. Survey evidence shows the second trend, called affective polarization, rising faster in the United States than in most comparable democracies. Researchers still disagree about how much the public’s actual policy views have split, and about what causes the hostility.

In plain words: people have always disagreed about politics. What has changed is how people feel about those who disagree with them. Many now see the other party as immoral or dishonest, not just wrong. People also tend to think the other side is more extreme than it really is. That makes compromise harder, even on issues where many people already agree.

The problem of polarization, defined precisely

Polarization is a word that hides several different measurements, and the debate about it is often a debate about which one is meant.

Ideological polarization. The distance between groups on policy positions, and how consistently individuals hold left or right positions across many issues.

Affective polarization. Dislike and distrust of the opposing party and its supporters, regardless of policy views. Iyengar, Sood, and Lelkes proposed treating polarization as social distance and reported that “both Republicans and Democrats increasingly dislike, even loathe, their opponents”.⁠³

Partisan sorting. The process by which people with liberal views end up in one party and people with conservative views in the other. Sorting can make parties look more different without any individual becoming more extreme.

Elite polarization. The distance between the parties’ officeholders, usually measured from roll call votes in Congress.

Perceived polarization and metaperceptions. What people believe the other side thinks. A metaperception is a belief about another group’s beliefs. It can be wrong in measurable ways.

These measures can move independently. A country can have sharply divided legislators and a moderately divided public, or a public that agrees on many policies while its partisans despise each other. Clear thinking about remedies depends on naming the measure first. For how survey measures and denominators limit what any such number can claim, see Measurement.

Scale and evidence of polarization

Polarization in the American public, measured by Pew Research Center telephone surveys, rose sharply between 1994 and 2014. In a 2014 survey of 10,013 adults, the share of Americans expressing consistently conservative or consistently liberal opinions had doubled from 10% to 21% over two decades. In 2014, 92% of Republicans were to the right of the median Democrat and 94% of Democrats were to the left of the median Republican, compared with 64% and 70% in 1994. Very unfavorable views of the other party rose from 17% of Republicans and 16% of Democrats in 1994 to 43% and 38% in 2014.⁠¹

Hostility continued to grow. In a Pew survey of 6,174 adults conducted June 27 to July 4, 2022, 62% of Republicans and 54% of Democrats held a very unfavorable view of the other party. The share of Republicans who described Democrats as more immoral than other Americans rose from 47% in 2016 to 72% in 2022; the share of Democrats who said the same about Republicans rose from 35% to 63%. Pew also found that 27% of Americans viewed both parties unfavorably, up from 6% in 1994.⁠²

The American trend is unusual internationally. Boxell, Gentzkow, and Shapiro measured affective polarization in twelve OECD countries over four decades. In their baseline estimates, the United States had the largest increase; five countries had smaller increases and six had decreases.⁠⁴

Elite polarization is measured differently. Pew’s analysis of DW-NOMINATE scores, which place each member of Congress on a scale from 1 (most conservative) to -1 (most liberal) based on roll call votes, found that the parties were farther apart ideologically than at any time in the past 50 years. About two dozen moderate members remained, compared with more than 160 in 1971 and 1972. Fifty years earlier, 144 House Republicans were less conservative than the most conservative Democrat; the House has had no ideological overlap between the parties since 2002. The movement was uneven: the average House Democrat moved from about -0.31 to -0.38, while the average House Republican moved from 0.25 to nearly 0.51.⁠⁵

Against this, Fiorina and Abrams reviewed the evidence on the mass public in 2008 and concluded that direct evidence on citizens’ policy positions showed “little or no indication of increased mass polarization over the past two to three decades,” while party sorting had clearly occurred.⁠⁶ The Pew consistency data and the Fiorina and Abrams review are not necessarily in conflict: consistency across issues can rise through sorting even if the average person’s position on each issue moves little.

Perceptions are further apart than reality. More in Common’s 2019 Perception Gap study, a YouGov survey of 2,100 Americans, found that Democrats and Republicans imagine almost twice as many of their opponents hold views they consider extreme as actually do. People who said they read the news most of the time were nearly three times more distorted in their perceptions than those who read it only now and then, though the authors note they cannot prove that one causes the other.⁠⁷ Misperception extends to violence: across studies with a combined 4,741 participants, partisans’ estimates of rival partisans’ support for political violence were 245% to 442% higher than actual levels.⁠⁸

Party labels themselves have weakened even as partisan dislike has grown. Gallup found that a record 45% of U.S. adults identified as political independents in 2025, with 27% each identifying as Democrats and Republicans; most independents still lean toward one party (20% Democratic leaners, 15% Republican leaners, 10% non-leaners).⁠¹⁴

Causes and competing explanations

Explanations for polarization fall into several families, and the evidence for each is partial.

Campaigns and elite cues. Iyengar, Sood, and Lelkes argue that partisan affect is only inconsistently tied to policy attitudes and that “exposure to messages attacking the out-group reinforces partisans’ biased views of their opponents”.⁠³ On this view, hostility is learned from political competition itself.

Partisan media. Martin and Yurukoglu estimated a model in which viewers choose slanted cable news and their ideologies shift as a result. They estimate that Fox News increases Republican vote shares by 0.3 points among viewers induced to watch 2.5 additional minutes per week, and use the model to assess media-driven polarization.⁠¹² The estimate concerns one network and one period and depends on the model’s assumptions.

Social media. The experimental record is mixed. Allcott, Braghieri, Eichmeyer, and Gentzkow randomly paid users to deactivate Facebook for the four weeks before the 2018 midterm election and found that deactivation “reduced both factual news knowledge and political polarization”.⁠¹¹ A large 2023 study with Meta moved consenting users from algorithmic to reverse-chronological feeds for three months. Time on the platforms fell, and exposure to political and untrustworthy content rose, but the change “did not significantly alter levels of issue polarization” or other key attitudes.⁠¹⁰ One reading is that platforms matter less than critics claim; another is that three months of one feed change is too short and too narrow to detect cumulative effects.

Sorting and identity. Fiorina and Abrams emphasize sorting rather than polarization of views.⁠⁶ When party, ideology, religion, and place line up together, party becomes a broader social identity, and disagreement feels personal.

Misperception. The Perception Gap and the violence studies suggest that a large part of the hostility rests on false beliefs about the other side.⁠^7,8

Cross-national variation. Because the rise in affective polarization differs so much across wealthy democracies,⁠⁴ explanations that apply equally to all of them, such as the spread of the internet, face a challenge. That is an inference from the cross-country pattern, not a finding stated in the paper’s abstract.

Institutional explanations, including primary elections, district lines, and the decline of local news, are treated on their own pages: Primary elections, Gerrymandering, Local news collapse, and Misinformation.

Who is harmed by polarization

The costs of polarization fall unevenly, and some are matters of judgment rather than measurement.

• People who hold mixed views. When parties sort tightly, voters whose views cross party lines have fewer candidates who match them. The decline in moderate members of Congress, from more than 160 in 1971 and 1972 to about two dozen,⁠⁵ means fewer representatives occupy that middle ground.
• People targeted by hostility. Misjudging the other side’s support for violence is not harmless. Mernyk and colleagues found that correcting the misperception reduced support for violence, which implies that the misperception was contributing to it.⁠⁸
• The policy process. Hostility raises the cost of compromise for representatives, because cooperation can be punished by their own supporters. See Gridlock.
• Trust in shared institutions. When elections, courts, and agencies are seen through a partisan lens, confidence depends on which side won. See Trust in elections.

Whether polarization also harms by clarifying choices less than it hardens them is a contested question, discussed below.

What has been tried to reduce polarization

Attempts to reduce polarization have targeted beliefs, contact, platforms, and rules. Results differ by target.

Correcting misperceptions. In preregistered studies, a brief correction of inflated beliefs about rival partisans reduced support for partisan violence by 34% and willingness to engage in violence by 44%, and the effect on support persisted about a month later.⁠⁸

Testing many interventions at once. A megastudy of 32,059 participants tested 25 treatments designed by academics and practitioners. Many reduced partisan animosity, “most strongly by highlighting relatable sympathetic individuals” with different views, or by emphasizing shared identities. Several reduced support for undemocratic practices, most strongly by “correcting misperceptions of rival partisans’ views” or highlighting the threat of democratic collapse.⁠⁹ The abstract does not report how long effects lasted, and reducing animosity and reducing antidemocratic attitudes were not achieved by the same treatments.

Structured deliberation. The “America in One Room” deliberative poll brought more than 500 registered voters together for a long weekend on five national issues, with a control group answering the same questions. Deliberators “showed large, depolarizing changes in their policy attitudes and large decreases in affective polarization”.⁠¹³ The event was small, expensive, and carefully moderated, and whether such effects can be reproduced at scale is unresolved. See Comparative cases for other deliberative processes and their limits.

Platform changes. Switching feeds to chronological order did not significantly change polarization in three months.⁠¹⁰ Leaving Facebook reduced polarization but also reduced news knowledge,⁠¹¹ a tradeoff that complicates any simple prescription.

Rule changes. Reforms to primaries, districting, and voting methods are often proposed as remedies for elite polarization. Their records are discussed on the Primary elections and Gerrymandering pages rather than repeated here.

The pattern across these attempts is that individual attitudes can be moved in controlled settings, while evidence that any intervention changes national trends is absent.

The strongest arguments about polarization across the political spectrum

Polarization is described by every side as a problem, but the diagnosis differs. Each view below is stated in its strongest form.

Polarization as clarity. One argument holds that sharp party differences give voters real choices and make parties accountable for programs. On this view, the problem is not distance between parties but contempt between citizens, and remedies should target hostility without blurring policy differences.

Asymmetric polarization. Some scholars and commentators on the left argue that the parties did not move apart equally, citing roll call data showing that House Republicans moved further from the center than House Democrats.⁠⁵ They argue that treating both sides as equally responsible misdiagnoses the problem. Critics reply that roll call scores measure votes on the agenda each Congress actually faced, not fixed positions, and that cultural and institutional shifts on the other side do not appear in those scores.

Polarization as a reaction. Many on the right argue that hostility toward the left grew in response to perceived one-sided control of major cultural institutions, including universities, large media organizations, and parts of the federal government. On this view, distrust is a rational response to institutions people see as unrepresentative, and calls for civility that do not address that perception will fail.

Polarization as a product of centralization. A libertarian argument holds that national politics became more hostile because more decisions of personal importance are made nationally. When the winner decides more of everyone’s life, losing becomes more threatening. The remedy, on this view, is to lower the stakes by decentralizing decisions, not to change citizens’ feelings.

Polarization as an institutional failure. Institutionalists argue that rules such as closed primaries, safe districts, and procedures that reward obstruction amplify a modest public divide into a large elite one. On this view, Fiorina and Abrams’s finding that the public’s views moved less than elites’ is the key evidence.⁠⁶

Polarization as misperception. Researchers focused on metaperceptions argue that much hostility is aimed at an imagined opponent, and that accurate information about what the other side actually believes is the cheapest remedy.⁠^7,8

These positions are not exclusive. A person can believe that roll call data show asymmetry, that cultural institutions are unrepresentative, and that misperception inflates both.

Unresolved questions about polarization

• Causal direction. Does partisan media create hostility, or do hostile partisans seek partisan media? Observational correlations such as the news exposure gap⁠⁷ cannot settle this.
• Durability. Most interventions are measured over days or weeks. Whether any effect lasts through an election cycle is largely unknown.
• Measurement. Feeling thermometers and trait ratings may capture expressive answers rather than real willingness to cooperate. How much of measured hostility is sincere?
• Asymmetry. How should researchers weigh roll call evidence of asymmetric elite movement against cultural and institutional evidence that the roll call record does not capture?
• Desirability. If some disagreement reflects real conflicts of value or interest, how much depolarization is the right goal, and who decides?
• Application to this framework. Would published group priorities reduce misperception, or would partisans read them as scores? This has not been tested. See Evaluation and Questions.