Case Study 1 — Reading the 2010-2025 Polarization Measurements

What this case asks

What do the actual numbers say about American polarization between 2010 and 2025? This case study walks through the four most influential data sources — Pew Research Center's typology surveys, the Voter Study Group's panel surveys, the American National Election Studies (ANES) feeling-thermometer time series, and the DW-NOMINATE legislative behavior data — and shows what each tells us. The point is not to settle the polarization debate but to give you the empirical foundation on which the rest of the conversation rests. By the end you should be able to evaluate political commentary about polarization with at least a working knowledge of what the numbers actually are.

The four data sources

Pew Research Center's typology surveys have been run periodically since 1987, with major updates in 1994, 2004, 2014, 2017, 2021, and 2024. Pew asks voters a battery of policy-position questions and uses the answers to construct an "ideological consistency" measure — how many liberal or conservative answers a respondent gives across the questions.

The Voter Study Group, a project of the Democracy Fund, has run the VOTER Survey panel since 2011, re-interviewing the same respondents over multiple waves to track individual-level change. This is one of the only panel data sets that captures what individual Americans' views look like over time, rather than just snapshots of different cross-sections.

The American National Election Studies is the longest-running political-attitudes survey in the United States, with feeling-thermometer questions on the parties dating back to 1964. ANES is the gold standard for affective-polarization measurement.

DW-NOMINATE, hosted at Voteview.com and maintained by Howard Rosenthal, Keith Poole, and successors, scores every member of Congress on a left-right scale based on every roll-call vote in U.S. history. It is the gold standard for measuring legislative-behavior polarization.

What Pew shows

The 2014 Pew typology report ("Political Polarization in the American Public") was a watershed in the popular discussion of polarization. Pew's central finding: the share of Americans holding consistently liberal or consistently conservative views across multiple policy domains had roughly doubled from 1994 to 2014. In 1994, about 10 percent of Americans held views consistent with one of the two parties' positions across most issues. By 2014, that figure was about 21 percent.

By 2024, the trend had continued. Pew's later work showed that ideological consistency had risen further among the politically engaged, and the median Republican and median Democrat had pulled further apart. Crucially, the rise in consistency was concentrated among the most engaged citizens — those who follow news closely, vote consistently, and identify strongly with a party.

But Pew also showed something easy to miss: the broader public, including the substantial share of Americans who do not strongly identify with either party, remained more moderate than partisans within either party. The polarization Pew documented was real but concentrated. The "average American" was still in the middle. The "average partisan," however, had pulled toward the wing.

What the Voter Study Group shows

The VOTER Survey's panel structure — interviewing the same respondents over time — produced a finding that cross-sectional surveys cannot: most individual Americans do not change their views dramatically over time. Polarization is not primarily caused by individual voters becoming more extreme; it is primarily caused by sorting, with voters with consistent worldviews increasingly aligning with the party that matches.

This is a subtle but important finding. If you ask "are voters becoming more extreme?" the answer is "somewhat, especially among the engaged." If you ask "are voters becoming more sorted into the right party for their views?" the answer is "substantially yes." The two phenomena look similar in cross-sectional snapshots but are causally distinct.

The VOTER Survey also captured the Trump-era realignment in real time. Many voters who had described themselves as Republican in 2011 but held positions on immigration, trade, or social policy that diverged from the party's mainstream in 2011 became more committed Republicans by 2017 — not because they had changed positions, but because the party had moved toward them. The mirror process happened on the Democratic side: voters whose positions on identity, education, and the role of the state had been outside the Democratic mainstream in 2011 found their party moving toward those positions by 2021.

What ANES shows

The ANES feeling-thermometer data is, in many ways, the most striking. The trajectory is clear, the magnitudes are large, and the inflection points map cleanly onto identifiable political events.

In 1980, partisans rated their own party at about 71 (warm) and the opposing party at about 47 (slightly below neutral). The cross-party gap was 24 points.

By 2000, the cross-party gap had grown to about 32 points, with the decline driven mostly by colder feelings toward the opposing party.

By 2012, the gap was about 40 points. The cooling toward opposing parties had accelerated through the Gingrich-Clinton confrontations of the late 1990s and the Bush-Obama transitions of the 2000s.

By 2024, the gap was approximately 53 points — own party still around 71, opposing party at roughly 18. This is not "we disagree with them"; this is "we cannot stand them." Iyengar's research frames this as the most powerful out-group hostility in modern American social life, exceeding hostility toward racial out-groups, religious out-groups, or any other major identity category.

What DW-NOMINATE shows

The DW-NOMINATE data tells a story about Congress, not about voters. The story is unmistakable.

In 1970, the median House Republican and median House Democrat were separated by roughly 0.45 units on the NOMINATE first-dimension scale (which runs from approximately -1 to +1). There was substantial overlap between the parties: liberal Republicans (Jacob Javits of New York, Charles Mathias of Maryland) and conservative Democrats (the Southern bloc) populated the middle of the distribution.

By 1990, the gap was about 0.65 units, and the overlap was shrinking.

By 2010, the gap was about 0.85 units, and the overlap had essentially vanished. The most conservative House Democrat was to the left of the most liberal House Republican.

By 2024, the gap was approximately 0.95 units — the largest measured since Reconstruction. The bipartisan center, in roll-call-vote terms, had ceased to exist. Cross-party voting on major legislation, which had been routine in the 1970s, had become unusual.

The Senate trajectory tracked the House trajectory closely, with slightly less extreme magnitudes and a slight lag.

What the four sources together show

The four data sources converge on a coherent picture, though each captures a different facet:

Congress is highly polarized. The legislative behavior data (DW-NOMINATE) shows the largest divide since Reconstruction. Cross-party coalitions are rare. Procedural conflict is common. Compromise is punished by primary electorates.

Engaged citizens are highly polarized. Pew's ideological-consistency measures and ANES affective-polarization measures both show substantial growth in polarization among the politically engaged.

The broader public is more moderate than the engaged segment. Pew's typology data show that most Americans do not hold consistently liberal or consistently conservative views; they hold mixed positions. The engaged are not representative of the whole.

Sorting drives much of what looks like polarization. The VOTER Survey panel data show that individual Americans change their views less than cross-sectional snapshots imply. Much of what we call polarization is voters being more accurately matched to the party that fits their views.

Affective polarization has grown faster than ideological polarization. Americans dislike the other party more than the substance of their disagreements would seem to justify.

What the data does not settle

The data does not settle the causal questions. What is driving polarization? Pew, ANES, the VOTER Survey, and DW-NOMINATE describe the phenomenon; they do not by themselves identify the causes. Disentangling the contributions of geographic sorting, media fragmentation, the primary system, gerrymandering, money in politics, identity sorting, political entrepreneurship, social media, and the rural-urban divide requires additional analytical work, much of it contested.

The data does not settle the asymmetric-polarization debate. DW-NOMINATE shows that Republican legislators have moved further from the center than Democratic legislators on the NOMINATE first dimension. But NOMINATE captures roll-call behavior; it does not capture all dimensions of policy substance, and the issue space measured changes over decades. Whether the asymmetry is as large as some scholars claim, smaller than some critics suggest, or somewhere in between is a real and ongoing argument.

The data does not settle what to do. Empirical description does not automatically translate into normative prescription. Reasonable people who agree that polarization is real and rising disagree about whether the appropriate response is structural reform, civic education, partisan victory, individual cross-cutting effort, or some combination.

Conclusion

Empirically, American polarization is real and rising. The data is consistent across Pew, the VOTER Survey, ANES, and DW-NOMINATE. The phenomenon has multiple dimensions (ideological, affective, sorting), is concentrated among the politically engaged, and has shown its largest growth since approximately 1990.

The causes are multiple and contested. The asymmetric-polarization claim has empirical support, particularly in legislative-behavior data, but is contested in important ways. The most honest framing: both major parties have polarized over the past four decades; the magnitudes and directions are different on different measures; serious scholars disagree about how to weight different measures.

A reader of political commentary who takes only the empirical findings of one of these data sources — or who accepts only the version of the asymmetric-polarization debate that flatters their tribe — is consuming an incomplete picture. The full picture is messier, less satisfying tribally, and more useful analytically.

For a citizen attempting to navigate the polarization conversation responsibly, the most important practical takeaway from these data sources is that you can check them yourself. Voteview.com is free; ANES data is downloadable; Pew reports are public PDFs; the VOTER Survey publishes summaries openly. A reader who is told that "the data show" something, on either side of the partisan divide, can verify the underlying claim in fifteen minutes of work. That capacity to verify is the deepest defense against being misled by partisan framing.