Americans have come to debate when and why the United States became the way it is.politically Divided. Was polarization a slow burn going back decades, or was it a sudden collapse related to more recent events?
A new large-scale empirical study Polarization in the United States over political issues This occurred after 2008, as a result of the rapid and sustained expansion of ideological cleavages from the late 1980s through the late 2000s and 2010s.
In a study published in Royal Society Open ScienceIn , researchers introduce a machine learning-based approach to measuring polarization on political issues and apply it to decades of U.S. survey data and comparative data from countries around the world.
Rather than relying on party labels or self-identified ideology, this study unsupervised learning techniques To detect how citizens’ policy views naturally cluster and how far apart those clusters have moved over time.
“Our research shows that 2008 marked a major turning point in the left-right divide on many issues that define modern American politics,” co-author Dr Lee DeWitt, a professor at the University of Cambridge, said in the paper. press release. “Sentiments vary by topic, but Americans have moved to the left on many issues throughout the 21st century. This shift may come as a surprise to those familiar with the rightward shift of Republican leaders over the same period and recent U.S. headlines.”
In this study, researchers claim that: polarization It’s not just about people with strong opinions, it’s about societies fragmenting into internally coherent, distinct groups that at the same time increasingly clash over a wide range of issues. by allowing machine learning algorithms By identifying these groups directly from survey responses, researchers say they can measure polarization more objectively and with much greater precision than previously possible.
“There can be many reasons why voters feel they belong to a particular party, even if many of their attitudes do not align. Clustering algorithms provide an objective, bottom-up way to analyze changes in political opinion, going beyond party ideology or voters’ self-perceptions,” explains co-author Dr. David Young, professor of psychology at the University of Cambridge.
A new way to measure polarized light
Much of the debate about polarization hinges on how it is measured. Traditional approaches are often compared democrats and republicansor distinguish between liberals and conservatives and track how far apart those groups are on certain issues.
However, these methods have drawbacks. Political party coalitions change over time, and people “categorise” themselves into the following political parties: more in line with their opinionsand many citizens do not identify strongly with any political label at all.
To address these issues, researchers A machine learning algorithm called K-means clustering. Rather than asking which political party someone belongs to, the algorithm looks at how individuals answer broad policy questions and groups them into two clusters that are internally similar and externally as different as possible.
For the U.S. analysis, researchers applied the method to U.S. National Election Study data from 1988 to 2024, covering a wide range of issues from abortion and traditional family values to government spending, economic inequality, and race-related policies.
of algorithm We consistently identified two broad opinion clusters that roughly correspond to left-leaning and right-leaning worldviews, but importantly, we did so without using party identification as an input.
Once these clusters were identified, the team measured polarization along three dimensions: the average distance between the clusters, the internal cohesion of each cluster, and how uniform the clusters were in size. Among them, one thing stood out.
Waves of polarization since 2008
Between 1988 and 2008, the distance between America’s two major opinion groups did little to shrink. But since 2008, things have changed dramatically. The spacing between clusters increased sharply over 2020, but has leveled off slightly in the latest data.
“Between 1988 and 2024, separation distance increased by 0.14 scale points, from 0.22 to 0.36,” the authors report, noting that this represents an overall increase of 64 percent, almost all of which occurred after 2008.
Importantly, the other two dimensions of polarization remain approximately stable. The clusters never became more homogeneous internally, nor did one side come to numerically dominate the population. Instead, Americans have simply fallen into two separate camps on more and more issues.
This pattern calls into question the narrative that frames polarization as a steady rise over decades. Rather, the data shows a long period of stagnation followed by rapid escalation. This finding is consistent with broader social and technological changes in the late 2000s. social media such as economic shocks and political conflicts leading to nationalization.
Not just a party, but a packaged belief
One of the insights emerging from this study is how polarization manifests itself across individual issues. Over time, the researchers found that the two clusters diverged for every issue they examined, not just a few high-profile flashpoints.
In some cases, both clusters move in opposite directions, which is a typical polarization situation. In other groups, both groups moved in the same direction, but at different speeds, still increasing the distance between the groups.
The result is what researchers describe as a “packaged” ideology. That is, people on one side were more likely to hold consistently consistent positions across social, economic, and cultural domains.
“What this effectively means is that in 1988, people with generally right-leaning views did not necessarily have more restrictive views on abortion than those with generally left-leaning views,” the researchers explain. In contrast, “right-leaning views now tend to be ‘packaged’ with restrictive views on abortion.” This shows that polarization is increasingly consolidating once-separate issues into coherent ideological camps.
This cohesive effect helps explain why political disagreements today feel so totalizing. It’s not just that Americans disagree a lot, they disagree with the same people on almost everything.
Machine learning meets political science
Beyond its conclusions about U.S. politics, this study computational psychology We aim to demonstrate how unsupervised algorithms can reveal the latent structure of public opinion that traditional analysis misses.
Clustering allows you to measure polarization without making any prior assumptions about who the “side” is. This is an important advantage in cross-country comparisons.
The paper’s second major analysis applies the same approach to survey data from more than 100 countries and finds that cultural issues, rather than economic issues, tend to drive polarization globally, but in very different ways depending on a country’s level of development.
In the case of the United States, this technique also helps disentangle two often confused phenomena: ideological classification and true attitude differences. Researchers found evidence that both happened. Americans became better organized into political parties that reflected their views, but their views themselves became further apart.
“Developing countries typically have a culturally conservative majority and a culturally liberal minority. Highly developed countries tend to have a socially liberal majority, even if only by a small margin,” Dr. Young said. “The United States is unusual in that the left and right are roughly equal in size. This has been the case for a long time and may help explain why polarization in the United States feels so intense.”
A tipping point or a pause?
One interesting wrinkle in the data is what happens after 2020. The most recent election cycle shows a slight decline in measured separation, raising the possibility that the long-term rise in polarization has slowed, or at least paused.
Researchers caution against concluding that Americans are becoming less politically divided. No single data point constitutes a trend, and levels of polarization remain much higher than before 2008. Still, this pattern emphasizes that polarization is not an inevitable, ever-growing force, but one that waxes and wanes in response to political, social, and institutional pressures.
Ultimately, this study provides one of the clearest empirical demonstrations to date that political polarization in the United States has not increased gradually over the past four decades, but has accelerated sharply since 2008.
What remains to be determined is why such changes occurred, or which forces are most responsible for further polarizing American opinion.
“In the past, someone who had a left-wing view on one issue might have a right-wing view on another issue. That’s less likely now,” Dr. Young said. He suggested that clearer ideological signaling from political leaders could play a role, adding: “I think this consolidation will occur as people take clear cues from political elites about what they align with. It makes average positions more extreme and widens the gap between Americans.”
Nevertheless, researchers urge people to refrain from allowing extreme polarization to become the new normal. Rather, polarization has a history of inflection points, and with the right tools, polarization can be measured, understood, and perhaps ultimately addressed.
Dr. DeWitt added, “Although political groups have become more divided in recent years, these divisions remain blurred and Americans are not completely divided into opposing tribes on this issue.”
Tim McMillan is a former law enforcement executive, investigative reporter, and co-founder of The Debrief. His writings typically focus on topics related to defense, national security, the intelligence community, and psychology. You can follow Tim on Twitter. @TimMcMillan. Tim can be reached by email at tim@thedebrief.org or by encrypted email. LtTimMcMillan@protonmail.com
