What polarization data from 9 countries reveals about the US

America is an outlier, but not in the way you’d think.

My book on US political polarization is about to publish (preorder here!), which means I’ve begun getting a lot of questions about polarization. One of the most common is: Do political polarization trends in other countries track what we’ve seen in America?

It’s a good question! If they don’t, then what’s happening in American politics is distinct to our system, society, and history. If they do, then it’s part of a larger story, perhaps triggered by changes in global technology, migration, or trade.

The only problem with the question is I couldn’t find a good answer. Political polarization is a tricky thing to measure, and there weren’t good data sets that covered comparable countries over extended time periods.

Until now, that is.

In an important new working paper, Levi Boxell, Matthew Gentzkow, and Jesse Shapiro built the closest thing to a usable data set I’ve seen. Using survey results, they construct a measure of polarization that permits comparison across the US, Canada, Britain, Germany, Australia, New Zealand, Norway, Sweden, and Switzerland over the past four decades. What they found is striking: In five of those countries, polarization actually declined in recent decades. In no country did it rise as quickly as in the US.

Take a look at the chart, and then let’s dig into the findings in detail:

Cross-Country Trends in Affective Polarization, NBER.

A few things to note here. First, this is measuring “affective polarization” — that is to say, the difference between how warmly people view the political party they favor and the political party they oppose. The surveys they’re building off of often use slightly different language. In America, the dominant measure is the American National Election Survey’s feeling thermometer, which rates feelings toward the two parties on a 1-100, cold-to-hot scale. Other surveys in the international data set ask about sympathy, or favorability, or how much people like the parties. They’re all measuring similar things, but the fact that they’re not the exact same measurement is worth knowing. This is the best data set we have, not the perfect data set we wish we had.

That said, a couple of findings jump out:

• Party polarization is not uniformly increasing in Western democracies. In a number of countries, it’s actually decreasing. That is to say, there’s nothing inevitable about rising party polarization in this era. That allows us to reject a couple possible explanations.
• One theory this lets us reject is that polarization is a byproduct of internet penetration or digital media usage. Internet usage has risen fastest in countries with falling polarization, and much of the run-up in US polarization predates digital media and is concentrated among older populations with more analogue news habits.
• The researchers also test measures of migration, trade, and income inequality, and find that none of them track international trends in polarization.
• There is some evidence that rising ethnic diversity correlates with polarization. The study’s authors note that “the increase in the non-white share has been twice as large in countries with rising affective polarization as in those with falling affective polarization.”
• To correct for the fact that some of the countries in the data set are multi-party systems, the researchers also ran the numbers restricting them to the top two parties. The results didn’t change much: Canada polarized faster, and Britain, Norway, Sweden, and Germany de-polarized faster, but that might have been an artifact of fewer usable surveys and restrictions that didn’t quite fit the political systems being measured. (On a related note, Lee Drutman has an interesting piece arguing multiparty democracy can curb polarization.)
• In the US, “the timing of the introduction of Fox News appears roughly consistent with the acceleration of the growth in affective polarization during the 1990s.” Hmmmm.

All in all, the researchers conclude that the “findings are most consistent with explanations of polarization based on changes (e.g., changing party composition, growing racial divisions, the emergence of partisan cable news) that are more distinctive to the US.”

But I’d argue the most important implication is one they don’t quite state, and for reasons I’ll explain, may not even endorse: America is less of a polarization outlier today than in the past.

The authors kindly shared the underlying data set with me over email, and it reveals something interesting. If you average out the earliest polarization measures in the non-US countries, the difference in measured feeling between the party people favor and the party they disfavor is 41. If you average out the final readings in the series, you get a score of 40.

Now look at the United States: the earliest measure in the data set is a score of 27 — 14 points less than the international average. If you look at the final score in the data set, it’s 46 — 5 points more than the international average.

A hitch here is that the polarization data starts at different times in different countries. But if I restrict the analysis to countries where the first reading is within five years of the first reading in the US, the difference is still more than 10 points — more than twice the difference at the end of the data series.

Again, this is a rough cut of rough data. But it’s interesting to me, because everything I’ve learned in my polarization research suggests that America’s relatively non-polarized mid-century period was the aberration, and the much more polarized party system we have now is closer to both historical and international norms.

In the mid-20th century, divisions within both the Republican and Democratic parties on race and civil rights warped the party system, with the two main parties containing what was functionally a four-party system: Democrats, Dixiecrats, liberal Northeastern Republicans, and Republicans. The Dixiecrats, in particular, were unusual: They were often conservative, but for reasons of history (the Republican Party had invaded and occupied the South in living memory) and strategy (amassing congressional seniority in the majority party helped them protect white supremacy in the South), they remained in the Democratic Party until the Civil Rights Act set the rupture of that alliance into motion.

This is also a reminder that there are worse things than polarization: the acceptance of white supremacy that held the Democrats and the Dixiecrats together was an abhorrent compromise, and reflects the reality that the alternative to polarization is often suppression, not consensus.

Because American political memory uses the 20th century as a baseline, there’s a dominant assumption that the current, high state of polarization is the aberration, and the relative comity of mid-century American politics the state of nature. But the truth is more likely the opposite: the mid-century system was weird, party polarization is natural and here to stay, and the path toward a functioning political system runs through reforming the structures of American government to work amidst polarized parties.

I wouldn’t go so far as to assert that this international data set proves that argument right. But it’s consistent with it. The simplest interpretation of cross-country polarization trends is that America’s level of party polarization is reverting toward the international mean as the unusual forces that restrained party polarization here dissolve. That isn’t to say polarization isn’t creating a crisis in American governance — I think it is. But that crisis reflects the interaction between polarized parties and our weird political system, which requires high levels of cross-party consensus to function. In other political systems, where winning elections means gaining a governing majority, polarized parties are much less of a long-term threat.