The return of political culture.

Two days before Donald Trump’s inauguration, the economist Daron Acemoglu published an essay in Foreign Policy under the title “We Are the Last Defense Against Trump.” American institutions, he argued, were less able to restrain a president than most people believed, and the defence would have to come from citizens and civil society. Coming from an economist best known for arguing that institutions, not culture, explain why some nations prosper, it was a striking turn.

Many political scientists said the same in the months after the election, writing in newspapers, on blogs and on Twitter that democratic norms among politicians and voters, and resistance from civil society, were what would keep American democracy intact.

I wrote this essay for a forum in the newsletter of the American Political Science Association’s Comparative Democratization section, which asked contributors for new ways to measure democracy. My answer was that measuring democracy in the age of Trump meant returning to political culture. That raised two questions: how much does political culture matter for whether regimes survive, and how can we measure it?

The rediscovery sat awkwardly with the field. The most prominent accounts of democratisation, democratic breakdown and authoritarian survival, by Barbara Geddes, Adam Przeworski and colleagues, Carles Boix, Daron Acemoglu and James Robinson, Jennifer Gandhi, Milan Svolik and Naunihal Singh, give little reason to think that falling public support for democracy threatens its survival. Most treat political culture as a product of regimes and their incentives, if they treat it as relevant at all. Beatriz Magaloni, and Steven Levitsky and Lucan Way, were partial exceptions.

If the evidence before 2016 said that political cultures hostile to democracy could not change institutions, the election of one president should not change that view. The real question was whether the evidence had ever said so.

By political culture I meant “deep-seated values and practices with relevance for the demands which citizens formulate in the political sphere, and which condition the nature and intensity of their political behavior.” On that definition, votes for authoritarian candidates and protests against stolen elections are political culture too, not only answers to a survey.

So even if fatigue and disillusion with liberal democracy were rising in Western countries and elsewhere, did we need to worry? I answered yes, and pointed to four channels that the comparative literature already documented.

None of these channels needs culture to be fixed or ancient. They need only that what citizens want, and how hard they act on it, can tip the outcome when a regime is under strain.

The same holds inside authoritarian regimes. Scholars of dictatorship usually study four survival strategies: repressing opponents, coopting them, governing well enough to satisfy people, and persuading them that the regime has a right to rule. Mass support and political culture condition each of them.

If political culture matters, the next problem is measuring it in ways that allow comparison across people, places and years. I divided the measures into two kinds. Textual measures capture what people say and write, their sentiments and preferences. Behavioural measures capture what they do. The division runs across quantitative and qualitative research and across observational and experimental designs. Laboratory experiments compare individuals measured close together in time; observational studies and long-running field experiments need measures that hold up between units and over years.

For decades the default was the survey. Ronald Inglehart and the World Values Survey pioneered cross-national comparison, and regional surveys such as the Afrobarometer and the European Social Survey followed. Tellingly, Michael Coppedge’s 2012 book Democratization and Research Methods has a 35-page chapter on “Political culture and survey research” that mentions other approaches only in passing.

Large-scale analysis of texts has since become feasible: parliamentary speeches, media reports, books and social media posts, which I called “primary bearers and expressions of political culture.” Behavioural measures have spread too, helped by economists’ long preference for revealed over stated preferences. The most striking example I gave was Nico Voigtländer and Joachim Voth’s study of German cities, which traces anti-Semitism from medieval pogroms to violence in the Nazi era. I expected the future to belong to behavioural measures drawn from automated digital data, such as searches, purchases and open government data, often with precise locations.

My dissertation uses both kinds. On the Danish–German border, turnout from 1871 and gymnastics club membership in 1865 measure civic habits long before anyone ran a survey. In African regions and among immigrants in Europe, the evidence comes from survey interviews.

This afterword was written in October 2026 and is not part of the essay.

Three and a half years after the essay appeared, Trump lost the 2020 presidential election to Joe Biden and refused to accept the result. On 6 January 2021, while Congress met to count the electoral votes and certify Biden’s victory, a crowd of his supporters stormed the Capitol. A week later the House of Representatives impeached him for incitement of insurrection, by 232 votes to 197. After he had left office, the Senate voted 57 to 43 to convict, short of the two-thirds majority required, and he was acquitted. He won the presidency again in 2024 and began a second term on 20 January 2025.

In February 2017, in a piece for the Danish science site videnskab.dk, I had listed his threat not to accept a lost election among the campaign remarks that worried scholars of democracy, and described the scenario of a president who loses an election and refuses to step down, as a risk rather than a forecast.

What the essay adds is a reminder of where to look. Much of what decided those weeks was behaviour rather than opinion: officials who certified results, legislators who counted the votes, and the people who tried to stop the count. That is political culture in the essay’s sense, and it can be measured.