We benchmarked the performance of our approach against a range of existing measures of related estimands. Our measure outperformed these relevant alternatives. We show the performance of our large language model estimators (“SBERT-LLM”) versus considered alternatives in the table below.

We use this method in a case study of U.S. news website coverage of the war in Ukraine. We show that low quality U.S. news sites were more likely than mainstream U.S. news sites to have overlapping claims (narrative similarity) with Russian newspapers.

5/ Using a counterfactual simulation, we estimate that if the difference between Republicans and Democrats had remained at its 1992 size, the aggregate subsequent downturn in inequality perceptions would have been halved.

4/ These figures show that the downturn was asymmetrically concentrated among identified Republicans, especially Republicans with middle and high socio-economic status. Patterns of inequality perceptions reflect a growing class divide within the Republican party.

2/ We show that trends in perceptions of economic polarization are inversely correlated with trends in actual distributional inequality, with a surprising decline in the share of Americans who perceive inequality to be growing during the 1990s and early 2000s.

- In contrast to popular speculation, we show that scripted content is not only on ideological topics
- Using a case study of domestic coverage of COVID-19, we demonstrate how the regime uses scripting to shape, constrain, and delay information during crises

- We show that in China, scripted state propaganda is a daily phenomenon
- On 90% of days from 2012-2022, the vast majority of party newspapers printed scripted propaganda
- Scripted propaganda is also common in commercial newspapers, and its prevalence increased 2012-2022