22 October 2025

When American lawmakers stand up to speak in the U.S. Congress, their words do more than fill the record — they shape the laws that govern society. In a new doctoral dissertation at Linköping University, sociologist Hendrik Erz has analyzed more than 17 million speeches and 100,000 roll-call votes to reveal how political language reflects, and even predicts, the way representatives vote.

Hendrik Erz.
Hendrik Erz, newly awarded PhD in Sociology at Linköping University. Photo: Vsevolod Suschevskiy

On October 20, 2025, Hendrik Erz successfully defended his PhD thesis On the Record: Understanding a Century of Lawmaking in U.S. Congress Through Speech and Vote Behavior at Campus Norrköping. His research offers a rare, data-driven view of how American democracy actually works — not only through votes, but through the everyday practice of speaking, persuading, and negotiating in Congress.

Words That Shape Decisions

“Lawmaking is not only about how people vote,” says Hendrik Erz. “It’s also about how they talk, the words they choose, and how they justify their positions. The way representatives speak reveals a lot about the motives and pressures behind their decisions.”

Using advanced text-analysis methods and computational social science, Hendrik Erz studied more than 130 years of congressional debate to trace long-term changes in political behaviour.

His findings show that speeches and votes are deeply intertwined — and that shifts in language often foreshadow shifts in policy.

From Crisis to Change: The 1970s and 1980s

One part of the dissertation focuses on the 1970s and 1980s, a period of economic crisis and political transformation often described as the “neoliberal revolution.” During these decades, the United States moved away from the post-war consensus toward market-oriented policies.

Hendrik Erz’s analysis of congressional debates from this era revealed how politicians’ social backgrounds, education, and party affiliations shaped the way they spoke about taxation, regulation, and the federal budget. Lawmakers with higher education and economic experience were more likely to promote market-driven solutions, while others defended the traditional role of the state.

Although representatives across party lines often discussed the same issues, their tone and framing differed sharply.

“The crisis years became a turning point in how American politicians talked about the economy,” Hendrik Erz explains. “Their shared experience of uncertainty and crisis forced them to redefine what responsibility meant in economic policy.”

Measuring Polarization Through Language

Another part of the research explored the growing polarization of U.S. politics. Using network analysis, Hendrik Erz mapped how members of Congress connected through the language they used.

His results showed that polarization, while strong at the Congressional level, was less visible within individual policy areas. Lawmakers from different parties still used more similar language when discussing specific topics such as trade or welfare, even as overall partisan divisions deepened.

“The way politicians talk is a mirror of how parties build their identities,” Hendrik Erz notes. “They increasingly talk about different things, not necessarily in completely different ways.”

When Words Predict Votes

In the final section of the thesis, Hendrik Erz linked congressional speeches directly to voting behaviour. By studying how representatives spoke before roll-call votes, he discovered that it was often possible to predict when a member would break ranks with their party.

He developed a new measure of “party pressure” — the degree to which parties influence their members’ voting behaviour — and found that this pressure has fluctuated significantly over time. In some periods, lawmakers enjoyed substantial freedom to vote according to their convictions, while in others, party discipline was much tighter.

“Speeches are often signals,” Hendrik Erz says. “When someone plans to defy their party, you can often hear it between the lines in what they say on the floor.”

Beyond the Votes: Understanding Democracy Through Data

Taken together, the results show that congressional lawmaking is not merely the mechanical counting of votes, but a complex social process shaped by language, identity, and institutional power. Hendrik Erz’s work bridges sociology, political science, and computational linguistics, demonstrating how big data and text analysis can reveal the underlying mechanisms of democratic decision-making.

“By combining speech and voting data, we can better understand why certain policies emerge and others fail,” he concludes. “It’s not just about parties or ideology — it’s about people, their experiences, their values, and the words they use to express them.”

Contact

Read more about Computational Text Analysis

Organisation

Latest news from LiU

Red blood cell.

AI method for determining time of death is awarded prize

Jenny Arpe’s master’s thesis at LiU presents a new method for determining the time of death using AI. For this she has now been awarded a master of science in engineering thesis prize awarded in memory of Christopher Polhem.

Linköping University to AI-proof its study programmes and courses

LiU is introducing a new action plan to AI-proof all its programmes and courses.“AI is already changing the way we learn. We need to ensure high quality and remain relevant in an AI-driven future,” says Deputy Vice-Chancellor Karin Axelsson.

Jendrik Seipp.

Research on next-generation AI planning receives SEK 15 million

LiU researcher Jendrik Seipp has been awarded SEK 15 million to develop an AI planning system that uses multi-core processors for parallel computation. This could lead to more efficient logistics and large-scale energy optimisation, among much else.