How to talk like a Democrat or Republican: Language divide between U.S. political parties reaches historic High

New Study Finds U.S. Congressional Speech Began to Polarize in the 1980s

Democrats and Republicans in the United States live in different worlds. Democrats talk about "undocumented workers," "tax breaks for the rich," and "estate taxes," while Republicans refer to "illegal aliens," "tax reform," and "death taxes."

A new study finds that American political speech has become more polarized across party lines over time, with a clear trend break around 1980, and that current levels are unprecedented.

In the working paper, "Measuring Polarization in High-Dimensional Data: Method and Application to Congressional Speech," by University of Chicago Booth School of Business Associate Professor Matt Taddy, Stanford University Professor Matthew Gentzkow and Brown University Professor Jesse M. Shapiro, researchers developed a new approach for measuring political polarization that breaks down and analyzes the differences in rhetorical choices among U.S. political party members.

This model-based measure of polarization improves upon traditional standard measures and finds that, in contrary to previous studies, the polarization of speech was in fact quite low until the 1980s when it began a rapid rise.

"We've come up with a new way to measure segregation where the old methods didn't work," said Taddy. "When using this new method to model word choices, we find this pattern of clearly increasing polarization after 1980. And this just confirms what most people expect—that the parties are growing further apart in the subjects they care about and how they talk about issues."

To develop their model, the researchers used an automated script that sorted the text from the U.S. Congressional Record from 1873 to 2009 into individual speeches. They also matched the speaker of each line of text to the same lawmaker listed in the Database of Congressional Historical Statistics.

The "bag of words" approach disregards grammar and word order and focuses instead on counting the frequency of so-called bigrams, or two-word phrases appearing across speeches. The method allowed the researches to compile a sample of 723,198 unique phrases spoken a total of 271 million times by 7,285 unique speakers. The measurement method corrects for sample bias, a factor that may have led previous research to conclude that partisan speech was higher at the turn of the 20th century.

"These differences in speech may contribute to cross-party animus and ultimately to gridlock and dysfunction in the political system," the researchers said in the study.

The new approach to measuring polarization has implications for analyzing a wide range of divides. Researchers have already begun to apply the method to measuring segregation and web browsing behavior according to race, as well as segregation in product choices according to education level.