Social media rewards politicians for negative partisan attacks, study finds

Members of the United States Congress have become increasingly partisan and negative in their online communications over the last decade. A massive study evaluating millions of social media posts and emails found that politicians receive disproportionately high engagement when they attack political opponents rather than sharing legislative accomplishments. The findings were published in the American Political Science Review.

Political representatives have long used a variety of strategies to connect with their constituents and secure reelection. In the 1970s, political scientists established that legislator communications traditionally fell into a few universal goals, such as promoting a personal brand, claiming credit for legislative accomplishments, or taking stances on issues voters care about. For decades, politicians relied on mailers, televised press conferences, and local newspapers to achieve these goals.

The rise of the internet fundamentally altered this dynamic. Digital platforms allowed elected officials to bypass traditional news outlets and broadcast their messages directly to a national audience. This technological shift occurred alongside a growing ideological divide within the American electorate.

Michael Kistner, a political scientist at the University of Houston, and a team of colleagues wanted to understand how this shift to digital platforms altered the way elected officials talk to the public. The research team noted that previous investigations often focused on short timeframes or single platforms. Those limited scopes made it difficult to spot long-term trends or compare different types of audiences.

To build a more comprehensive picture, the researchers gathered a massive dataset of online communications from members of Congress spanning from 2009 to 2022. They collected approximately 4.7 million posts from Twitter, 2.4 million posts from Facebook, and 184,000 official email newsletters. The dataset encompassed nearly the entire era of widespread social media use in politics.

Free daily newsletter

To analyze this vast amount of text, the team utilized computational language models to categorize messages based on their intended purpose. The researchers trained these models on a subset of messages manually reviewed by human experts. The models then evaluated the remaining millions of sentences to determine if a politician was advertising an event, claiming credit for a policy, assisting constituents, or engaging in negative partisan attacks.

The system also analyzed the vocabulary used in each message to scale the text on a continuous partisan spectrum. Extreme partisan speech often features vocabulary used almost exclusively by one political party, while moderate speech employs language shared by both sides. A computer algorithm assigned each post a partisanship score based on how strongly its language identified the speaker’s political affiliation.

Before running their main analysis, the researchers validated their computer models against other common measures of political behavior. They compared the language scores to members’ positions based on roll call votes, how often they voted in line with the president’s stated preferences, how they raised campaign funds, and the issues they highlighted on their official websites. The text-based scores aligned well with these established benchmarks. Members who sponsored bipartisan bills in real life tended to use more bipartisan language online, and those who frequently mentioned local towns in floor speeches were more likely to post about constituent services.

The researchers then mapped out the rhetorical trends across the 14-year period. The analysis revealed a steady, substantial increase in partisan polarization in congressional rhetoric across all three communication platforms between 2011 and 2022. The language used by Republicans and Democrats diverged markedly as the decade progressed. Topics of conversation shifted away from routine governance and toward partisan conflict.

While the partisanship of Democratic speech remained largely flat over the decade, with no statistically significant trend, Republican rhetoric shifted heavily toward the extremes. The data showed that Democrats increasingly focused their messaging on progressive priorities, using words related to climate, democracy, and infrastructure. Republicans pivoted toward immigration and combative rhetoric, with frequent references to the border and the opposing party.

In addition to becoming more ideologically separated, the specific types of messages politicians prioritized shifted over time. The portion of communications dedicated to negative partisan attacks grew rapidly. Negative attacks represented about five to ten percent of messages in the early years of the study but increased to nearly twenty percent by the end of 2022. Position-taking, where a politician simply states their stance on a controversial issue, also grew to dominate the digital conversation.

At the same time, routine governance topics became less prominent overall. Messages about delivering federal funding for a local district or offering constituent assistance failed to keep pace with the rise in partisan attacks. The researchers noted that the sharpest increase in negative partisan messaging by Democrats occurred after the 2016 presidential election. Republicans showed a similar spike in negative messaging following the 2020 election.

To explore a potential incentive driving these rhetorical shifts, the research team looked at how the public responded to different types of messages on Facebook and Twitter. Social media platforms provide immediate, quantifiable feedback that politicians and their staff use to evaluate their communication strategies. The research team tracked engagement metrics such as likes, shares, retweets, and comments for each post in their database.

The researchers then compared the engagement levels of negative or highly partisan posts against those focused on bipartisan cooperation or local constituent services. Negative partisan attacks generated vastly more attention than any other type of communication. On Twitter, negative attacks received 60 percent more likes, 80 percent more retweets, and 79 percent more replies than an average post by the same politician. Facebook yielded similar results, with partisan attacks gathering 86 percent more shares and 93 percent more comments than routine updates.

Conversely, messages about bipartisan collaboration or local district services received less engagement than the typical social media post. Credit claiming, which involves a politician taking responsibility for a positive government action, also failed to attract high levels of online interaction. The data suggests that social media algorithms and user behaviors heavily reward hostility and ideological extremity over traditional legislative work.

A casual reader might assume that the high engagement for negative posts reflects the genuine desires of the broader electorate. The researchers caution against this interpretation, pointing to evidence that the messaging that attracts engagement online differs from the messaging voters say they prefer.

In separate survey data, researchers regularly find that most voters prefer representatives who focus on legislative achievements rather than partisan bickering. Politicians may be falling victim to a “social media feedback mirage,” mistaking digital likes and shares from a polarized online audience for true political support. When elected officials rely on these engagement metrics to evaluate their performance, they risk becoming trapped in a cycle of escalating negativity that fails to serve the broader public.

The study’s data concludes in 2022, which means it does not capture the recent rise of alternative platforms like Bluesky and Truth Social. These newer networks cater to specific ideological groups and may fragment political audiences further in the future. However, the researchers noted that the vast majority of elected officials still maintain active, primary profiles on Facebook and X, the platform formerly known as Twitter. Future research will need to examine whether politicians adapt their strategies as these digital environments evolve and as visual platforms like YouTube and Instagram gain prominence.

The study, “Measuring Partisanship and Representation in Online Congressional Communications,” was authored by Michael Kistner, Michael Heseltine, Robert Alvarez, Maya Fitch, Lucas Lothamer, and Elizabeth Simas.