A new study published in the Proceedings of the National Academy of Sciences suggests that positive group emotions, such as solidarity, can drive social media virality during moments of political crisis. During the 2024 U.S. election, partisan users engaged more with posts expressing ingroup support than with hostile content following threats to their own party’s leadership, particularly the attempted assassination of Donald Trump and Joe Biden’s withdrawal from the race.
Social media platforms have long been criticized for reinforcing division by rewarding extreme content. Prior studies have consistently found that posts expressing moral outrage or hostility toward opposing groups tend to attract more engagement in politically polarized environments. But researchers have also noted that positive group emotions, such as solidarity and collective identity, become more visible after shared traumatic experiences like terror attacks or military invasions.
One recent example comes from a 2022 study of the Russian invasion of Ukraine, which found that expressions of ingroup solidarity—not hostility toward the enemy—were the strongest predictors of engagement among Ukrainian social media users. That research provided early evidence that during moments of group threat, online communities may rally around shared identity and support, rather than amplifying aggression.
This line of thinking led the authors of the new U.S.-based study to consider whether similar dynamics might unfold in a domestic political context—not following a national tragedy, but after high-stakes events that disrupt party leadership. Lead author Malia Marks, a doctoral researcher in psychology at the University of Cambridge, was particularly struck by the expressions of unity she observed on social media in response to both Trump’s shooting at the Butler Rally on July 13 and Biden’s re-election campaign suspension on July 21.
“During the 2024 campaign trail chaos, I began to wonder if my colleague Yara’s findings from Ukraine would replicate in my home country despite the situational differences — this wasn’t a military invasion or terror attack, which have previously caused this ‘rally ’round the flag’ effect,” Marks told PsyPost. “Plus, America’s two political parties are famously polarized and hostile. However, as I saw reactions to the Butler Rally shooting and Biden’s exit on my own social media feed, I observed so many expressions of in-party solidarity that I began to wonder if threats to group leadership could be enough to trigger this pattern even amongst partisan Americans.”
To explore these questions, the researchers used CrowdTangle, a public insights tool from Facebook, to gather over 62,000 posts from 484 accounts affiliated with American politicians, major partisan news outlets, and high-profile political commentators. The posts were collected over a 25-day period—from July 5 through July 29, 2024—spanning the eight days before the Trump assassination attempt and the eight days following Biden’s campaign withdrawal.
Each post was assessed for content that expressed either solidarity with the poster’s political group (termed ingroup solidarity) or hostility toward the opposing party (outgroup hostility). These classifications were made using a combination of a large language model (GPT-4o) and manual coding. To ensure reliability, two human coders independently labeled a sample of 3,000 posts, and the classification model showed strong accuracy.
The team then used statistical models to test whether posts that expressed solidarity or hostility received more engagement. Engagement was measured by combining all Facebook reactions, including likes, shares, and emotional responses such as anger or care. The analysis was broken down by political affiliation and time period—before and after each major event.
Before either crisis took place, posts that expressed hostility toward the opposing political group were the most powerful drivers of engagement for both Democrats and Republicans. This pattern aligns with much of the existing literature showing that online outrage is often rewarded with attention and interaction.
However, this dynamic shifted sharply after each party faced a leadership crisis. For Republican accounts, posts that expressed solidarity received 36 percent more engagement than those without such content in the days leading up to the Trump shooting. After the assassination attempt, this jumped to 53 percent. Notably, during this same period, the impact of outgroup hostility on engagement decreased.
The pattern flipped again after Biden suspended his campaign. At that point, Republican engagement with solidarity posts fell to a 26 percent increase, while engagement with hostile content jumped to 51 percent above baseline. In short, Republican users gravitated toward unity in the wake of the Trump attack but returned to antagonism once the opposing party faced its own setback.
A similar pattern emerged on the Democratic side. Before the Trump shooting, solidarity posts received 20 percent more engagement than others. That impact disappeared in the aftermath of the attack, suggesting that Democrats were not rallying emotionally during an event that directly affected their opponents. But after Biden’s withdrawal, engagement with posts expressing Democratic solidarity nearly doubled, rising to 91 percent more engagement than neutral posts.
Meanwhile, Democrats’ engagement with posts attacking Republicans increased steadily across all three time periods, but solidarity overtook hostility as the dominant driver of engagement only after Biden’s campaign suspension.
These results suggest that users’ responses were closely tied to which party experienced the disruption. Expressions of unity became more engaging when one’s own group was perceived to be under threat, while reactions to the other group’s crisis often triggered a rise in hostility.
“Unlike previous papers testing the effect of threat on ingroup solidarity, we were able to evaluate both sides of the competition,” Marks explained. “What surprised me was that threats to one group seemed to trigger changes in the opposition group, not just the threatened group. Specifically, Democrats interacted much more with content hostile to Republicans after Trump was shot, whereas Republicans’ interactions with posts hostile to Democrats shot up after Biden’s withdrawal. I’m excited to do some follow-up work on this, exploring the idea that people may be sensitive to weakness in an outgroup as a prime moment for attack.”
The researchers suggest that their findings point to a more context-dependent understanding of online engagement than is often presented. While previous studies have highlighted the prominence of hostility in driving virality, this new work shows that expressions of solidarity can also play a leading role—but mostly in response to shared group threats.
“There’s lots of prominent literature showing that hostility and negativity are strong predictors of online engagement,” Marks said. “However, most of this work looks at static averages, rather than testing different predictors over time, matched to major historical events. By taking a more dynamic approach, we found that hostility is sometimes much less important than positive social emotions like solidarity.”
This echoes theories from social psychology, including social identity and intergroup emotions frameworks, which propose that people tend to favor their own group and that this favoritism becomes more pronounced when group identity is made salient by external events.
But as with all research, there are some limitations. The study’s timeline was relatively short, covering just a few weeks around the two political events. While the dataset was large and detailed, a longer observation period might have offered deeper insights into how long these shifts in behavior persist.
“I’d love to conduct a follow-up study using a much longer timeline, which could more robustly parse which types of threats lead to certain effects, but it’s becoming increasingly difficult and costly to access social media data,” Marks said.
Another limitation relates to the data source. Social media behavior may not reflect the emotional reactions of the general population. Those who post and engage with political content on platforms like Facebook may represent a more active or polarized segment of the public. The study also relied on textual signals, meaning nonverbal or visual content was not included in the analysis.
The research team is now developing experimental approaches to test these patterns more directly. “We are now working on an experimental study, whereby we can manipulate people’s exposure to social threat and different types of content,” Marks explained. “This will allow us to better understand the causal relationships between these variables.”
The study, “Ingroup solidarity drives social media engagement after political crises,” was authored by Malia Marks, Yara Kyrychenko, Johan Gärdebo and Jon Roozenbeek.
