At the end of June, Nancy Pelosi and House Democrats announced that they will form a select committee to investigate the Jan. 6 insurrection. The investigation will examine why and how a group of extremists — armed with pepper spray, stun guns, baseball bats, lead pipes and flagpoles — could storm the U.S. Capitol building in an effort to overturn the presidential election.
But while some in Donald Trump’s voting base may have approved of the attack, not every Republican did. Using data from millions of Twitter profiles, our research finds that the riot dramatically decreased expressions of identification with the Republican Party and Trumpism across the country. And that drop lasted.
This highlights the choice Republican senators recently faced when rejecting a bipartisan independent commission to investigate the Jan. 6 attack: they risk alienating either the radical partisan base who approved of the insurrection or moderates who disapprove of political violence.
How Twitter reveals the insurrection’s effect on political self-identification
We investigated how the Capitol insurrection affected the way millions of social media users choose to present themselves politically in their Twitter “bios.” Social media bios are useful measures of identity because they unobtrusively document how ordinary people self-identify.
To conduct our research, we collected a sample of roughly 3 million U.S.-based Twitter users and tracked their bios each day beginning June 1, 2020, noting any changes in the text. To date, we have amassed roughly 1 billion daily snapshots of the public bios of these users.
To be sure, Twitter users are not a representative slice of the U.S. population. According to the Pew Research Center, Twitter users tend to be somewhat younger, more educated and more politically engaged than Americans on the whole. However, since people who are more politically engaged tend to have more fixed political identities, this skew should work against seeing changes in how people identify. If expressions of political identities are malleable in this relatively more engaged subset of Americans, the same is likely the case among the rest.
Twitter allows users to describe themselves using any text or emoji that can fit into 280 characters. To measure whether a user expresses identification with the Republican or Democratic Party, we use a machine-learning technique to find keywords associated with Republican and Democratic identification, such as “trump,” “#maga” and the blue wave emoji 🌊. We then consider any user listing one or more of these keywords as expressing a party identification.
After the Capitol riot, identification with the GOP and Trumpism dropped dramatically
On most days, we found little change in the overall number of users identifying themselves with the Republican Party: users added and removed Republican terms in roughly equal numbers.
We found two notable exceptions. After Joe Biden was declared the winner of the November election, we observe an increase in users who removed terms identifying themselves as Republicans. But after the Jan. 6 Capitol insurrection, the number of users removing these terms was much more pronounced.
We wondered whether the removal of political identification was because Republicans were appalled by the insurrection or whether both Democratic- and Republican-identifying users felt distaste for politics and turned away. To test for this, we use a statistical technique that allows us to compare changes in Republican Party identification relative to changes in Democratic Party identification.
We found that before Jan. 6, changes among Republican and Democratic-identifying users remained roughly equivalent. But afterward, Republicans become far more likely than Democrats to remove mentions of partisanship from their bios. In the three weeks immediately following the insurrection, we see a substantial 1 in 14 users (a little over 7 percent) net drop in the use of partisanship terms among Republican-identifying users relative to changes in expressions of partisanship among Democratic-identifying users.
On Twitter, Trump is more popular than Jesus
Those Republicans who shed their identification after Jan. 6 didn’t add it back
If users are so quick to remove expressions of partisanship association from their bios as a result of the insurrection, did they simply add them back shortly thereafter? Because we track every user each day, we can answer this. We examine those users who removed partisan-identifying information in the immediate aftermath of the insurrection, and tracked changes in their bios over the following month to see if they once again add partisan terms.
This was not the case. Only 5 percent of those who removed their Republican identification added it back within the following month.
In other words, the Capitol insurrection pushed a large number of previously Republican-identifying Twitter users to step away from explicitly identifying with the Republican Party.
Why Facebook really, really doesn’t want to discourage extremism
What does this tell us?
The Capitol insurrection was one of the most remarkable examples of a violent attack on U.S. democratic institutions in recent times. And it seems to have pushed some portion of citizens away from identifying with the Republican Party, which has been affiliated with the insurrection. In other words, while political violence that violates democratic norms may hearten extreme partisans, it can also bring intra-partisan backlash or demobilization. This is consistent with political scientist Omar Wasow’s recent scholarship finding that while nonviolent 1960s civil rights protests boosted the local Democratic share of the vote, violent Black-led protests did the opposite, depressing the Democratic vote. By contrast, political scientist Ryan Enos and co-authors found that the 1992 Los Angeles riots after the Rodney King verdict may have increased public support for its causes.
Nevertheless, our findings suggest that partisan violence has political costs, which might be underestimated by the politicians who court it.
Don’t miss any of TMC’s smart analysis! Sign up for our newsletter.
Gregory Eady (@GregoryEady) is assistant professor of political science at the University of Copenhagen.
Frederik Hjorth (@fghjorth) is associate professor of political science at the University of Copenhagen.
Peter Thisted Dinesen (@ThistedPeter) is professor of political science at the University of Copenhagen.