Sheyril Agarwal, Aanchal Sagwal, Joyojeet Pal
(to cite: Agrawal, S., Sagwal, A., Pal, J. (2022) The Post-Uvalde Twitter Discourse in the US, https://joyojeet.people.si.umich.edu/uvalde)
We examined what politicians across the US tweeted about following the Uvalde school shooting in the days immediately following the incident to see how the two main political parties reacted to the incident, and how, in turn, their followers engaged them.
We see two broad patterns. First, Democrats were 6 times more likely to talk about ‘Guns’ compared to Republicans, while Republicans were over twice as likely to talk about ‘Prayers’ or ‘God’.
Second, we find that both sides dig into their respective positions on the underlying issue – gun control, and that they are rewarded online when they stick to their position. Thus when Democrats take an aggressive stance of outrage or message in favor of stronger gun control, these tend to be more widely retweeted, and when they dodge or vacillate on the issue, they get strong pushback. On the Republican side, we find that those offering prayers or condolences get pushback, whereas digging in on not budging on gun control gets rewarded by their base.
Third, we find that states that lean Democrat were more likely to have politicians discuss the issue, while states that lean Republican were less likely to have politicians to discuss the issue. Several politicians on the Republican side, including leading Senators and prominent Representatives did not address the shootings at all during the first week following the incident.
We explain these findings in greater detail.
We collected the data for this work using two processes. The first approach sought tweets from the general public in the aftermath of the incident. These tweets were collected using a bag of words process through which we created a list of keywords and word combinations relevant to the incident (such as ‘Uvalde, Texas Shooting, School Shooting’ and several combinations of these words) as well as hashtags (#Uvalde, #RightToBearArm, #TexasSchoolMassacre, #EndGunViolence, #BanAssaultWeapons, #GunReformNow, #UvaldeMassacre, #RobbElementary) for the dates: 13/05/2022 to 27/05/2022…. We did a ground truth analysis to check what proportion were accurately identified and found our accuracy to be 95.3%.
The second approach sought all tweets from every known US politician handle in the US, using the University of Michigan database. Here, we pulled every tweet from the handles of 515 known politicians for the dates 21/05/2022 to 03/06/2022 to add a buffer of days prior to the incident. The goal with this data was to understand the extent to which politicians were talking about Uvalde, how they discussed the issue, and the kind of response they got when they discussed it.
We mapped the politicians tweets to conceptual buckets and enumerated the frequency of their occurrence. We identified 7799 tweets and retweets from Democrats and 4967 tweets and retweets from Republicans and classified them based on the frequency of conceptual buckets in each. The results show the broad spread of topics across the politicians of the two parties.
Table 1: Conceptual spread of topics discussed by Republican and Democrat politicians in the week following the Uvalde shootings
Figure 1: Daily visualization of common words used in tweets by Republicans and Democrats respectively
First, while only about 7 percent of the tweets explicitly mention Uvalde or Texas, the actual frequency is much higher, arguably the majority of tweets about weapons and school violence from politicians in both parties are related to the shootings. There are a few important patterns we see here, both parties are important for what is missing on their lists. Democrats overwhelmingly mention Guns and Assault weapons, violence and kids, whereas Republicans’ tweets focus instead on Biden, inflation & gas prices, and immigration – all of which the Democrats minimally talk about these. Both Republicans and Democrats talk about “America” and the nation, typically with varying tones and visions of what that means, but the concept is more than twice as central to Republican tweets.
There are a number of common topics. First, both parties talk about “honor” “sacrifice” and “family” at comparable rates. One reason for those (and the “America” tweets) is tweets related to memorial day. We see that God or prayers are more than two and a half times as likely to be invoked in tweets by Republicans as compared to tweets from Democrats, however, we see that mental health is much more likely to be talked about by Democrats. The overwhelming majority of the mental health-related tweets from Democrats were about the mental health awareness month rather than about Uvalde, whereas the majority of Republican tweets about mental health were post-Uvalde.
Figure 2: Interactive graph of politicians (Blue: Democrats, Red: Republicans) and the extent of positive and negative engagement with their messaging on Uvalde. The Y axis represents the sum of retweets to Uvalde related tweets for the corresponding Twitter user. The values on the X axis were calculated as the difference between the Total Retweets and the Total Replies to related tweets.
As we see in Figure 1 above, almost all the politicians were negatively engaged, including Barack Obama (who brought the George Floyd killing into his tweet about Uvalde), and POTUS. The vast majority of politicians fall to the left of 0 on the x-axis, with only @betoORourke @AOC and @joaquincastrotx getting significant overall positive engagement (higher retweets than replies). Only @laurenboebert from among major Republican representatives got overall positive engagement, and her tweets were generally unapologetic about their pro-gun stance.
Examining specific tweets, we find that the Uvalde shooting saw politicians from most parties facing a significant amount of attack from their constituents and those from citizens who vote for their opposite party. The way we measure this is two-fold. First, we look at the proportion of retweets to replies to their messages. This is a fairly consistent means of examining negative sentiment, since the main exception case where replies outnumber likes is expressions of personal loss or celebration to which people send condolences or greetings. While the tragic event did see a number of such tweets, which received a high number of replies from commiserations, this was not true for politicians. For this, we use the second method, of sentiment analysis of the responses, and found that replies, across the board, were largely a sign of negative sentiment aimed at the sender.
The most significantly attacked politicians were Greg Abbott, Ted Cruz, and Rep Tony Gonzales (Uvalde is part of his district). On the Democrat side, Kyrsten Sinema and Joe Manchin were most likely to get trolled. In general, politicians from both parties were attacked more than any other user type such as journalists or celebrities as the majority of the blue points (representing Politicians) lie on the negative end and Journalists/Commentators (represented by orange) and Media accounts (represented by yellow) find themselves on the positive end in Figure 2. We also find that about a week after the incident, politicians increased tweeting about immigration.
Figure 3: Interactive graph showing the extent of positive and negative engagement for Politicians, Journalists and Commentators and Media with their messaging on Uvalde. The Y axis represents the sum of retweets to Uvalde related tweets for the corresponding Twitter user. The values on the X axis were calculated as the difference between the Total Retweets and the Total Replies to related tweets.
We see in the examples of tweets in Figure 4 that most politicians get negatively engaged, compared to journalists and commentators whose messages do not get as much abuse. We find that the “thoughts and prayers” language invited much attack. In a series of tweets in Figure 1, we find that replies are higher than reposts or likes when prayers are mentioned by politicians. The responses to these typically suggest they are seen as inaction and deflection.
Figure 4: Responses to tweets calling for prayers
In all of these tweets above, we see that there are more negative engagements than positive, most went viral. We also see that in most such messaging, users with a significant Twitter following themselves engaging (we see three of the four tweets above with influencers responding to politicians). In general messaging from senior politicians got more blowback when they were seen as empty.
Canned responses were also often similar in tone, this got picked up by Twitter users and criticized for either being cursory and disrespectful to the memory of the event. In particular, when the responses were seen as too close to each other, coming from politicians known to be supportive of gun rights, the critiques were more pointed.
Figure 5: Similarly framed responses from politicians attacked by citizens
Another pattern was that of calling out politicians’ voting history or funding. Several republican politicians who tweeted about the issue were responded to by Twitter users pulling out records of their funding. While Ted Cruz was the most attacked from among senators, his colleague from Texas, John Cronyn, also got much negative engagement with his messaging (While the initial attacks against Cronyn were for eschewing Gun Control, the later attacks would be for proposing restrictions). We also saw across most politicians who tweeted condolences, if they had any NRA funding, that was called out and posted by other Twitter users. As we see here with Ohio senator Rob Portman, the tweet calling out his NRA monies is more engaged than his own condolence tweet.
Figure 6: Callouts against politicians who have taken pro-gun stances in the past.
In comparison, tweets that took a more antagonistic stance received less pushback, or at least comparatively more engagement from the core of the respective parties. For instance, both Representatives Marjoie Taylor Greene and Lauren Boebert got more positive affirmation for their aggressive stances on Twitter. Greene tweeted the blame on what she referred to a ‘Godless culture’ that ‘undermined masculinity’, while Boebert took aim at Beto O’Rourke, for his confrontation with Greg Abbott at a press conference in Uvalde. The incident of Uvalde Mayor Don McLaughlin abusing O’Rourke and telling him to leave the premises emerged as a rallying point for Republicans on Twitter.
Figure 7: Responses by Republican representatives Marjorie Taylor Greene and Lauren Boebert
While the more aggressively attacked politicians online were Republicans, Representatives and Senators from among Democrats who are typically more targeted by citizens with opposing views, such as New York representative Alexandria Ocasio-Cortez got high pushback. Besides Beto O’Rourke, the Democrat who got the most negative engagement from the Republican side was California representative Eric Swallwell, who explicitly called for a ban on assault weapons. We see in Figure 8 some of the retorts to the progressive politicians, in general, one consistent pattern is that influential accounts, typically commentators, activists or social media influencers with a significant following tend to be the most engaged when they reply to these Democrats. In both cases Brian Suits who confronted Rep Ocasio Cortez and Allie Beth Stuckey, who confronted Rep Swallwell, are radio/podcast hosts. Rep Swallwell was among the most vocal supporters or greater restrictions on access to guns.
Figure 8: Messaging by Democratic representatives, and responses from podcasters
A small number of Republicans took positions that deviated from the dominant stance taken by the majority of their party colleagues, keeping themselves open to conversations on gun control or critical of polarizing speech. These included Rep Adam Kinzinger of Illinois and Geoff Duncan, Lt. Governor to Brian Kemp in Georgia, both politicians who have taken explicit anti-Trump positions in the past. While such messages got less pushback compared to condolences, the engagements appear to be from across party lines, which may not necessarily be politically valuable to the candidates.
Figure 9: Messaging by Republicans critical of party’s dominant positions
While influencers with a following outside of social media spaces played a role in the messaging around the shootings including Guitarist Peter Frampton, basketball star LeBron James, and singer Taylor Swift, the message that was the most viral was by retired NBA star and coach of the Golden State Warriors, Steve Kerr. The message, an angry video clip from a pre-game press conference, went viral and directly criticized the “moments of silence” engagement in the aftermath of such events. A number of celebrities also engaged in exchanges with politicians – singer Nancy Sinatra posted a critical message to condolences expressed by Pennsylvania senator Pat Toomey. Actor Matthew McConaughey, who was born in Uvalde, would later become an important voice in the issue, make a plea in the White House for firmer gun control laws in the state.
Figure 10: Messaging from Nancy Sinatra, Steve Kerr, and Taylor Swift on the shooting
Most conversation around the Uvalde from influencers was critical of guns, a small number of well-recognized figures with significant social media following did come out on the side of defending gun rights. These were largely talk show hosts and podcasters or influencers who were primarily known for their online messaging. The influencer who made the most tweets, with the highest number of engagements on the side of gun rights or opposition to the Democrat positioning on the issue was author Tim Young. As with the tweets from politicians, when such messaging took an aggressive posturing against a politician, rather than directly pro-assault weapons or solely about mental health or prayer, these received more engagement. For instance, the tweet by talk show host Mark Davis received far more blowback for its focus on prayer and gun-grabbing than Young’s attack on Beto.
Figure 11: Highly engaged messages from conservative commentators
Progressive influencers who expressed a need for regulations around gun violence on average received more engagement online. Journalism professor Chris Christiansen, who put out a tweet about gun control in the United Kingdom, and screenwriter Adam Best who brought attention to AR-15s in a tweet (figure 7) both were very highly engaged. In general, we find that messages critical of access to deadly weapons from influencers had much more engagement than messages proposing that more access to guns (for instance by teachers and security personnel in schools). While the sample is likely skewed towards a more liberal audience, the gap we see here is very significant, not one of the top 100 most retweeted messages related to the Uvalde shooting in the first 48 hours were from conservative voices, and even the next 100 had just a handful, including a condolence tweet from Melania Trump, attacks on Beto O’Rourke, and a few tweets from Tim Young.
Figure 12: Viral messages with a critical take on gun policy
As with politicians who went aggressively against any conversation on gun control, we see also that the influencers, particularly those with a broad-based following, who used confrontational language in tweets significantly outperformed those who went with a more conciliatory tone. The Hodge Twins, a duo of comedians with a significant following on the right, and actors Randy Quaid and Kevin Sorbo all took unapologetic takes on defending the right to purchase assault weapons, or directly attacked Joe Biden, which in turn got a great deal of engagement. Framing the gun debate as an issue of emotive appeal to conservatives was a successful strategy. We see here that Kevin Sorbo adds transgender persons’ rights into his messaging on guns, while Mary Vought adds abortion – both of their tweets went massively viral. Another issue on which viral tweets from the right piggyback is that of COVID-19 vaccines.
Figure 13: Conservative influencers eschewing defensive positions on guns
We examined the differences between various states in terms of what language was used by politicians, as well as how citizens responded. California had the highest engagement rate in pure numbers, with 136 messages from representatives and senators on Uvalde within the first week (not including federal cabinet members from California), much higher than even Texas, at 79 tweets. However, as a share of the population, smaller northeastern states had much higher representation.
Figure 14: Map of US states based on normalized proportion of politicians talking about Uvalde in the study period
There was relatively little engagement on the subject from middle-America, and where there was, it was largely from Democrat handles. In Missouri, for instance, about a quarter of all messages from Democrat handles were about Uvalde, whereas of the 73 tweets from representatives, senators, and the Josh Hawley (hawleymo), for instance, posted several messages about China and his family, did not once address Uvalde. Jason Smith (JasonSmithMO) discussed gas, leftists, and SCOTUS, Rep Billy Long (USreplong) tweeted 33 times – including several tweets about inflation, memorial day, chinese communist party, baby formula, illegal immigration with no engagement on Uvalde during the study period.
A number of politicians on the Democrat side from states that are not traditional blue states also avoided the issue, including prominent senators. Joe Manchin, who tweeted several times on May 24, did not address the Uvalde shooting, while Kyrsten Sinema put out one tweet expressing condolences, and has since been trolled almost universally with all her messaging.
Figure 15: Messaging from Kyrsten Sinema being engaged by gun control advocates
We see that the share of conversation about guns versus prayers and mental health somewhat corresponds to the split between Republican and Democrat-dominated states, but because Democrats in largely Republican states were more vocal in areas like the Midwest and parts of the South, the conversations there appear to be more centered on Guns. This is also skewed by more Republicans defending guns in some of those states.
Figure 16: Map of US states based on normalized proportion of politicians between two selected bins of topics – “Guns” (including Assault Weapons, Second Amendment etc) vs “Prayers and Mental Health” (including God, Spirituality etc)
The most engaged messaging in the aftermath of the shooting was neither messages from influencers nor from politicians. Rather it was from citizens offering tribute to the children and teachers and their families.
Figure 17: Tributes to the children and adults who lost their lives at Uvalde
Finally, we see that Texas itself has significant variance in opinion on the path ahead. While Texas politicians like Cruz, Abbot, and Gonzales were at the receiving end of much negative engagement on Twitter, it is not clear any of it matters on the ground, most significantly seen through the election of Republican candidate and strong gun rights advocate Mayra Flores in an election that took place within three weeks of Uvalde. Flores, who posed with an AR15 during her campaign, spread misinformation on her Twitter feed that a border agent had killed the Uvalde shooter, and proposed armed security for schools following the shooting. While her opponent Dan Sanchez spoke in support of Republican John Cornyn for his Bipartisan gun deal, Flores remained consistent in her position.
Cornyn would eventually get booed in the Texas GOP meeting for his position on the gun deal, which was a much-diluted version of what the social media action for gun control was asking for.
Figure 18: Tweets by Mayra Flores, candidate for the TX 34th district, who won her election 3 weeks after the Uvalde shooting
Here is a link to the full Twitter data (tweet IDs).