One of the most retweeted tweets in the month following the shooting at Marjory Stoneman Douglas High School in Parkland, Florida, was this:
"High school students are talking on message boards to make April 20th the day they all walk out of school and don’t come back until Congress updates the gun laws as it’s their lives on the line. Plz retweet or share for visibility as this is a great idea. #FloridaSchoolShooting"
— Noah Evslin (@nevslin) February 16, 2018
With 101,705 retweets and 140,696 favorites, this tweet had a major impact in spreading the word about an event that would give voice to so many students across the country as they left their classrooms to protest gun violence in their schools. How has the social media conversation around mass shootings changed since Sandy Hook?
In this analysis, each event set includes thirty days worth of hashtag activity on Twitter related to each of six major mass shooting events. Starting with the Sandy Hook Elementary School shooting in Newtown, Connecticut, on December 14, 2012, the six events had varying levels of activity on Twitter, took place in a variety of settings, and seemed to have a major presence in the news cycle.
The Top 10 Hashtags of Each Event Set
More information on each of these events can be found in the Mother Jones Investigation US Mass Shootings, 1982-2018.
The Event Sets
First take a look at the basic stats of each event set. In comparing unique tweets, authors, and hashtags for each event set, we start to get an idea of the volume of each set. For example, while the Parkland event set had the most tweets, the Pulse event set had the most unique usernames. So, more people tweeted about the Pulse shooting using a hashtag, but more tweets were generated around the Parkland shooting. In fact, the mean number of tweets per username for Pulse is 1.6, and the Las Vegas and Charleston event sets have a similar mean. The Sandy Hook event set averages at 2 tweets per username, and the Charleston event set is 2.6. But the Parkland event set has an average of 3 tweets per username.
Another thing setting the Parkland event set apart is the split between the number of people who tweeted multiple times in the event set and the number of people who only tweeted once. For Parkland, 36% of the unique usernames in the dataset tweeted more than once in the 30 days after the event. For Charleston it was 30%, for Sandy Hook 25%, and 22% for the other three.
Next, start to look at the engagement generated by the hashtag activity in each event set. When you combine tweets, retweets, and favorites you can see that the Parkland event set has by far the largest reach of any of the event sets, totalling nearly 34 million. The activity generated by the next largest event set, Pulse, generated nearly 11 million.
When viewed as percentages, the 1,262,502 original tweets in the Parkland event set account for roughly 4% of the overall activity, whereas 795,936 original tweets account for 34% of the overall activity in the Sandy Hook event set. That's a drastic change in the effect of hashtag activity on the conversation around these events.
Hashtag Activity Timelines
Timelines of engagement (tweets, retweets, and favorites) from day 1 to day 30 for each event set show that the Parkland event set is again unique among the six event sets collected. For every other event set, activity peaked by the fifth day after the event, however the Parkland event set shows two peaks. The first occurred 8 days after the event on February 21, a day of action across Florida as groups of students travelled to Tallahassee to witness a gun control vote in the state legislature, met with the governor, staged walkouts and protests in support across the state, and appeared on national television in a CNN Town Hall event. The second peak occurred on day 29 or March 14, 2018, exactly one month after the mass shooting at Marjory Stoneman Douglas High School, and was also the date of the #Enough! National School Walkout.
Gun, Pray or Vote
Performing keyword searches in each event set reveals some details about what is being talked about in tweets that contain hashtags related to these events. In a comparison of number of tweets that mention "gun", "pray", or "vote" from day 1 to day 30 for each event set, the word "vote" least frequently appears in any set, really only showing up with any significance at all in the Parkland event set. The Parkland students responsible for organizing the March For Our Lives have made an effort to encourage young voters in much of their messaging, including launching a summer tour aimed at increasing voter registration. The Pulse event set overwhelmingly mentions "pray", including permutations like "prayer" or "praying", when compared to the other two keywords. When searching for mentions of guns in the event sets search terms included "gun", "rifle", "weapon", "ar15", "ar-15", "ar 15", "bump stocks", "bumpstocks", and "semiauto", in an attempt to capture several of the different specific conversations that have mounted in the gun debate after shooting events. Regardless of how large a portion of the conversation is occupied by "prayer" overall in each event set, it is a short-lived topic of conversation. For every event set, tweets that mention "pray" peak in the first five days, never to dominate the discussion again.
Identity, Politics, Weapons, or Emotions
In another attempt to suss out the content of conversation in hashtagged tweets, more extensive keyword searches got grouped into the following categories:
- Identity tweets mention an identity such as gay, black, or muslim
- Politics tweets mention a political figure, office or institution
- Weapons tweets mention guns, weapons, the 2nd amendment or the NRA
- Emotion tweets mention prayer, sorrow, anger, or similar expressions of emotion
It would make sense that identity would be a large factor of the conversation with both Pulse and Charleston, due to both shootings being classified as hate crimes. However, the Pulse event set is dominated by emotion, due in part to the large amount of activity around Lin Manuel Miranda's sonnet at the Tony Awards which contributed to #loveislove trending, while only the Charleston event set is heavily dominated by identity. In the Charleston event set there's a shift in the conversation around day 10, where tweets about emotion fall well below the amount of tweets about identity and politics. That's also the day President Obama eulogized shooting victim and pastor of Emanuel African Methodist Episcopal Church, Rev. Clementa Pinckney. Parkland is heavily dominated by discussions of weapons then politics, with the ranking being consistent for all four categories from day 3 to day 30 in the timeline.
Who are all these tweets directed toward? Among the tweets that include direct mentions of other Twitter users, tweets that mention the president (either the president's named account or the POTUS account) are far and away the largest group of mentions, and this holds true for every individual event except Charleston. In the Charleston event set, individual activists or activist groups (such as MoveOn) are mentioned most often. Tweets mentioning individuals during the Charleston event were directed most often at activists Bree Newsome and DeRay Mckesson, and the activity is the result of two very different hashtag campaigns. The #freebree campaign was a response to Newsome's arrest after she climbed a flag pole and tore down a confederate flag, and generated significant support for Newsome's release including fundraising $75,000 for her legal expenses. The #gohomederay hashtag sprung up in response to Mckesson's presence at Charleston protests after the event, and was rooted in hostility toward both the activist and his message.
Data collection method
Tweets were collected via python script using query search and limited to 30 days after and including the date of each event. Each event set began with location specific searches within the timeframe, for example #sandyhook or #parkland. Unique hashtags within the search results were aggregated and ranked by frequency. New searches were started from the most frequently occuring hashtags in the results. Each search was added to the event set, and unique hashtags aggregated and ranked by frequency again. Thus the event set grew organically until no new hashtags of note were discovered.
Sandy Hook Shooting, 12/12/12-01/14/13
#26acts, #26actsofkindness, #2a, #ctshooting, #demandaplan, #guncontrol, #guncontrolnow, #newtown, #newtownshooting, #newtowntragedy, #nowaynra, #prayforct, #prayfornewtown, #prayforsandyhook, #sandyhook, #sandyhookpromise, #sandyhookshooting, #standdownnra
Charleston Church Shooting, 06/16/15-07/17/15
#ameshooting, #black9, #blacklivesmatter, #breenewsome, #bridgetopeace, #charleston9, #charleston, #charlestonchurchshooting, #charlestonmassacre, #charlestonshooting, #charlestonstrong, #charlestonsyllabus, #charlestonunited, #chimewithcharleston, #clementapinckney, #confederateflag, #confederatetakedown, #droptheflag, #dylannroof, #emanuel9, #emanuelame, #freebree, #gohomederay, #guncontrol, #gunsense, #iamame, #iamcharleston, #notonemore, #prayersforcharleston, #prayforcharleston, #remembercharleston, #risingforcharleston, #standwithcharleston, #takedownthatflag, #takeitdown, #wewillshootback
Pulse Nightclub Shooting, 06/12/16-07/13/16
#gaysbreaktheinternet, #loveislove, #loveisloveislove, #lovewins, #nobillnobreak, #omarmateen, #oneorlando, #onepulse, #orlandohorror, #orlandolove, #orlandonightclubshooting, #orlandoshooting, #orlandostrong, #orlandounited, #prayersfororlando, #prayforflorida, #prayfororlando, #pulse, #pulsenightclub, #pulsenightclubshooting, #pulseorlando, #pulseshooting, #twomenkissing, #weareorlando
Las Vegas Strip Massacre, 10/01/17-11/02/17
#guncontrolnow, #jasonaldean, #jesuscampos, #lasvegasattack, #lasvegasmassacre, #lasvegasshooter, #lasvegasshooting, #lasvegasshootings, #lasvegasstrong, #lasvegasterrorattack, #mariloudanley, #massshooting, #paddock, #prayersforlasvegas, #prayersforvegas, #prayforlasvegas, #prayforvegas, #prayingforvegas, #stephenpaddock, #stopthehate, #vegasmassacre, #vegasshooter, #vegasshooting, #vegasstrong
Sutherland Springs Church Shooting, 11/04/17-12/05/17
#devinkelley, #devinpatrickkelley, #massshootings, #prayersfortexas, #prayforsutherlandsprings, #prayfortexas, #rylandward, #sutherland, #sutherlandshooting, #sutherlandsprings, #sutherlandspringsshooting, #sutherlandspringstexas, #texaschurch, #texaschurchmassacre, #texaschurchshooting, #texasmassacre, #texasshooter, #texasshooting, #texasstrong, #thoughtsandprayers
Parkland School Shooting, 02/14/18-03/15/18
#alexwind, #cameronkasky, #cnntownhall, #davidhogg, #davidhogg111, #emma4change, #emmagonzalez, #enoughisenough, #floridahighschoolshooting, #floridaschoolshooting, #floridashooter, #floridashooting, #jaclyncorin, #laurenhogg, #march4ourlives, #marchforourlives, #msdstrong, #nationalschoolwalkout, #nationalwalkoutday, #neveragain, #neveragainmovement, #neveragainmsd, #nikolascruz, #parkland, #parklandflorida, #parklandschoolshooting, #parklandshooting, #parklandstrong, #parklandstudents, #parklandstudentsspeak, #prayersforflorida, #prayersforparkland, #prayforflorida, #prayforparkland, #prayforstoneman, #prayforstonemandouglas, #schoolwalkout, #stonemandouglas, #stonemanshooting, #studentsdemandaction, #studentsstandup
Searches were centered around hashtags for multiple reasons. One is that it is an easy way to make sure search results were in some way directly related to the event being researched. Another reason is because the use of a hashtag implies some intentionality in being part of a larger conversation, and what I was really interested in with this project is the content and character of that larger conversation.
Six events of major impact in the six years between the Sandy Hook school shooting and the Parkland school shooting. Two events in schools, two events in churches, and two events in public entertainment venues.
Once each event set was collected, basic stats were aggregated. Overall engagement in terms of number of unique tweets and totals of retweets and favorites were calculated. Also calculated were the number and frequency of unique hashtags in the set, and the number of frequency of unique mentions. Finally, keyword searches were performed on each event set, and binary variables were created for each tweet with a 1 indicating the search string was present in the tweet and a 0 indicating it was not. Aggregated totals for keyword search results equals more than the total of unique tweets because multiple keywords may be present in the same tweet.