#BuildTheWall data visualizations & high-volume influencer networks
I captured tweets containing three hashtags in relation to Trump’s address to the nation on January 8, 2019: #BuildTheWall, #PaintOurCountryRed and #MAGA. Networks of high-volume accounts spread these hashtags, amplified the trends and drove up the quantity of tweets in each hashtag.
- #BuildTheWall — 25,724 tweets from January 8 to January 10
- #PaintOurCountryRed — 16,151 tweets from January 8 to January 12
- #MAGA — 17,560 tweets from January 8 to January 12
This is the second time I’ve created network graphs of hashtag #BuildTheWall and the results are similar to previous analysis which was connected to the Kate Steinle verdict. I’ve also documented similar high-volume activity in QAnon and Operation Backyard Brawl.
As I was analyzing these three hashtags, a video of students wearing MAGA hats from Covington high school mobbing Native American activists at the Lincoln Memorial in DC went viral causing a media shitstorm. The original tweet came from a suspicious account that was suspended and later discovered to be a paid influencer account from a website called shoutcart. The @2020fight account had 39,000 followers and tweeted hundreds of times per day. Shoutcart later told CNN that the account “was not part of any paid campaign” through its service.
There will likely be more news coverage of this account and paid influencers in general. But consider the following analysis to put this event into perspective. I have documented a large network of high-volume MAGA accounts. Each account has tens of thousands of followers and tweets hundreds of times per day. This army of influencers constantly boosts pro-Trump content on Twitter.
If a single influencer with 39K followers can make one video go viral, what kind of influence does this network of high-volume pro-Trump accounts wield?
#BuildTheWall — January 8 to January 10
User-to-hashtag network
I captured 25,724 tweets mentioning the hashtag #BuildTheWall from January 8 to January 10 using Tweet Archivist.
There are numerous high-volume accounts in this hashtag and QAnon accounts are a noticeable presence in the #BuildTheWall network.
The QAnon accounts are predominantly in the lower right quadrant of this graph, they are easily identifiable by their heavy edges and patterns created by hashtag spamming.
Here is the network from the overview screen in Gephi showing various accounts and hashtags connected to the #QAnon hashtag. They are scattered throughout the graph but mostly concentrated in the lower right quadrant.
Closeup of the QAnon cluster in 25,724 #BuildTheWall tweets from January 8 to January 10:
The same network filtered by degree range 20 reveals several hubs of activity coming from high-volume accounts. These accounts are mostly retweeting other #BuildTheWall tweets.
High-volume hubs are “super-spreaders” in hashtag networks. Most of these accounts predominantly retweet other people’s tweets. They act like an engine running in the background and driving up the quantity of tweets.
Viewing the filtered network in the overview screen (below) shows a QAnon hashtag cluster that stands out from the rest of the network. Those arrows generally appear in Gephi when accounts are spamming hashtags and/or when accounts are extremely active. In this case, the blue arrows are caused by an account spamming hashtags.
High-volume accounts
Here are some of the accounts that stand out due to their high-volume behavior and visibility in the #BuildTheWall hashtag network:
Alpha_Omega_Yah
This account was created December 7, 2018 and has already amassed a whopping 37K tweets — this is because the account is tweeting on average 1094 times per day. @Alpha_Omega_Yah was one of the top users for #BuildTheWall from January 8 to January 10.
ImmoralReport
ImmoralReport has appeared in previous networks and its behavior over time seems like a violation of Twitter’s automation rules. ImmoralReport is tweeting using a custom app called “StopMadness2” —the second iteration of this app. It was previously using an app called “StopMadness” which I documented in previous analysis. ImmoralReport has been active since 2015 and averages 717 tweets per day. This appears to be a clear example of malicious automation and spamming; the account is a bot and is a constant presence in hashtags relating to immigration, refugees and general anti-Muslim topics.
LadyRedWave & ouchinagirl
These two accounts are also familiar and have appeared in previous hashtags. LadyRedWave is currently averaging 127 tweets per day and ouchinagirl is averaging 94. Both accounts were hyperactive participants in #BuildTheWall and their accounts are high volume hubs in this network.
thebestcloser
This account is a political commentator with 11K followers. The account doesn’t tweet what I consider to be high volume (100K tweets per day or more) but it does spam hashtags which is why its activity creates these unusual patterns.
sherry25793049
This account was opened in June 2018 so it’s a relatively new account and is currently tweeting an average of 165 tweets per day.
SociallyMissy
sherry25793049 is a near neighbor to SociallyMissy, who doesn’t appear to be tweeting at a high volume but a glance at her timeline shows some social media marketing content (door dasher/work from home offers etc) along with retweeting Trump tweets and right-wing content.
SociallyMissy’s account also goes through periods of inactivity which can be viewed with the Account Analysis tool created by Luca Hammer. So calculating average daily volume of SociallyMissy’s account (total number of tweets divided by days since an account was created) does not appear to be very suspicious but the average tweets per day calculation is skewed by the periods of time where the account is almost dormant. SociallyMissy appears to be some kind of commercial/social media marketing account.
#BuildTheWall — January 8 to January 10
User-to-user network
Several of the accounts that appear in the user-to-user network are included in this network because many other accounts are mentioning them in tweets, not because they tweeted hashtag #BuildTheWall. The POTUS account and Donald Trump’s main account are a constant presence in these communities. The handles SenSchumer and SpeakerPelosi were also mentioned excessively in this dataset, which is why their accounts appear as prominent nodes.
In the upper half of the user-to-user graph, some of the high-volume accounts stand out clearly when I filter the network by degree range 50. thebestcloser is a near neighbor to Donald Trump’s account and has an abnormally heavy edge because he is constantly mentioning Trump’s account in tweets.
The lower half of this graph is a clusterfuck of the usual pro-Trump/QAnon/political spammers including heyitsCarolyn (290 tweets per day), LadyRedWave (127 tweets per day), steph93065 (163 tweets per day), and AnnaApp91838450 (290 tweets per day).
Here are the top users of the hashtag #BuildTheWall from January 8 to January 10 per Tweet Archivist:
The hashtag #BuildTheWall peaked on January 9 and per Tweet Archivist made over 182 million impressions in this 2 day timeframe.
I captured 16,151 tweets containing hashtag #PaintOurCountryRed from January 8 to January 12 using Tweet Archivist. This hashtag made over 193 impressions even though this dataset contains fewer tweets and represents 4 days in contrast to 2 days of the hashtag #BuildTheWall.
#PaintOurCountryRed — January 8 to January 12
User-to-hashtag network
Below is the user-to-hashtag network for this dataset which shows hashtag #BuildTheWall and a variety of other linked hashtags.
Two noticeable clusters in the upper left quadrant are patterns caused by two accounts spamming hashtags:
#PaintOurCountryRed — January 8 to January 12
User-to-user network
Below is the user-to-user network of 16,151 tweets containing hashtag #PaintOurCountry Red from Janary 8 to January 12. Once again Donald Trump, Nancy Pelosi, Chuck Schumer and the POTUS account appear in this network because many other accounts are mentioning their Twitter handles, not because they tweeted the hashtag.
I filtered this network by degree range 300, removing all nodes with less than 300 edges, which reveals a network of hyper-active accounts that tweeted this hashtag.
This hashtag was not as active as other hashtags in this dataset and according to Tweet Archivist, its volume peaked on January 11, however it still made over 193 million impressions in this four day timeframe.
#MAGA — January 8 to January 12
User-to-hashtag network
The user-to-hashtag network for #MAGA is connected to a lot of other hashtags. The #QAnon community is again distinct in the top left quadrant and #BuildTheWall is also very visible.
When I began filtering the #MAGA user-to-hashtag network by degree range, I found a very small number of accounts that are central hubs in this hashtag. Six accounts in particular stood out: Alpha_Omega_Yah, ImmoralReport, trumpnewsbots, PCadfael, curtislwalker and gbroh10. In the following graph which was filtered by degree range 30, the labels for these six accounts have been enlarged. All other labels shown in this graph are hashtags that these accounts are connecting to.
#MAGA — January 8 to January 12
User-to-user network
Below is the user-to-user network for 17,560 #MAGA tweets from January 8 to January 12. Again Donald Trump’s account was mentioned excessively in this hashtag. Lou Dobbs tweeted hashtag #MAGA in a tweet that received massive engagement so his account is very visible in this graph.
The hashtag #MAGA is in constant rotation on Twitter. During this timeframe it made over 104 million tweets according to Tweet Archivist.
Conclusion
I have documented these high volume accounts in various hashtags over the past two years and they seem to violate Twitter’s SPAM policy. I’ve kept a rolling thread of what I call “MAGA spammers” and update it when I find new high volume accounts. A few of the accounts have been suspended/deleted since I started the thread last July, but most are still active and cranking out high volumes of pro-Trump content.
I don’t know if these accounts are partly automated but as I’ve mentioned previously, I suspect many are run by real people.
Mike Rothschild published a fascinating piece in the Daily Dot about his experiment creating a fake account and joining this network of pro-Trump accounts that tweet constantly and have massive followings:
He refers to this hyperactive right-wing corner of Twitter as an “endless army of Twitter foot soldiers” and it’s accurate to call them an army of pro-Trump influencers.
They follow everyone they come across and participate in “follow trains” to increase their follower counts. Predictably, he found dubious content circulating in these communities:
My timeline became an inundated by the same accounts, some real and some fake, recycling the same conservative news, obsequious Trump worship, memes, gifs, Fox News clips, Bible quotes, patriotic glurge, follow trains, censorship complaints, random observations of how evil Hillary Clinton is, and posts pushing the right-wing outrage of the day.
Even more disturbingly, there was also a massive influx of fake news stories, particularly from notorious hoax sites True Pundit and YourNewsWire.
Their posts sent obviously fraudulent “news” careening through a network that has hundreds of thousands of people ready to believe it, and they’re retweeted thousands of times in the bubble.
If a single influencer account can catapult a video of kids in MAGA hats into a media shitstorm, then an army of large, pro-Trump accounts coordinating their tweets is a weaponized influencer network.
Gephi graphs in this blog were created using OpenOrd combined with Force Atlas 2 layout algorithms.
Tweets per day (TPD) cited above are via Social Bearing as of January 2018 and those metrics can fluctuate over time. TPD is calculated by the total number of tweets an account has tweeted divided by the number of days the account has been online. This metric can change if an account stops tweeting for a period of time or deletes tweets en masse.
For some accounts I used Luca Hammer’s Account Analysis tool to determine the tweet volume during the timeframes of the three hashtags I captured.
Impressions cited above are via Tweet Archivist who defines the metric as “the total number of times that the tweets of an archive have been delivered to Twitter streams. Of course, not everyone who receives a tweet will read it. As such, impressions are the largest possible audience for the given archive. Paid advertising works similarly; even though an ad was displayed on a website, there is no guarantee that a person actually saw it. Also, note that impressions does not deduplicate users, so if the same person sees a given hashtag twice, it counts as two impressions. Note that, because replies are only delivered to common followers’ timelines, they are calculated as a single impression.”
