Donald Trump has controlled the Twitter narrative about himself for the past 48 hours. Proof? Metrics. Here’s “How We Use Twitter, Part 2.” This one is centered on the hashtag #notmypresident, taken over the last 24 and 48 hours.
My initial plan to run a new test on the hashtags #medieval OR #medievaltwitter got deep-sixed yesterday morning when I had to restart my computer and in the process lost the data I had captured with Trendsmap, which I’m trying for a week. The quantity of Tweets fell off sharply, however, when I tried to recapture data in NodeXL (so that both captures would have roughly the same time stamp). The discussion would, therefore, be rather bland. So, with no immediate followup to the #medievaltwitter post in sight, I decided to go for current politics instead. Part 1 of this analysis can be considered a demo and recommendation of Trendsmap. Part 2 dives into data collected from Trendsmap and NodeXL.
Part 1 #notmypresident for the past 24 hours, c. 1730 hours January 4, 2017
Trendsmap is a growing and quite exciting platform for analyzing and visualizing Twitter data in real time. I’ve been testing it out for a few days with their gracious 1-week trial, running various searches on it. So far, I’m very impressed with Trendsmap, and though I can’t afford a subscription now, I am definitely interested in pursuing it later, or recommending it to my school.
In trial mode Trendsmap lets you examine tweets from the past 24 hours. Under analytics, we can see that there were 14, 500 tweets featuring the #notmypresident hashtag, with a 45%-55% female-male split, with peak activity around 8 p.m. on January 3.
When we look at the map, it seems almost all the activity comes from inside the U.S., with a couple exceptions–the UK, Australia, and New Zealand being the main exceptions, with other bursts in Israel, Nigeria, and the western Mediterranean.
Trendsmap lets you zoom in to examine the specific tweets that were the most influential, with interactive map features:
The details of the tweets reveal some interesting features. The second most popular hashtag was #theresistance, which could be further parsed to identify the community that opposes Trump the most (NodeXL’s “word pairs” feature indicates the same thing:).
Back to Trendsmap: New York City, Portland OR, Los Angeles, D.C., and Seattle account for a quarter of the tweets, though we can tell from the heat map that opposition to the incoming administration is either most widespread or most vocal east of the Mississippi. Roughly 44% of the tweets come from iPhones or Android devices, indicating (perhaps) a particular socio-economic background of the tweeters (someone dig up the research and correct me on this…).
Part 2: Specifics, NodeXL, and Analysis
In Trendsmap: When we come to look at the Top Retweets vs the Most Quoted Tweets, an interesting pattern emerges (again, one that NodeXL helps us go into in more detail. The top retweet had nothing to do with any specific Trump action or tweet, but rather was a meme reaction to “when you find out your friend voted Trump.” None of the top retweets came from the PEOTUS himself. However, he snagged four of the top 12 “Most Quoted Tweets,” three of which were retweeted far more than @KingBach’s tweet. The leader was his tweet about the intel briefing being moved to Friday, the second defended Assange, the third announced his press conference, and the fourth talked about Jackie Evanchos’ album sales after announcing she would perform at the inauguration.
When it comes to “Most Replied Tweets,” PEOTUS Trump took all twelve spots for the last 24 hours. The “Most Influencing Tweets” are a mix of tweets seen in previous categories, but none are Trump’s. Trendsmap calculates “influence” here as accounts with the most followers–in other words, closeness centrality.
In NodeXL: Comparing these results with NodeXL, which for January 3 and 4 pulled 11,272 individuals who tweeted, and 26, 563 connections (again, the graph and workbook can be accessed at the NodeXL Graph Gallery). Note that NodeXL and Trendsmap display data differently, so the individuals often count for more than one tweet.
Here’s the data clustered into groups:
And here’s the data, rearranged according to the Harel-Karem Fast Multiscale layout and with groups in their own boxes:
Without going into the different groups (perhaps a topic for a later date, though politics will have moved on by then), let’s confine ourselves to some visuals and statistics on Trump’s absolute dominance of Twitter.
Not only does NodeXL back up Trendsmap’s overall picture, it becomes clear from the different tests NodeXL allows you to run that Trump’s power on Twitter is far greater than is apparent from Trendsmap’s visuals. To illustrate this visually, here is a full-page image of the network:
And here is what it looks like when you click on the node @realdonaldtrump, which sits at the center of group 1 (dark blue):
Note further that the rest of group 1 forms essentially two rings, one which is retweets, replies, and quotes, and the other which is responding to the first ring.
When we look at the metrics, Trump crushes nearly every category:
This can be parsed further when we look at In-Degree (Out-Degree doesn’t matter, as Trump doesn’t reply to people, but people reply to him):
Betweenness Centrality, a much more potent measure of influence than Closeness Centrality:
Eigenvector Centrality and Page Rank:
So, depressing as this may be, Twitter is dominated by Donald Trump, and the opposition voices that engage with him do next to nothing to counter his influence–in fact they actually boost it. Most of the opposition, from what I can see in these tweets, has no immediate, specific message other than “he’s awful/unqualified/terrible/embarrassment/ridiculous” or some combination thereof. Additionally, if you look at the “Top Retweets” above, most have links of some kind, which require further reading from interested parties. Trump doesn’t have to provide links, and that is an adavantage–if you wanted to be snarky, you could say that to be in opposition to Trump requires more thought…At the very least, I’ve seen many academics acknowlege the limitations of 140 characters while tweeting away happily. Trump is just fine with these 140 characters, because they do deliver a message immediately, with no need for further explanation or links that would require prolonged reading or engagement. And by doing so, he maintains the image of talking directly to the people of the U.S. and the world (though much of the world isn’t on Twitter, or not Tweeting about the U.S.). Basically, as Amanda Hess opined back in February 2016, Trump is actually really good at Twitter, from a rhetorical and political standpoint. Given his reach, the quality of the positions advocated is largely moot–“quantity has a quality all its own” (almost certainly not said by Stalin).
All that being said, however, the counter-narrative is out there, and seems robust. As yet, however, it has not coalesced.
Plus: This evening, the hashtag #BLMKidnapping began trending. I have not had a chance to run a NodeXL analysis of it, but it would probably strain my computer to calculate. As of 12:05 a.m. on January 5, Trendsmap recorded the following:
With this heat signature map:
With these Top Retweets and Most Quoted tweets:
With these Most Replied and Most Influencing Tweets:
I didn’t watch the entire video, so maybe I missed where BLM is invoked. In any case, it seems the far-right have taken over the Twitterverse on this one. I’ve found Nina Turner to be invaluable for good counter-narrative on this horrible incident, and Shaun King later this morning should be as well.