Is Trump's Campaign Locking Him Out of Twitter?

People love @RealDonaldTrump as much as they love the real Donald Trump. Just like the candidate himself, the Twitter account is entertaining, coarse, rude, doting, blunt, fiery, cloying, politically incorrect, and superbly meme-ready. More importantly, it’s real.  You get the sense that each tweet shoots straight from Trump’s amygdala, typed furiously, rat-a-tat-tat—sad! It’s such a refreshing change from the stale fare you get everywhere else on campaign Twitter, where politicians literally sign their tweets to let their constituents know that, yep, it’s really them, not the social-media staffer, like it usually is.

That’s why I loved this text analysis of Trump’s tweets by David Robinson, a data scientist at the coding site StackOverflow. Robinson tests a well-circulated hypothesis: If a @RealDonaldTrump tweet is marked as being sent from an iPhone, it’s from a staffer. But if it’s sent from an Android phone—Trump’s Samsung Galaxy?—it might be from the candidate himself. Folks have noted that the Android tweets are a bit, uh, Trumpier:

Every non-hyperbolic tweet is from iPhone (his staff).

Every hyperbolic tweet is from Android (from him). pic.twitter.com/GWr6D8h5ed

— Todd Vaziri (@tvaziri) August 6, 2016

Robinson took this theory and tested it empirically, downloading more than 1,000 of Trump’s tweets and running them through a language parser. Even with a limited sample, the results were pretty conclusive:

My analysis, shown below, concludes that the Android and iPhone tweets are clearly from different people, posting during different times of day and using hashtags, links, and retweets in distinct ways. What’s more, we can see that the Android tweets are angrier and more negative, while the iPhone tweets tend to be benign announcements and pictures.

Tweets from the Android phone tend to come earlier in the day, while the iPhone’s pace picks up in the afternoon and the early evening. The Android account is also a bit of a grandpa when it comes to retweeting people, copying their tweets into its own and surrounding them in quotes, which the savvier iPhone account doesn’t do. The iPhone account loves to tweet pictures; the Android almost never does.

Most juicily, the Android phone—oh, let’s just call it “Trump”—is much, much angrier. By Robinson’s calculations, it uses 40-80 percent more negative words in tweets than the iPhone, including Trump mainstays like “badly,” “crazy,” and “weak.” Tweets from that device frequently evoke sadness, fear, anger, and disgust, according to Robinson’s language processing, while the iPhone is more likely to express anticipation, trust and joy.

Plus, the iPhone tweets out event times. That’s a staffer job. C’mon.

So I got curious. If tweets from the Android phone represent Trump’s contributions to his Twitter account, and the iPhone signals his staff’s input, how much is the New York billionaire actually tweeting these days? As the stakes grow higher—and flubs more damaging—has Trump’s campaign locked him out of his account for his own safety?

To figure this out, I needed to grab as many of Trump’s tweets as I could, preferably from the start of his campaign. Turns out this is pretty hard: Twitter clamps down on your ability to download tweets programmatically after 3,000 or so, which only took me back to December. I gave the social media company a call, but was told it would probably cost thousands to extract the data I needed.

So I turned to Oren Tsur, a visiting scholar at Harvard and associate professor at Ben Gurion University in Israel. He’s written about Trump’s Twitter account before and has been scraping the candidate’s tweets since 2015. With his help, I scanned his archive and matched it with more current data—resulting in some 6,000 tweets—and counted the number of times @RealDonaldTrump tweeted from each device, grouping by month. If you convert each month to a percentage, you get this:

If the Trump-Android proxy holds true, the Republican nominee’s participation in his own Twitter account has dropped sharply over the past year. From July 2015 to October, Trump largely ruled the account, with up to three-quarters of @RealDonaldTrump tweets coming from the Android phone. The staff takeover began in November, growing to around 50 percent of tweets before plateauing through the early primary months. But the last month has seen another major shakeup. So far in August, almost two-thirds of his tweets have been sent from an iPhone, suggesting a serious crackdown by the campaign staff.

(Note: A number of readers have asked whether the Android phone is actually tweeting less, or if the iPhone is just tweeting more, diluting Trump’s share. Perhaps Trump is just being out-talked by an effective campaign staff? Answer: Trump is indeed sending fewer tweets from the Android device. He posted an average of 371 tweets per month from July through December; that average is down to 200 per month in 2016, not counting August. The iPhone’s tweeting has also ticked up substantially.)

Granted, it’s still early in the month. But in the past two weeks , Trump has already picked a fight with a Gold Star mother, repeatedly misled supporters about an Iranian money transfer, and problematically joked that the “Second Amendment people” could stop Clinton before she appoints new justices. With that in mind, it’s hard to view this uptick as anything other than campaign manager Paul Manafort firmly taking the reins and changing the account password.

That is, if you accept my premise at all. This analysis comes with caveats, and it’s worth noting some inconsistencies between Robinson’s analysis and other published literature on Trump’s tweeting habits.

• Trump has said he dictates tweets to staffers. “During the day, I’m in the office, I just shout it out to one of the young ladies,” he said in a CNN town hall. “I’ll just shout it out, and they’ll do it.” So some of the iPhone tweets might actually be Trump’s own words, even if he didn’t type them out.
• Trump says he does his own tweeting after 7 p.m. But Robinson’s analysis shows the iPhone tweets spike at around 8 p.m., well past Trump’s supposed deadline. By my knowledge, you can’t schedule tweets with an iPhone, so that means a staffer is manually tapping them out. That said, the Android tweets typically pick up a fair amount around 9 p.m. And I’d bet those 8 p.m. tweets often relate to rallies, when Trump is onstage and can’t handle his phone.
• It’s entirely possible Trump has handed off his Twitter account entirely. Robinson’s analysis suggested the users of the two phones had distinctly different styles. But that doesn’t mean Trump is one of them. The Android phone could be in the hands of a trusted aide—perhaps his anger translator?

Please understand that I recognize the ridiculousness of all this. I've expended a great deal of time and coding effort—enlisting a literal Harvard scholar—to figure out how often a presidential campaign tweets on an iPhone. But that’s a testament to the central role Twitter plays in Trump's campaign. In a recent interview with the Washington Post reporter Philip Rucker, Trump said he didn’t need the media, because he has mastery of social media. “I’m head and shoulders above everybody else,” he said. “I’ve read now 22 million people on Twitter, Facebook and Instagram. More than 22 million people. Nobody else is even close.” He’s right. And when a candidate has the biggest megaphone around, it’d be nice to know if he’s actually using it.