By: Patrick Zimmerman

Trump’s Twitter account is not a great window into his policies; his word is far too fickle and disconnected from the executive actions he takes.  However, it can be a great way to monitor the state of mind of the President of the United States.  And, you know what?  @realDonaldTrump’s Twitter habit is something that we’ve spent some time quantifying and cataloging.  So that’s convenient.  For this project, we’ll look at two metrics as proxies for Trump’s level of belligerence (or confidence) and panic: his rate of diplomatic insults and of rage tweets. 

The question

Is there a pattern to Trump’s rage tweets or insults, and can that be matched up with external stimuli on the President, such as the Müller investigation, the progress of legislation (or lack of it), or turnover in his administration?

The short-short version

The initial hypothesis about a relationship (even an inverse one) between insults and rage tweets turned out to be a big nope.  However, in terms of anxiety, it sure looks like his rage tweet rate ebbs and flows with his level of angst about the Russia investigation.

Defining rage tweets & insults

For this project, we’re making the following assumptions:

• Most of the time he overtly insults a country or sends out a rage tweet (see definition below), it is Trump himself tweeting, not a staffer.
• His rage tweets are, in fact, indicative of a lack of control and not some grand strategy. That is to say, Trump is what he appears to be and not a Keyser Söze-esque master of disguise.

Diplomatic insults are pretty self explanatory; any time Trump insults a country either overtly or implicitly.  For example, if he complains about immigrants of uncertain character streaming over “our southern border,” there is no other way to read this than as an insult to Mexico (the Mexicans sure take it that way).

Rage tweets are evidence that the president has temporarily lost control over his public voice.  They come in two types:

all-caps shouting…

and excessively long rants.

The results

We’ve mapped Trump’s 2-week rolling averages for his insults and rage tweets, both expressed as a percentage of his total tweets, as well as the 2-week rolling averages of potential stimuli, the rate of revelations about Russian interference, the departure of senior staff from his administration, and blockages to the execution of his agenda (primarily through the courts and through legislative defeats).

Note: each axis on the below dashboards is normalized relative to the max value of each series.  This doesn’t affect any of the correlation calculations, but makes the curves much easier to see, and we’re looking for trends here.

What is clear is that, if our hypothesis about rage tweets reflecting Trump’s level of anxiety is correct, then Trump had a heightened level of anxiety from January through May 2017, then things calmed down during the latter half of 2017.  2018 has seen a marked rise in Trump’s anxiety level.

Mouseover for details.

Take home: Russia is the only outside stimulus that fits that pattern, with a correlation of 0.480.  Most importantly, it’s the only story that could be vexing the President that pretty much went away for a while during 2017. Everything else lacks that dip in frequency.

Mouseover for details.

Trump’s insults? Nope.  They pretty much have been constant throughout Trump’s tenure.  Correlation of 0.032.  Aka diddly/squat.

Mouseover for details.

Nothing here, either.  Major departures have plagued this administration, and they’ve figured in some of Trump’s rage tweets.  However, they don’t seem to be the driving force behind his anxiety-induced rants and screams, and there are major periods of turnover that barely provoked a blip on the rage-o-meter.  Correlation of 0.783.

Mouseover for details.

Trump has also had a series of frustrations in enacting his agenda (though many of these were subsequently overcome, we’re interested in his reaction in the moment). Did, say, the blockages on his travel bans in the courts, the failure of the Obamacare repeal, or the government shutdown over his budget cause an uptick in Trump’s Twitter stress?  Nope.  Correlation: 0.094.

To sum up, it sure looks like the Big Bad Special Prosecutor is making @realDonaldTrump sad.

Adjusting for the Twitter expansion

One important note on the tweet rates: due to the November 7 2017 expansion of Twitter’s character limit from 140 characters to 280, we changed the way we defined rage tweets, to keep things consistent.  Our original criteria was 3 140-character tweets, or 420 characters, which worked out to 3.04x the median length of all the The Donald’s non-rage tweets (138 characters). For tweets made after November 7 2017 doubling of the character limit, we scaled our definition of a rant appropriately, to anything over 681 characters (the median length of @realDonaldTrump’s non-rage tweets after the change is 227).

Mouseover for details.

If we hadn’t adjusted accordingly, Trump would have been credited with a significantly higher number of rant-style rage tweets to this point (57 v. 34), which would have overstated our argument.  All-caps tweets were not affected.

Also, counted as normal (rather than rage) tweets that are clearly intended to be read as positive, boastful, or obviously not rage-induced, even though they would otherwise meet the shouting or rant criteria, such as:

What’s next?

Well, the big question: is this trend predictive of behavior towards the Russia investigation?  Will he try to fire Rosenstein or Müller or otherwise hamstring the investigation even more than he already has (such as with the appointment of a second Special Counsel to investigate the investigation that led to the first Special Counsel)?  Is this trend going to continue, or will Trump calm down as he did in the months after Comey’s testimony and the establishment of the Special Counsel?

There’s likely some upper limit to the rage tweet rate, but there’s no way to really know if we’ve approached that or not.  As in, there is a world where 100% of his tweets are all-cap rants, but one imagines that he would have been removed from office before he would reach a state that apoplectic.