Donald Trump once said in a tweet “I have never seen a thin person drinking Diet Coke.” Now that he is President of the United States, we expect him to act differently on Twitter. However, although he may not be as much of a diet-coke hater he certainly has his bad days (don’t we all?). This post seeks to answer the popular questions about Trump’s birdy-pulpit, such as:
In seeking to answer these questions, I hope we can gain a look into the otherwise private mind of our 45th President, focusing specifically on the content and emotion of his tweets since January 20th, 2017.
To wrap our minds around the analysis, let’s get an idea of the polar opposites of Trump’s tweets. Starting with his most positive, then looking at the most negative. We measure whether or not a tweet is positive using a tool called sentiment analysis, which simply looks at the emotion behind the words and sentences being used in Trump’s tweets.
To get these words, I simply download all of the tweets that have been sent from the @RealDonaldTrump twitter account since 1/20/2017. Going a step further, I also ensure I’m only looking through tweets that have been sent by Trump’s personal device. I will discuss this later.
What an amazing comeback and win by the Patriots. Tom Brady, Bob Kraft and Coach B are total winners. Wow! Donald J. Trump
Had a great meeting at CIA Headquarters yesterday, packed house, paid great respect to Wall, long standing ovations, amazing people. WIN! Donald J. Trump
Hope you like my nomination of Judge Neil Gorsuch for the United States Supreme Court. He is a good and brilliant man, respected by all. Donald J. Trump
The fake news media is going crazy with their conspiracy theories and blind hatred. @MSNBC & @CNN are unwatchable. @foxandfriends is great! Donald J. Trump
The FAKE NEWS media (failing @nytimes, @NBCNews, @ABC, @CBS, @CNN) is not my enemy, it is the enemy of the American People! Donald J. Trump
While on FAKE NEWS @CNN, Bernie Sanders was cut off for using the term fake news to describe the network. They said technical difficulties! Donald J. Trump
In the former category we see that Trump is quite gracious with praise, although in two of the three most positive tweets Trump is essentially praising himself. In the negative category, we start to see a trend emerge: when Trump is tweeting about the media, he is tweeting about the quite negatively. It’s also notable that Trump’s three most negative tweets are about the press. If you’re familiar with his feed — or the president at all, really — this should not surprise you.
Still, it’s useful for us to gauge just how much President Trump is tweeting about the “fake news” press.
From this graph of Trump’s frequently used words we see that he does in fact talk about the press (usually the New York Times) more than almost any other subject. That’s not too surprising all things considered but is definitely an important discovery.
You will notice that Trump also talks about the country quite a bit, using words like “nation,” “people” and “America.” This is a little hint at the nationalist rhetoric found in Trump’s first joint address to Congress that I wrote about last week.
Now that we have an idea of what Trump is tweeting, it would be fun to explore when he is tweeting.
It’s often the case that the president is a very busy man. Hell, Lyndon B Johnson used to hold conversations with people through the bathroom door. In theory, that does not leave much time in the day for Trump to spout off online. Or does it? Here, we look at the times and days Trump is more likely to tweet.
We see what we would expect to see; Trump tweets a lot in the morning. Specifically, POTUS has tweeted 100 times between 5 and 9:00 AM since January 20th. Presumably, the Commander in Chief has some extra time in the morning.
You will also note that the bars around 6:00 AM are darker than some of the other bars in the graph; that symbolizes that Trump tweets more negatively in those morning hours than other hours, though his most negative hour of tweeting is during most Americans’ lunch hour, at noon.
Trump also spends a lot of time tweeting on the weekend, at which time he is also the most negatively.
On the bright side, Trump is most positive on Mondays. Cheers to you, Mr. President; we aren’t all that jolly at the start of the week.
What we are arguably most interested in is how emotive Trump is. In other words, are his tweets more negative than positive? More angry than joyous? More sad than happy?
As it turns out, all of these are true. But allow me a few sentences to explain this process, and justify the decision above to only use tweets sent by Trump’s personal android phone.
Android vs. iPhone
It was hypothesized in the late stages of the 2016 campaign that tweets sent from Android phones were often the personal tweets of President Trump. On the other hand, tweets from iPhones were those sent by the campaign. This was corroborated by David Robinson last October when he analyzed the statistical differences in usage of pictures, links, and (you guessed it!) emotions by Trump and his campaign.
In exploring whether or not this was still the case, we stumble across some interesting results.
Firstly, here is a look at the different words used in tweets sent from the president’s personal Android phone and those from staff iPhones. These are only tweets sent after the inauguration of POTUS Trump.
There is a very clear difference between the words Trump uses and those of his staff. In the above graph, that difference is measured simply by the log ratio of words mentioned in Android-phone tweets over those used in iPhone tweets. That’s just statistics talk for “I tell my computer to look at words used more often in Android-phone tweets.” We see those on the right half of the graph, with iPhone tweets on the left.
From Trump, we get words like “failing,” “mess,” and “fake” and mentions of various media networks. From the iPhone, AKA the staff account, we get words that look to be from mostly promotional tweets — like Trump’s appearances on Morning Joe or his press conferences.
You’ll notice that the biggest difference is from tweets where Trump is still hung up about the 2016 election.
So, if Trump’s words are different than those of his staff, are his emotions distinct as well?
You betcha. Here is a similar graph that shows Trump’s propensity to tweet with different emotions that his current White House twitter intern.
Above, each emotion we look at has its own panel of words associated with that emotion. In each panel, words that are orange are words from President Trump, and red words are from his staff.
Immediately it’s clear that the words with negative emotions (anger, fear, sadness, etc.) are almost completely unique to tweets from POTUS’ Android. Joyous words are split almost 50-50, and words that signify Trust are used much more frequently by Trump’s staff.
Here’s a simpler version of that graph, where emotions with bars extending to the right half of the graph are sentiments present in Trump’s tweets that are not in his staff’s.
Again, it’s abundantly clear that Trump is the more negative person in the white house. In fact, Trump is 400% more likely to tweet sad tweets than his staff, 200% more likely to send angry, negative, or disgusting messages, and 150% as likely to tweet fearful words.
These quick comparisons of word usage and emotional valence confirm that the tweets in Trump’s timeline that are sent from Android phones are tweets sent by the President himself, whereas tweets from iPhones are often tweets from his staff.
Since Trump is very different from his iPhone crew, it only makes sense to look at the tweets from his Android in analyzing his word choice and emotions. Here, we look at the latter component, as I’ve already reviewed the former.
First, let’s look at the percent of Trump’s tweets that correspond with a certain emotion.
Roughly 13% of Trump’s tweets are negative, with anywhere from 5 to 10% containing emotions of disgust, anticipation, anger, sadness, fear, or distrust.
Only 4% of his tweets are joyous (happy).
Although this does not have any immediate implications for our analysis, the frequency of certain emotions in his tweets may indicate the frequency with which he has those emotions in real life. Of course, we cannot be sure, but it makes sense that Trump’s tweets are reflections of his character. If they are, our commander in chief is not the calm (ahem, “presidential”) man he and his staff would like us to believe. Indeed, the opposite is probably true.
Finally, it’s appropriate for us to look at Trump’s attitude over time. The graph below goes through all of the steps we went over in the last section, recording just the positive and negative sentiments for Trump’s tweets on any given day. Through this, we get a gain insight into how Trump’s emotions have evolved since election day, and even a little before then.
Peculiarly, the otherwise “happiest” days for POTUS look to be his worst. After each of his biggest events — the election and inauguration — Trump has taken an immediate and sustained dive into negative rhetoric. Although he has “so far” always come out of those negative swings, he has also “so far” always progressed to hit new lows after the next big event.
Here, we also get a good look at the difference in sentiment between Trump and his staff that we discussed above.
It’s not clear where Trump’s heart will turn in the future, but if the past is any indication he is set to take a sustained downturn in the next couple of days.
In the end, here are the answers to those questions I initially posed to you readers.
In this post, I have shown that Trump is predisposed to tweeting messages that are much more negative than that of his staff. Indeed, the tweets sent from his Android device seem to indicate that he is often rather emotionally distressed.
Like I said, the future is ultimately hard to predict, but Trump will almost certainly continue tweeting about “FAKE NEWS” and the New York Times while his staff covers tweets about policy, press conference, and otherwise “normal” events that transpire under any presidency.
Future research could compare Trump’s proclivity to negatively valenced emotions with that of other political leaders. Perhaps his compatriots in the Senate will be of some help to our understanding of just what is is that makes Trump tick — or, rather, what makes him tweet.