Skip to main content

The Multiple Voices of Donald Trump

@POTUS and @realDonaldTrump -- the voice of the Trump administration and the voice of Trump himself. Is their a difference between the two? And, if so, what characterizes that difference?

The first question can be answered, quite easily, in the affirmative. The answer to the second question is, as the results from an analysis of @POTUS vs. @realDonaldTrump Tweets reveals, dependent on how one defines "difference."

The Data
The results found via the analysis discussed below are based off of a growing dataset composed of @POTUS and @realDonaldTrump Tweets. The earliest Tweets in the data date to February 16, 2017 with the most recent dated to March 6, 2017. [Note: the program I've been using to collect Tweets reports the time per Tweet several hours in advance of eastern standard time, so the most recent Tweet in the dataset is technically dated to March 7.]

The Analysis
I used several methods of text analysis to compare @POTUS to @realDonaldTrump Tweets. These include sentiment analysis using the NRC Emotion Lexicon (a growing favorite of mine) and term frequency - inverse document frequency (tf-idf). The former allows for inference of positive vs. negative sentiment in text, while the latter allows for the identification of top words per a given corpus of data vs. another corpus, adjusted for the word's uniqueness. Both of these methods permit making an assessment of quantifiable similarities and differences between @POTUS (official Twitter account of the Trump Administration) and @realDonaldTrump (the personal account of Donald Trump) Tweets.

Findings
Tweets over Time
Over the timeframe included in the analysis, there is, aside from a clear spike in @POTUS Tweets around the time of Trump's address to a joint session of Congress, a seemingly random distribution of Tweets from February 16 to now.



Also, it appears that, while there are slightly more Tweets from @POTUS (97), @realDonaldTrump is active daily, while @POTUS is occasionally inactive during certain days.


Sentiment Analysis
Sentiment valence (the positive vs. negative sentiment score per Tweet) ranged from as high as 5 to as low as -3, with the mean sentiment hovering just slightly above zero at 0.51.

The below figure displays the sentiment per Tweet by date and by whether it was Tweeted by @POTUS (blue) or @realDonaldTrump (red).



A quick glance at the above figure doesn't suggest an overt difference in average sentiment per account. And, in fact, a Welch two-sample T-test confirms that there is not a statistically significant difference in mean sentiment per @POTUS vs. @realDonaldTrump. There is a slightly higher average sentiment per @POTUS Tweets, but, at this point, it's not certain whether this difference is a meaningful one.


tf-idf Words
While the above analysis fails to find much of a difference in positive vs. negative sentiment between Twitter accounts, tf-idf reveals some telling differences in word usage per account. The below figure displays the top tf-idf words per @POTUS Tweet.



Meanwhile, the below figure displays top tf-idf words per @realDonaldTrump Tweet.



While it's still early in this analysis to make any hard inferences regarding differences in @POTUS vs. @realDonaldTrump Tweets, what seems most clear, at least at face value, is that the latter appears to uniquely focus on (for lack of a better phrase) palace intrigue ("russia," "leaking," "classified") and criticism ("obama," "failing," "nytimes"). Meanwhile, the former appears to focus more on patriotism ("americanspirit"), mentions of individuals in the administration, policy ("repealandreplace"), and promotion ("cpac2017"). Thus, while one face of the Trump Administration is portrayed on social media via @POTUS, @realDonaldTrump continues to serve as the unique online presence of Donald Trump, apart from the administration.

Why Care?
Differences in the Twitter activity of @POTUS and @realDonaldTrump reveal a lot about divergences in voice between Trump himself and that of the Trump Administration (staffers, etc.). This difference matters because, while 15.7 million people currently follow @POTUS, 25.7 million follow @realDonaldTrump. Whichever account people take more seriously as the true representation of the administration, its positions, its priorities, its goals, and its trajectory matters. Whether we focus on palace intrigue or substantive policy discussion may depend upon which voice (@POTUS or @realDonaldTrump) about which we are more concerned.

Comments

Popular posts from this blog

A Network Analysis of Foreign Aid Commitments

International Relations scholars often talk about the "diffusion" of norms, behaviors, security worries, etc. throughout the international system. Foreign aid policy is one such norm -- one that developed, democratic countries often are peer-pressured into sharing. But which countries lead the way in terms of aid commitments? Why Network Analysis? The study of networks in the social sciences has largely been restricted to sociology; however, more recently, other fields such as political science (international relations in particular) have adopted network science as a tool in the study of social phenomena. Networks provide a visually intuitive graphical representation of the multiple connections among numerous actors. Aside from being a visually appealing representation of a network of relationships, network analysis of the international system helps to bring to light (and also account for) the fact that international politics is inherently multilateral . Most analyses in...

A Sentiment Analysis of President Trump's Inaugural Address

The day has come and gone on this 20th of January in 2017. Donald Trump has been sworn in as the 45th President of the United States, and like his predecessors, Trump gave an inaugural address during today's ceremony. Lots of pundits, of course, have spent today closely analyzing, scrutinizing, praising, and whatever-else-ing Trump's speech. Like everyone of these pundits, I have personal thoughts about the events that have transpired today as well, but, rather than simply contribute to the discordant  cacophony of punditry a contrived opinion of my own, I'm choosing  to do something more constructive (or at least less politically volatile ) in an effort to offer something of novel import to discussion about President Trump's address. Rather than qualitatively sift through every word and nuance of Trump's inaugural address, I'll do something political science-y: use some textual analysis techniques to leverage quantifiable insights from Trump's speech. The...

What Factors Explain Variation in Median Household Income (Part II)

As promised, I've completed further analysis of the impact of various factors on median household income at the metropolitan area level, the results of which are displayed and discussed below. Rather than focus on median household income in general, I've gone a step further with this analysis by looking at variation in median household income by income tier -- e.g., lower income, middle income, and upper income tiers. The results are interesting (at least I think so...), and they arguably have significance for public policy. Also, if you don't care for all the glorious (ok...boring) details in the following analysis, just scroll on down to the conclusion. In my previous post I alluded to what I viewed as a misguided hyper-focus on security, crime, and undocumented immigration in this current presidential election season (most notably because of the rhetoric of the Republican nominee) that has distracted from other important social and economic issues. Now, I certainly thi...