Thread by @compstorylab

Thread for a new paper of ours on the arXiv:

“Ratioing the President: An exploration of public engagement with Obama and Trump on Twitter”

arxiv.org/abs/2006.03526

J. R. Minot, M. V. Arnold, T. Alshaabi, C. M. Danforth, P. S. Dodds

We explore the dynamics of how Twitter users have responded to tweets made by Obama and Trump from their main accounts, @BarackObama and @realDonaldTrump.

For each tweet, we track three main characteristics as they evolve over time:

- Number of Favorites
- Number of Retweets
- Number of Replies (hard to measure—see our paper)

The three characteristics are different in nature.

- Favorites are light and easy to make.

- Retweets require more effort as the user is broadcasting to the their followers.

- Replies involve some creative action, however little.

We’re interested especially in levels of being ratioed, a concept that emerged towards the end of 2017:

urbandictionary.com/define.php?ter…

(This Urban Dictionary definition for “ratioed” has not been ratioed.)

For these three measures, ternary plots make for powerful visualizations.

In October of 2017 (approximately three thousand years ago), @FiveThirtyEight put out an analysis of US political figures using these kinds of plots:

fivethirtyeight.com/features/the-w…

Tweets are located according to the relative balance of
# replies : # retweets: # favorites.

Top: 0:100:0 (sharing something important and serious)

Bottom left corner: 100:0:0 (users have something to say)

Bottom right corner: 0:0:100 (festival of love)

Generally, if replies outweigh retweets and/or favorites, a tweet is getting ratioed.

Which means that the bottom left corner of these ternary plots is not where you want to be (unless that was the plan).

A recently ratioed tweet:

https://twitter.com/realDonaldTrump/status/1267227396341669889?s=20

Note: We have never included tweets in our papers before, for both privacy and for the misprepresentation that a single tweet might convey. However, for this paper we do show POTUS tweets as they are considered part of the official record.

npr.org/2019/10/25/772…

Celebrities tend to live in the massively favorited zone of the bottom right corner.

A typical response to a @bts_twt tweet:

https://twitter.com/BTS_twt/status/1264758066546798592?s=20

Our paper.

Here’s an example of how the # retweets to # replies and their ratio behaved over time for a specific Trump tweet.

After some initial transients in the ratio during the first few minutes, the ratio settles down after about a day.

We also track how the response to a tweet moves around the ternary plot over time:

We create time series for the numbers of the three characteristics, using a histogram structure because there are just so many tweets.

Obama and Trump:

Ratio histograms for three time periods based around the 2016 election show strong changes.

- The increasing gravity of the favorites count point to it being less informative.

- The bottom time series show the key ratio of # retweets to # replies.

Leaving favorites aside, here’s a comparison of Obama and Trump for the # retweets / # replies ratio.

Obama’s tweets rarely cross the line marking more replies than retweets (dashed line, <1%).

Around 18% of Trump’s tweets find themselves in ratio territory.

And that’s the thread.

Our paper represents a start on POTUSometrics and ratiometrics.

We have more to come in this area and much more to come in general from our team at @compstorylab.

Congrats to Josh Minot on his first first author paper making it’s way into the world.