Trump’s in a Twitter Funk

Hillary is Beating Donald at His Own Game.

What if I told you everything we’ve come to believe about Donald Trump’s Twitter feed is wrong? Okay, that would just be click-bait. But let’s say I told you that over the past few months @HillaryClinton has:

  1. Posted more often than Trump, which is a higher twitter frequency
  2. Reached more users, resulting in more potential impressions
  3. Had bigger moments, earning more retweet echoes
  4. Gained more followers, expanding the campaign’s audience

None of this makes intuitive sense based on what we know of the 2016 campaign for President of the United States. Our social media research at George Washington University’s Graduate School of Political Management’s PEORIA Project has consistently shown that Trump has dominated the platform, which has helped him overtake his primary opponents and close the gap in polling with Clinton.

As it turns out, however, all four are true.


Higher Tweet Frequency

We define tweet frequency as the total number of tweets originating from a specific Twitter account. If one campaign has higher frequency, that campaign is posting more often. In terms of traditional campaign media, this translates to a number of free ads on the platform for followers of the campaigns.

Scrolling through and transcribing tweets by hand is extremely time consuming and eventually you reach a point where a page doesn’t scroll back any farther and/or your eyes fail you. On its website, Twitter tells you to stop with a research limiting, yet deeply merciful, “Back to top” message at the bottom. This was the end of the line for me.

Thankfully, transcription isn’t the only way to do this. Twitter provides developers to access a maximum of 3,200 statuses through its API. Deployed with a fairly straightforward Python script, you can download a wealth of information to a .csv file about the tweets including when it was created, the URL (which includes how it was posted), the number of retweets, number of times it was favorited, and the exact text of the tweet.

Sample output of the Python script to download the most recent 3,200 tweets for @realDonaldTrump

Next, the task is to make some sense of the data. First a note about time and date: it will be set to GMT so you’ll need to adjust the data to compensate. I’m in Washington, DC so I needed to add five hours. For this analysis by date, I formatted the cells in Excel so I would just see date, hiding the time stamp. Then I sorted the dates in reverse order and saw the surprise: Clinton’s oldest tweet of the 3,200 downloaded was on February 20, 2016. Trump’s oldest tweet was on December 13, 2015.

At first I thought this was a mistake so I emailed the author of the script David Robinson, a brilliant data scientist for Stack Overflow. Obviously, Trump would out-tweet Clinton, even if it was her staff that was doing most of the account’s work. He patiently suggested that I double-check my assumptions.

So I went back and eyeballed the timelines on the Twitter website for as long as they would allow and it was true: @HillaryClinton had out-tweeted @realDonaldTrump during this period. Here’s a graph showing the difference between the accounts.

The data shows Trump posted 2,104 times from February 20 to August 15, 2016. Clinton, who is so dubious of public engagement that she rarely holds press conferences and had a private email server to protect or hide her emails from prying eyes (your choice on the characterization), tweeted the full 3,200 times during the same period. Clinton out-tweeted Trump by more than half, or 1,081 posts.

As an aside, let’s look at who is doing the tweeting. In contrast to Clinton, whose tweets are noted with a “-H” to show she tweeted it herself, Trump’s have no such designation. According David’s analysis, almost all of Trump’s tweets originate from either the Twitter app on an Android or iPhone.

But his main finding was more interesting: based on his sentiment analysis of @realDonaldTrump, the tweets originating from the Android device were angrier and had a more negative tone while the ones posted from an iPhone were more neutral or positive. As The Verge noted, Trump tweets from a Samsung Galaxy.

In our August 16 download of the most recent tweets from @realDonaldTrump, 1,641 were from Twitter for Android while 1,320 were from Twitter for iPhone. The remaining 234 were from the Twitter Web Client, two were re-posts from Instagram, two were from Twitter for iPad, and one was from Mobile Web.

According to our data, most of the @HillaryClinton tweets originate from TweetDeck (2,497), Twitter for Web (601), or Twitter for iPhone (102). We’ll save our impressions of that account’s voice for later but it’s obvious that the vast minority of the tweets (124 out of 3,200) are signed “-H” for Hillary.


More Potential Impressions

Our partners at Crimson Hexagon define the metric potential impressions as the number of times that a particular post could have been seen. Potential impression is calculated by adding the sender’s followers and the followers of all users who have retweeted the post. In campaign terms, this is the potential reach of any tweet.

Let’s narrow the lens to look at the impact of these tweets over the past month, which corresponds to the night before the Republican National Convention. Over this period, Clinton earned about 11 billion total potential impressions from her tweets. Trump earned about 9.9 billion. Not only is Clinton outspending Trump to gain potential impressions on traditional media, she’s has also been more active on Twitter, gaining her more access to users on that platform.

One can argue that Trump is getting more bang for his tweet as Team Clinton is spending more time tweeting to gain this edge on Twitter. If these were actual dollars being spent on traditional media Trump’s message would be more efficient. But this is Twitter and the marginal cost of an additional tweet does not cost the campaign more money. Clinton’s strategy of tweeting more often results in a greater reach to Twitter users.


Earning More Retweet Echoes

Twitter offers its users a simple way to replicate another user’s tweet on their own feed: the retweet. The original tweet is then passed along to all of the retweeter’s followers, amplifying the potential audience for the tweet exactly as it was intended. We view these in campaign terms as the perfect message echo.

Last month in Unclosed Deals, we documented how Team Clinton’s “Delete your account” tweet was the most powerful moment of the 2016 campaign season on Twitter with 413,667 retweets. Trump tried to respond with “How long did it take your staff of 823 people to think that up — and where are your 33,000 emails that you deleted?” but that was retweeted less than half the number of times (200,266).

Here’s our chart from the deck.

The next moment when both candidates were retweeted was during and right after the Democratic National Convention. Clinton earned 283,974 retweets from a blizzard of 103 tweets on the last day of the convention, July 28. The next day, Trump got 268,481 retweets over a relative sprinkling of 23 separate tweets. While Trump’s were more efficient, Clinton earned more retweets, which we define at PEORIA as perfect campaign message echoes. Again, retweets ≠ endorsements but they do amplify the messages.


Expanding Campaign Audience

As of this post, Trump remains the undisputed leader in Twitter followers with 10,937,072 compared with Clinton, who has 8,328,931. Remember, our previous reporting demonstrated how Trump’s Twitter account grew at a much faster rate than Clinton’s, overtaking her in mid-October of last year. More recently, we showed that Trump’s growth on the platform was leveling off to equal Clinton’s growth.

Now, we can report that @HillaryClinton is growing faster than @realDonaldTrump, albeit slightly. According to our partners at Crimson Hexagon, over the past month Clinton is up 13% while Trump is up 12%. Even though it is a 1% difference — and impossible to eyeball from the graph below — the fact that Clinton’s account is growing faster than Trump’s at this stage of the campaign is news. However, it’s tempered by the fact that in real numbers of followers, Trump gained 221,013 more than Clinton.


While there’s not enough time for Clinton to close the gap in followers, her team’s use of the platform has made it significantly more valuable than we might have expected. In the current phase of the campaign, Team Clinton is engaging people more often on Twitter, reaching more users, effectively leveraging moments, and growing the @HillaryClinton account faster vs. @realDonaldTrump.

The implications of these findings are instructive. As we found in our earlier research on Twitter metrics, Trump’s campaign on Twitter is reflective and at times predictive of where the campaign is going. Trump’s growth on the platform predicted his rise to the nomination and now his Twitter funk appears to be explaining his general election fall. One more thing for his (second) new team to fix.


If you enjoyed this article, click the💚 below so other people will see this here on Medium. You can follow our research on our website or on Twitter @PEORIAProject, which is funded by a generous grant from Mark R. Shenkman. To learn more about the Graduate School of Political Management visit our website or follow us on Twitter @GSPMgwu.

For more from and about David Robinson, data scientist for Stack Overflow, please visit his website. Speaking of researchers with common names, I’m definitely not this guy. You can follow me on Twitter @michaelcohen.

A single golf clap? Or a long standing ovation?

By clapping more or less, you can signal to us which stories really stand out.