Research & Insights

By Justin Tenuto, August 11, 2015

Who won the GOP debate?

If you read much news after the Republican debate in Cleveland last Thursday, chances are you were slightly frightened that reputable outlets like CNN were all agog over Donald Trump, a man with complicated hair who has a habit of saying bad and gross things pretty much every time he opens his gob. But it wasn’t his performance, per se, that had news outlets excited; it was his presence.

We mean this: Trump gets attention. The GOP Debate in Ohio had better ratings than the NBA Finals, which is borderline staggering when you consider that the election is 455 days from now. We wanted to understand what people thought of the Cleveland debate, so we looked at 20,000 randomly selected tweets from the night in question and ran the numbers. What issues resonated with voters? Which candidates were viewed most negatively? And are we really considering voting for a well-monied, sentient toupee?

Here’s what we found:

Nobody will shut up about Trump

We’ll cut to the chase: Donald Trump dominated the conversation. He received 44% of all candidate mentions on social media during (and directly after) the debate. The next highest? Jeb Bush, at about 11%. That’s four times as much attention as the presumptive frontrunner.


That said, people weren’t particularly positive when they mentioned Trump. 62% of all tweets were negative (compared to 21.6% positives). This wasn’t a groundswell of popularity as much as it was thousands of people muttering “get a load of this clown” under their breaths.

It’s worth noting here that Nate Silver of FiveThirtyEight fame (and our keynote at the Rich Data Summit–ahem, buy tickets here) has pegged Trump’s chances of winning the nomination at a robust and heady 2%. In other words, there’s a lot of sound and fury here, but you can do well to ignore any poll anointing the WWE Hall of Famer as the front-runner.

Everybody hates Chris

Since Twitter skews younger than most social networks and since younger people skew left on the political spectrum, it’s likely not too surprising that no Republican candidate got overwhelmingly positive marks on their performance last Thursday. Relative unknown John Kasich was the only candidate who had higher positives than negatives (41.6% to 35%, respectively, with the rest marked as “neutral”) but his message is more likely to resonate with Twitter’s demographic than, say, Mike Huckabee’s.

What was staggering, however, was how straight-up execrable Chris Christie’s scores were.


See that? The green is positive and Christie didn’t have much. In fact, he didn’t reach 3% positive. There are invasive surgeries with higher positives than that. Most invasive surgeries, frankly.

Jeb Bush didn’t fare much better (3.2% positive), while Marco Rubio (28.1%) enjoyed the highest marks outside of the aforementioned Kasich. Nobody really comes out looking all that good, mind you, but having less than three out of every hundred people being even vaguely positive about your campaign seems troubling at best.

The medium over the message

Much of the spotlight on the GOP debate was focused not on the candidates themselves, but FOX News and the moderators who ran the discussion. There were a bevy of articles, including some by noted FOX critics like Frank Bruni, who applauded the way the debate was run. The questions were more pointed than most expected and candidates weren’t given carte blanche to recite shabby boilerplate while America nodded in unison.

In fact, over half of all tweets concerned with subject matter were about the network itself.


That said, in contrast to the fairly glowing reviews from umpteen media outlets, a lot of these tweets were negative. About three in every four, actually. A sizable amount of these were from Trump supporters, which is unsurprisingly, being that his camp felt the debate was “unfair” and his followers have taken to delivering death threats to moderator Megyn Kelly. Maybe don’t do that, guys.

Oh, you vote too?

This crop of GOP might have a problem when they realize that half of America happens to be women. On women’s issues (but not abortion), 2.7% of tweets we sampled had a positive sentiment. When you consider the outsized percent of Trump-related snippets in this dataset and remember some of the pearls of wisdom he dropped onstage last Thursday, that number starts to make a bit more sense.


For that matter, candidates views on immigration were held in the highest esteem (though still far more negatively than positively). Again, we stress that it’s worth remembering the demographics on Twitter and the fact that we’re about 15 months away from voting, but the drastic negativity here isn’t something that should make members of the Grand Old Party all that excited either. Maybe when Trump innevitably bows out, we’ll see the quality of discourse improve. At the very least, the quality of Rosie-O’Donnell-related discourse should look a bit better.

How we did it

First, we used Twitter’s API to scrape 20,000 message with relevant hashtags like #GOPDebate. Then, we built a job where our contributors looked at that source data and answered four simple questions.

1- Is this tweet relevant and from a person? We’re not interested in tweets from news outlets or brands for this. We wanted to see what the public thought, not your local news affiliate.

2- What candidate was mentioned in the tweet? Next, we provided our contributor base with a dropdown menu of all candidates from the evening debate, as well as an option to select “No candidate mentioned.” That allowed us to get sentiment about the debate in total or a certain issue, even if no particular candidate was called out by name.

3- What subject was mentioned in the tweet? Similarly, users selected from another dropdown with issues most discussed that evening.

4- What was the sentiment of the tweet? Lastly, we of course asked for sentiment on a simple positive, negative, and neutral scale. Strictly informational tweets were to be marked neutral.

Here’s how it looked:


Once that was done, we downloaded our enriched dataset, dug into the findings, and graphed it all out above. You can download the entire dataset on our Data for Everyone page if you’d like to dig in, do some visualizations, or just read a whole bunch of tweets about Donald Trump for fun.

Lastly, if you’d like to learn some best practices on how to run your sentiment analysis jobs, we’re hosting a webinar with Oracle on Thursday on this week. We’ll be using real-world use-cases to show how people-powered sentiment analysis works and there will be a Q&A after if you have any questions. You can register for free here.