Instant debate feedback
The issues Twitter users felt the two candidates handled the best during the final presidential debate Wednesday night:
Donald Trump
1. Foreign affairs and conflict
2. Fitness to be president
3. Trade/economy
Hillary Clinton
1. Post election support/rigged election
2. Economy
3. Obama administration’s stimulus package
Source: Twitris
Donald Trump gets mentioned on Twitter three times more than Hillary Clinton, but researchers at Wright State University said it’s the hashtags that provide clues to whose Ohio supporters will turn out for the November election.
While millions watched the final presidential debate Wednesday night, 25 researchers at Wright State were watching on campus from a debate “war room” — gauging support for the presidential candidates by analyzing tweets in real-time.
“We’re taking this huge volume of tweets and saying ‘what does this mean?’ ” said James Mainord, CEO of Cognovi Labs, a Dayton firm working with WSU researchers on analysis.
The group analyzed more than 15,000 tweets from Ohio and 180,000 from across the country during the last debate. The researchers also have pulled data from more than 100 million election-related tweets over the past year, Mainord said.
To predict the election outcome, researchers are separating the talk on Twitter from actual action. To do that, they are using a program called “Twitris,” created by Wright State Computer Science and Engineering Professor Amit Sheth.
The program can single out hashtags, which are phrases people post and follow on the social media platform Twitter.
A few the researchers have been following are #OHVotesEarly, #OhioVotesEarly and the phrase “vote early.” They are used by Ohioans who are taking advantage of early voting.
Between Oct. 12 and Oct. 20, 1,100 tweets in Ohio were found to carry those hashtags. Of those tweets, 77.5 percent supported Clinton and 22.5 percent were for Trump, according to Twitris data.
There may be more tweets about Trump in general, but they don’t mean voters are casting ballots for him, Mainord said.
“Clinton leads with people in Ohio who actually talk about voting,” Sheth said.
A lot of Trump’s Twitter supporters are not “real people,” Mainord said. “Significantly more” supporters of Trump than Clinton use “robots,” which are accounts set to automatically tweet support during debates and rallies, Mainord said.
To get a clearer take on who is winning the election, researchers filter out the fakes so they don’t skew results. Not only will that make predictions more precise, it allowed them to better gauge sentiment toward the candidates before and after the final debate.
Hard feelings
Recent scandals affecting both presidential campaigns seem to have shifted public sentiment in a downward direction for each candidate on Twitter, Mainord said.
Public sentiment toward Trump dropped after a video showed him making lewd comments about women, according to a Twitris analysis. Opinions of Trump on Twitter plummeted when women started making allegations about his behavior.
“It’s like he can’t maintain a lead,” Mainord said.
The many WikiLeaks releases of emails from Clinton’s campaign also sank opinions of her on Twitter over the past week, Mainord said.
Before the final debate, it became clear that Clinton is winning the fight for public opinion, not because she’s popular but because Twitter users dislike her less than Trump.
During the debates, researchers observed a map of the United States for each candidate. States that were shaded red on their monitors represented negative tweets about a candidate. States shaded yellow were neutral and states showed up as green when there were more positive tweets.
Clinton has effectively “flipped the map,” said Mike Wiehe, director of the applied policy research institute at WSU.
“She had a lot more red during the first two debates,” Wiehe said.
Trump’s map went from being more positive and neutral to negative since the first debate, Wiehe said.
Ohio was shaded red for each candidate during most of the final debate, but it occasionally turned green.
The issues
Ohio’s Twitter users were posting mostly negative thoughts about the candidates Wednesday night, but there were a few times they seemed to support their stances.
Around 9:12 p.m., Ohio shifted from red to a slight shade of green on the map for both Trump and Clinton as they spoke about the U.S. Supreme Court and then abortion.
“It might just be kind of a rush of tweets where people are just cheering for their team,” Mainord said.
Researchers were able to match up Twitter reaction to debate topics with the help of Jacqueline Gill, a WSU graduate student studying public administration. Gill took notes on what was said in the debate so researchers could line up spikes and drops in reactions to topics.
“Suddenly you see what everybody else on Twitter is thinking,” Gill said. “That’s a really big deal.”
Clinton received more positive reactions than Trump on most issues, including immigration, abortion, the economy and entitlements. Trump only surpassed her in reaction on the topics of “fitness to be president,” areas of foreign affairs and on trade.
Trump’s numbers on issues tend to fluctuate while Clinton receives consistent response on all topics, Mainord said. It’s a reflection, researchers said, of how many times Trump has “shot himself in the foot” and how Clinton supports standard policy positions.
Track record
As the election nears, a deadline looms for Mainord and Sheth to make their prediction. Using data from Twitris, they will likely do that on Nov. 8, Election Day.
In 2012, they predicted Barack Obama would be re-elected and that the election results would be known around 11 p.m., Sheth said. They were right.
Twitris is known as one of the only analytic tools to make the correct call on Brexit, the United Kingdom’s vote to leave the European Union.
There was steady support for the UK to stay in the union, but on the day of the vote support for the country to leave surged on Twitter.
“We were looking at data and it felt so different from what everyone else was seeing,” Mainord said of the Brexit vote. “Finally, we said OK, there’s a pretty strong break for ‘leave.’ ”
About the Author