November 21, 2016

Did social media help elect Donald Trump?

WEST LAFAYETTE, Ind. — Communication professor Sorin Adam Matei is a dedicated social scientist who believes in political polling. But he saw early trends in Donald Trump’s social media efforts that were not consistent with the many political pollsters’ findings.

“I always laugh at people who make fun of surveys and polls. Usually, they don’t understand how polling works,” Matei said. “No one makes this stuff up. There were many polls, over periods of time, all going one direction. I looked at the polls and then I looked at social media. There was this big cognitive dissonance. Social media was not aligned with the reality of polling. Back in June, when I first noticed that Trump was getting several times more likes on Facebook and retweets on Twitter, I was more likely to say that Trump’s numbers were a speculative bubble, divorced from reality. Something akin to the early dot.com era, when anything digital attracted big money regardless of profitability prospects. I solved the cognitive dissonance by saying the social media lives in a world of its own. A virtual world.”

Matei, a professor in the Brian Lamb School of Communication at Purdue University, counted the number of tweets from each candidate’s official Twitter account and tracked the average social engagement of each tweet by retweets and “favorite” tags.

“As Election Day was coming closer I was looking at the polls, and there was something interesting; Trump was getting closer in the polls but it was still a long shot. Yet, his social media dominance was holding. I kept thinking he was on the losing side, but then I’d look at the social media numbers side by side with the polling again and it was looking at me right in the face. Yet, again, I am only human. Until Wisconsin was called I still did not believe the social media data.”

The final numbers, which he calculated a few days after the election, showed that Trump significantly outperformed Hillary Clinton on social media, specifically Twitter, throughout the campaign. From the primary through Election Day, Trump’s famous, and occasionally, infamous tweets were being re-tweeted by a margin of nearly 3-1 over Clinton’s tweets.

“And if you go by what we know about the demographics of the two candidates, Hillary should have done better on social media,” Matei said. “Her supporters were supposedly better educated, more cosmopolitan, urban and tech-savvy.” According to the Pew Internet surveys, Twitter is, traditionally, used by the younger, more urban, educated Americans (http://www.pewinternet.org/2015/08/19/the-demographics-of-social-media-users/). They are also more likely to vote Democratic. Naturally, Hillary should’ve attracted more Twitter followers and engagement than Trump.

The trend continued and became more pronounced.

“On Election Day, Trump went ‘boom’,” Matei said. “Sure, Hillary’s “enthusiasm factor” (re-tweets and favorites) also went from a few thousand to about 20,000. His, however, went from a healthy number of over 10,000 to 60,000. He got a lot of mileage out of social media. Those re-tweets showed that people were willing to work with him, help him out and spread the word to the very end.”

Facebook is the largest social media platform in the world. Matei looked at similar engagement variables on Facebook in June and found Trump did even better. “There were times he’d have 400 or 500 percent more interaction.”

There were, however, recent claims that Trump’s social media traffic was artificial, created by bots and fake news. “It is quite possible that Trump’s margins in social media were inflated by both “white” (legit) and “black” (spammy) automated Twitter accounts, or bots,’” Matei said. A bot posts brief messages at given intervals without human interventions. They can produce more content that is highly targeted and favorable to a candidate than human posters.

“My colleagues at the Oxford Internet Institute noticed a lot of tweets generated by automated Twitter accounts that “colonized” hashtags dedicated to Hillary Clinton. I am not sure, however, if what I measured, retweeting, was affected to the same degree by automated accounts,” he said. “This is something we need to look into. Also, the early data coming from Facebook, especially the number of comments and likes, which are much harder to fake, went in a similar direction. It would be interesting to look at the complete Facebook data series and compare it with the complete Twitter series, to get a better picture.”

Matei will release soon an analysis of Facebook likes and shares for the entire period of the primaries and the electoral campaign.

Writer: Howard Hewitt, 765-4949541, hhewitt@purdue.edu 

Source: Sorin Matei, 317-416-5807, smatei@purdue.edu  

Note to Journalists: Matei is available for phone and e-mail interviews.

Related Web sites:

College of Liberal Arts


Research News

More Research News

Purdue University, 610 Purdue Mall, West Lafayette, IN 47907, (765) 494-4600

© 2015-17 Purdue University | An equal access/equal opportunity university | Copyright Complaints | Maintained by Office of Strategic Communications

Trouble with this page? Disability-related accessibility issue? Please contact News Service at purduenews@purdue.edu.