Mining Social Media for Presidential Primary Insights
More people now get their political news from social media than network news, radio or newspapers, a new report by the Pew Research Center finds. But how that’s shaped the presidential primaries so far is anyone’s guess.
Greg Wawro, a political scientist at Columbia University, organized a hackathon last month to explore the question. Five teams of students sifted through 60,000 records tied to the presidential primaries, ranging from candidate press releases to posts on Twitter, Facebook and You Tube, to see how candidates used the medium to their advantage.
The time period under analysis—April through December—saw at least three outsiders emerge as serious contenders: Donald Trump and Ben Carson on the right, and Vermont senator Bernie Sanders on the left. Their rise coincided with terrorist attacks in Paris and San Bernardino, Calif., touching off a bitter debate about immigration that continues to dominate the election.
Some students used text analysis tools to see what words recurred most often on social media and to gauge whether posts had a positive or negative tone. Others looked at how candidates capitalized on current events to create buzz, and whether speeches, debates and controversial remarks helped boost their popularity.
Data was provided by voxgov, a company that makes unedited documents scraped from federal government websites available to subscribers; Columbia’s Data Science Institute and Department of Political Science co-sponsored the event.
Two teams were singled out for special recognition.
The first, calling itself P = MV, the equation for momentum, found that candidates boosted their popularity by using social media to capitalize on televised debates and speeches.
“The popularity surge begins immediately, sometimes minutes, after a controversial interview or during the live debate,” said team member Bernat Ivancsics, a master’s student at Columbia's Graduate School of Journalism.
For Trump and Sanders, surging numbers of Twitter followers translated into a sharp climb in the polls. From June through January, Trump’s Twitter following jumped 86 percent, to 5.6 million, and his position in Republican polls grew from 4.9 percent to 36.9 percent, according to aggregated poll data from the Huffington Post. During the same time frame, Sanders’s Twitter following grew 1,644 percent to 1.1 million, and his standing in the Democratic polls grew from 12.4 percent to 33.3 percent.
The second team, calling itself Hacks and Hack Nots, analyzed the words most commonly used by candidates on Twitter. In June and July, words like “tax” and “independence” dominated the discourse, but by late fall, “state,” “debate” and “refugees” rose to prominence.
"We analyzed key words to assess how the candidates used social media to stay on message and bolster their policy positions after big events like the Paris attacks or the gay marriage bill," said team member Jack Goodman, a master's student at Columbia's Graduate School of Journalism.
Social media has given politicians a new way to reach the public. Once filtered by mainstream media, candidates can now speak directly to voters. And voters can talk back, bringing a level of intimacy and interaction to politics that once seemed unimaginable.
President Obama famously seized on social media during the 2012 election. Through a two-week period in June, Obama’s campaign published 614 posts to Mitt Romney’s 168, a 2012 study by the Pew Research Center found.
The gap on Twitter was even bigger. Obama’s campaign fired off an average of 29 updates per day to Romney’s one. Obama’s mastery of this powerful, new communication tool translated into victory just as it had for Roosevelt, with radio, and John F. Kennedy, with TV.
The low cost of social media may play an increasingly important role in future elections by defraying campaign costs. “The real revolution might come by making it cheaper to run for political office,” said Wawro. “Ad buys and commercials typically constitute the largest proportion of a campaign's budget. Social media allows candidates to reach voters at much lower cost.”
Some teams are updating their projects as results come in from New Hampshire and other presidential primaries. Among the questions that Wawro says he’d like to explore in future hackathons is how positive and negative messages on social media translate into perceptions of a candidate's viability, and whether social media responses track how well candidates’ messages resonate with voters.
Team P=MV: Ethan Cooper (Columbia College), Han Cui (Statistics), Bernat Ivancsics (Journalism), Sarah Salvadore (Journalism) and Xinjia Zhang (Statistics)
Team Hacks and Hack Nots: Jingjing Feng (Statistics), Jack Goodman (Journalism), Sitao Luan (Statistics), Aditi Sangal (Journalism) and Jeremy Staub (Columbia College).
— Kim Martineau