As the Democrats debate tonight and tomorrow, many people — journalists included — will be watching two screens: on one, the debates themselves, and on the other, Twitter’s stream of hot takes. Recent reports remind us that Twitter does not mirror the world at large. Yet journalists’ reliance on it shapes what becomes the news.
Journalists have long wanted to see and report public opinion clearly. My new research shows that amid declining trust in polls and crises in polling methodology, journalists have been turning to Twitter to understand public opinion. Here’s how that might be slanting the news.
Journalist use social media — mostly Twitter — to understand and report on public opinion.
In a study recently published in Journalism, I examined how news stories treat social media posts as representing public opinion and how these practices shape journalistic routines, especially during campaign reporting. To do this, I analyzed 385 news stories about the 2016 U.S. presidential election that used social media to report public opinion. To find these stories, I searched MediaCloud’s election database, focusing in on a variety of outlets classified by Pew’s State of the Media report. I hand-coded the stories, noting the presence of social media metrics and individual social media posts, and classifying how they were deployed by journalists as public opinion in the stories. This exploratory analysis then guided textual-analysis interviews I did with journalists who wrote stories appearing in my data set. I spoke with 18 journalists from outlets including The Washington Post, FiveThirtyEight, BuzzFeed, ABC News and USA Today.
Even though social media users don’t reflect the electorate, journalists regularly reported online sentiments and trends, mostly from Twitter, as voicing public opinion. Nearly 75 percent of the stories used tweets and other social media posts or metrics to illustrate the horse-race narrative, emphasizing who was ahead or behind, and to complement survey polling and “person on the street” quotes.
Social media posts are the new vox pop
A number of news stories simply reported on “takes” people tweeted about politicians or issues, amplifying social media posts as a new, intriguing way in which Americans expressed their opinions.
Many of the journalists I interviewed talked about starting with a “theory” for the story — and finding particular posts that could be used to illustrate that. In 2016, social media posts provided easy and ample evidence for journalists’ “meta-narratives” about candidates: Trump as a joke, Sanders with grass-roots support. Journalists told me they found these posts through their own timelines, even though research shows that journalists mostly follow other journalists — not exactly a representative sample of Americans. Many journalists also reported culling tweets from DataMinr, a platform that highlights tweets that are getting “attention” — and which several journalists told me they used as a de facto assignment editor.
Social media posts give journalists another way to quote people, much as with traditional interviews in local diners or at rallies. Like polls, they’re used to establish the appearance of objectivity while still allowing journalists to interpret events through their own expectations.
Journalists used social media posts to illustrate “horse race” coverage
In 2016, a large majority of news stories using social media to represent public opinion did so to showcase public evaluations of a certain party, candidate or their campaign strategies. This trend continues. A December 2018 Newsweek story touted Beto O’Rourke’s video views across social media platforms as showing his potential for beating both Sen. Elizabeth Warren (D-Mass.) and President Trump.
James Hohmann of this newspaper told me that after key events, he was able to turn to Twitter to help him interpret events such as primary debates, saying, “Like, the next morning, I would be able to say definitively, ‘Bernie Sanders won the debate because of these online mentions.” He used social media posts as evidence of the candidates’ missteps or successes, enabling him to, for instance, quote “30 really good tweets … to bolster the argument.”
Interpreting social media metrics challenge newsrooms
Journalists find social media metrics an attractive indicator of public opinion because they’re quantifiable and easy to access. They’re used to help counteract the fact that high-quality polling is scarce and that impressive early primary polling may just reflect name recognition.
Journalists often embedded social media metrics in their stories to indicate a candidate’s popularity. During the 2016 campaign, Facebook, Google, and Twitter data — say, about which candidate gained the most followers — were used in reporting to indicate how well candidates were performing during debates. As the crowded Democratic primary kicks into high gear this week with two nights of debates, we’re sure to see headlines and stories about which candidates got the most mentions and followers, and which moments went viral.
In 2016, while some outlets, including Slate, calculated their own Twitter metrics, most relied on metrics reported directly from the social media companies. That year, FiveThirtyEight and USA Today worked with data from Facebook for ongoing features. ABC News partnered with Facebook to live-stream general election debates, broadcasting metrics displaying elements such as which topics got the most engagement on their platform. Already, FiveThirtyEight is doing deep dives into Twitter data to compare and rank 2020 contenders.
Social media firms in 2016 worked to get their data into public opinion reporting — sharing metrics, sending news releases and partnering with outlets to provide data. This practice gives social media companies immense power as purveyors of a new form of public opinion that shapes media coverage of elections.
What does this mean for news coverage of the 2020 election campaigns?
For journalists, social media posts have become routine sources for instant responses and reactions to events. This presents a particular version of public opinion that may inform future news judgments and shape new narratives. This approach risks promoting the loudest and most polarized viewpoints, exacerbating the already highly partisan, conflict-focused nature of U.S. political reporting — and it risks bringing bots, trolls and even coordinated foreign efforts to mislead the American electorate into news stories as examples of public opinion.
While my study looked at the 2016 election, the findings have important implications for news coverage of the 2020 election. When journalists draw conclusions about public opinion from their own Twitter timelines, rely on metrics provided by social media companies and use Twitter attention to represent public attention, the news media is presenting a skewed version of the public.
Shannon C. McGregor (@shannimcg) researches political communication, social media, and public opinion as an assistant professor in the department of communication at the University of Utah.