Did Fake News Help Trump Win?

By Matthew Gentzkow

Posted on February 16, 2018


American democracy has been repeatedly buffeted by changes in media technology. From the introduction of cheap “penny papers” in the 19th century, to the rise of radio and television, to the early days of the internet, new media have both produced profound effects, and provoked outsized anxieties.


Following the 2016 election, the focus of concern has shifted to social media. Social media has rapidly become one of the most important channels by which people consume political news and information. This may have important upsides for democracy. Social networks have engaged a large swath of voters who do not consume traditional media, allowed more direct communication between politicians and voters, and, by most accounts, made the “marketplace of ideas” more diverse and competitive.


However, as with all previous media technologies, the potential downsides loom far larger, at least in the public imagination. These include segregation of voters into “echo chambers” or “filter bubbles”, vulnerability to manipulation by bad actors via algorithmic targeting or botnets, and above all the proliferation of misinformation or, as the ubiquitous cliché would have it, “fake news”.


Research teams in academia, non-profits, and private firms have begun to quantify the scale of these risks, and to evaluate potential solutions. Though the evidence remains fragmentary, this collective effort provides some much-needed discipline to the broader conversation.


In a recent paper with NYU economist Hunt Allcott (“Social Media and Fake News in the 2016 Election”), we offer one contribution in this vein, combining audience data, fact-checking archives, and a new, specially commissioned, 1,200-person post-election survey to gauge the reach of fake news in the run-up to the 2016 election and the economic forces at play in its distribution.


We define fake news to be news articles that are intentionally and verifiably false, with the potential to mislead readers. This definition includes intentionally fabricated news stories, such as a widely shared article from the now-defunct website denverguardian.com with the headline “FBI agent suspected in Hillary email leads found dead in apparent murder-suicide”. It also includes many articles that originate on satirical websites but could be misunderstood as factual when viewed in isolation on Facebook or Twitter. For example, in July 2016, the now defunct wtoe5news. com reported that Pope Francis had endorsed Donald Trump’s presidential candidacy. According to the site’s “about” page, “Most articles on wtoe5news.com are satire or pure fantasy.” But this disclaimer was not included in the article itself, and the story was shared more than one million times on Facebook.


Our definition rules out several close cousins of fake news: (1) unintentional reporting mistakes, (2) rumors that do not originate from a particular news article, (3) conspiracy theories spread by people who believe them to be true, (4) satire unlikely to be misconstrued, (5) false statements by politicians, and (6) reporting that is slanted or misleading but not outright false.


As a first step in our analysis, we sketched a stylized theoretical model of the market for news on social media. The model clarifies the key forces that lead fake news to arise and proliferate: low costs of production and distribution, difficulty for consumers in separating false from true stories, tastes on the part of consumers for partisan confirmation, and large economic returns from advertising. The model also highlights the fact that the potential social costs include not just biased beliefs about the fake news stories at issue, but also a reduction in trust in media outlets more generally.


Informed by this model, we then studied the role of fake news in the 2016 American presidential election.


First, we assessed the importance of social media relative to other sources of political news and information. Figure 1 shows the media which respondents to our survey described as their “most important” sources of news about the 2016 election. Television remains by far the most influential source, chosen as most important by a majority of voters. Only 14 percent of voters cite social media as their most important source. We conclude that social media was important, but far from dominant, and probably less important than some might infer from its prominence in the public discussion post-election.

Image: Pie Chart

We also assembled a database of 156 election-related news stories that were categorized as false by leading fact-checking websites in the three months before the election. This enabled us to confirm that fake news was both widely shared and heavily tilted in favor of Donald Trump. Our database contains 115 pro-Trump fake stories that were shared on Facebook a total of 30 million times, and 41 pro-Clinton fake stories shared a total of 7.6 million times.


Several benchmarks allowed us to measure the rate at which voters were exposed to fake news. The upper end of previously reported statistics for the ratio of page visits to shares of stories on Facebook would suggest that each of the 38 million shares of fake news in our database lead each to on average 20 clicks on the websites posting the stories. This translates into 760 million instances of a user clicking through and reading a fake news story, or about three stories read per American adult. A list of fake news websites, on which just over half of articles appear to be false, received 159 million visits during the month of the election, or 0.64 per US adult. In our post-election survey, about 15 percent of respondents recalled seeing each of 14 major pre-election fake news headlines, but about 14 percent also recalled seeing a set of placebo fake news headlines - untrue headlines that we invented and that never actually circulated. Using the difference between “real” fake news headlines and our placebos as a measure of true recall and projecting this to the universe of fake news articles in our database, we estimate that the average adult saw and remembered 1.14 stories. Taken together, these estimates suggest that the average US adult might have seen perhaps one or several (but not ten or a hundred) fake news stories in the months before the election.


Finally, we studied the ability of our survey respondents to distinguish false and true headlines. Education, age, and total media consumption are strongly associated with more accurate beliefs about whether headlines are true or false. Democrats and Republicans are both about 15 percent more likely to believe ideologically aligned headlines, and this ideologically tilted inference is stronger for those with more ideologically segregated social media networks.


This analysis is not sufficient to allow us to say anything conclusive about the potential impact of fake news on the 2016 election outcome, because we do not have an estimate of the way fake news exposure affected votes. However, back-of-the-envelope calculations suggest that seeing a single fake news story would have needed to be many times more persuasive than seeing a single TV commercial in order for fake news to have changed the election outcome.



The preceding is republished on TAP with permission by its author, Stanford economics professor Matthew Gentzkow and the Toulouse Network for Information Technology (TNIT). “Did Fake News Help Trump Win?” will soon be published in TNIT’s newsletter.