After the shock of the 2016 presidential election, many Americans found psychological refuge in a simple explanation for why Donald Trump won: "fake news." False or misleading information published by dubious for-profit websites had spread widely on Facebook, reaching millions of people in the final months of the campaign. This development provided a tidy narrative that resonated with concerns about potential online echo chambers.
More than two years later, we can now evaluate these claims. And it turns out that many of the initial conclusions that observers reached about the scope of fake news consumption, and its effects on our politics, were exaggerated or incorrect. Relatively few people consumed this form of content directly during the 2016 campaign, and even fewer did so before the 2018 election. Fake news consumption is concentrated among a narrow subset of Americans with the most conservative news diets. And, most notably, no credible evidence exists that exposure to fake news changed the outcome of the 2016 election.
The fake news panic echoes fears that prior forms of communication would brainwash the public. Just as exaggerated accounts of hysteria over Orson Welles' "War of the Worlds" broadcast took advantage of doubts about radio, claims about the reach and influence of fake news express people's broader concerns about social media and the internet.
Many important concerns about online misinformation still remain, including the influence of the fake news audience, the difficulty of countering fake news at scale, the dangers of Facebook's size, and the threat of YouTube-based radicalization. But none of these questions can be adequately addressed without creating a reality-based debate that puts fake news in context as just one of the many sources of misinformation in our politics.
Any conversation about fake news has to start with hard data on the extent of the problem. Unfortunately, these data are lacking. Most discussions of fake news exposure rely on simple counts of views or readers that lack essential context on who was exposed to the content and how frequently and what other information they also consume. By contrast, a study I conducted with political scientists Andrew Guess and Jason Reifler drew on nationally representative laptop/desktop web traffic data from an online panel, allowing us to measure who visited fake news sites before the 2016 election with unprecedented precision.
We found that the reach of fake news declined dramatically in the period before the 2018 midterm elections.
We find that only 27 percent of Americans visited fake news websites, which we define as recently created sites that frequently published false or misleading claims that overwhelmingly favor one of the presidential candidates, in the weeks before the 2016 election. These visits show the expected political skew — Clinton and Trump supporters tended to prefer pro-Clinton and pro-Trump sites, respectively — but made up only about 2 percent of the information people consumed from websites focusing on hard news topics.
Consistent with behavioral evidence showing that online echo chambers are relatively rare, fake news consumption was concentrated among the 10 percent of Americans with the most conservative news diets, who were responsible for approximately six in 10 visits to fake news websites during this period. Even in that group, however, fake news made up less than 8 percent of their total news diet. Finally, people ages 60 and over consumed more fake news than other cohorts, which may reflect a lack of digital literacy or simply having more time to read news. (Other scholars have found similar patterns in Facebook sharing and Twitter sharing and consumption of fake news.)
Moreover, the reach of fake news declined dramatically in the period before the 2018 midterm elections. In a new report co-authored with Benjamin Lyons and Jacob Montgomery, Guess, Reifler, and I found that just 7 percent of Americans visited one of the fake news sites that we previously identified in 2018 — a decline of approximately 75 percent in relative terms. (Consumption differences between groups by age, partisanship, and news diets remained similar to 2016.)
Moreover, the role of Facebook in the spread of fake news appears to have changed. In 2016, the site differentially appeared in web traffic just before visits to fake news sites, suggesting it played a key role in enabling the spread of fake news. No such pattern is apparent in the 2018 data. This result, which echoes findings from other studies and holds with an updated set of websites we compiled before the 2018 study, suggests that the platform's efforts to limit the reach of fake news are having some impact. (Such inferences are necessarily indirect because Facebook remains largely closed to outside research for now.)
Research is also providing new insights into another limitation on the effects of fake news: who believes it. One key factor is directionally motivated reasoning — people's increased willingness to accept dubious claims that are consistent with their partisan or candidate preferences. When my co-authors and I tested the perceived accuracy of a number of fake news headlines, belief in the accuracy of headlines that favored respondents' preferred party (17 to 48 percent) was much higher than headlines that favored the opposition party (10 to 22 percent). Another important element, as psychologists Gordon Pennycook and David Rand emphasize, is analytical thinking ability. People who score low on this measure are especially prone to endorse false headlines. By contrast, those who score high are more likely to reject fake news even when it supports their political viewpoint.
Finally, there remains no evidence that fake news changed the result of the 2016 election. Any such claim must take into account not just the reach of fake news but also the proportion of those exposed to it whose behavior could be changed. As noted above, approximately six in 10 visits to fake news websites came from the 10 percent of Americans with the most conservative news diets — a group that was already especially likely to vote and to support Donald Trump. Accordingly, my colleagues and I find no association between pro-Trump fake news exposure and differential shifts in candidate support or voter turnout.
These findings do not alleviate every concern about fake news, of course. First, even if relatively few people consume fake news, those consumers may be especially politically active and thus disproportionately influential in our politics. In particular, fake news consumers may be especially important in party politics, which is highly responsive to people with intense preferences who vote in primaries. Fake news readers are also likely to disseminate the information they encounter from fake news websites via online and social networks, indirectly exposing many more people than would consume it directly.
Second, fake news is likely to have negative effects that extend beyond election outcomes. Most important, fake news websites distort and demean public debate in our democracy by spreading untruths. In addition, fake news exposure may intensify the intense loathing that partisans increasingly feel toward the opposition party. Content on these sites often portrays opposition candidates and party members in vitriolic terms, claiming, for instance, that Hillary Clinton sold weapons to ISIS. We find no measurable effect of exposure to a single fake news article on feelings toward the opposition party, but the effects of fake news exposure may accumulate over time.
Third, more needs to be learned about how to most effectively counter fake news. Providing online fact-checks reduces belief in headlines from these sites, but the scale of Facebook and other platforms outstrips the capacity of fact-checkers to keep up. Further interventions would therefore necessarily rely on algorithms, which would outsource further power over political speech to private companies. Another important answer is education, but current efforts to promote digital literacy have not been carefully tested and will struggle to reach the older Americans who are overrepresented in the fake news audience. Moreover, we must be careful to teach people to distinguish between good and bad information rather than promoting generalized cynicism. Broad warnings about fake news can create unintended spillovers that reduce belief in legitimate news coverage.
We found that the reach of fake news declined dramatically in the period before the 2018 midterm elections.
Fourth, relatively little is known about the effects of video. Amplification of extremism and false content is especially worrisome on YouTube given the amount of time some audiences spend on the platform and the way its algorithms may amplify misinformation (though YouTube now says it is trying to limit the reach of dubious content ). However, the claim that YouTube radicalizes large numbers of viewers is still a conjecture based on anecdotes that consider only the people who become extremists (not the vast majority, who do not).
Finally, the unprecedented scale of Facebook has created a dangerously alluring target that invites endless attacks from both fake news entrepreneurs and political opportunists. The company's resources allow it to deploy unprecedented countermeasures to defend itself, but we are concentrating risk in a single closed platform. Inevitably, Facebook's defenses will fail again, and, as in 2016, we may not realize that such a failure has taken place until it is too late. These failures can be especially pernicious in less developed democracies and authoritarian states, where Facebook and WhatsApp have already been linked to violence. (Disclosure: I have received funding from the company to study how to counter misinformation on WhatsApp in India.)
Fears about the harmful effects of fake news should not be dismissed — as noted above, serious questions remain about the potentially pernicious effects of online misinformation — but the evidence is far more nuanced than current debate allows. Unfortunately, blaming fake news for what's wrong with our politics is easier than recognizing it as a response to the demand for misinformation that polarization has created.
We must also recognize that fake news entrepreneurs aren't the only people trying to meet the demand for this kind of content. The most worrisome misinformation in U.S. politics remains the old-fashioned kind: false and misleading statements made by elected officials who dominate news coverage and wield the powers of government. As 2016 illustrated, the costs of making unsupported claims are low in highly partisan contexts, which limits the incentive for politicians to avoid them. Reading a fact-check of Trump's convention speech, for instance, reduced false beliefs that crime was increasing in the long term but did not affect his support.
Trump has gone on to make more than 8,000 false claims during his first two years in office, many of which are amplified in cable news chyrons or in credulous online news headlines. As a result, a sizable minority of Americans still believes some of his most frequently repeated false claims. These beliefs persist despite unprecedented fact-checking efforts, which struggle to overcome unprecedented levels of polarization in media trust. Even more corrosively, Trump's supporters are increasingly rationalizing those falsehoods. Belief in the importance of presidential candidates being honest has declined from 71 percent among Republicans in 2007 to just 49 percent today, threatening the previously uncontested norm that the president should be expected to say things that are true, or at least not obviously false.
Ultimately, fake news helped alert us to the threat, but it is Trump who has most effectively weaponized partisan misinformation in our politics. Understanding how to prevent our leaders from exploiting this vulnerability further must therefore be a top democratic priority.
Brendan Nyhan is a professor of public policy at the University of Michigan.