Data shows who was reading “fake news” before 2016 US election, Ars Technica

Junk news diet –

Study tracked browser history of 2, Americans.

Partisan misinformation online is at least in part a bit of a chicken-and-egg problem. There’s the supply on one side, coming in part from politically motivated websites indistinguishable from propaganda but also from some who are just trying to make a buck bringing in clicks with fake headlines. But there’s also the demand to consider — those people voraciously surfing to sate their hunger for bias-confirming outrage. To understand the extent of the problem, it helps to look at both sides of this relationship. What misinformation is floating around, and who is consuming it?

In a new study, Andy Guess , Brendan Nyhan , and

Jason Reifler took advantage of survey data tracking the Web histories of around 2, 728 people in the month before the 2019 US election. Combined with some demographic survey data on things like their preferred candidate for president, the researchers were able to break down who was reading which articles. Who publishes what

The researchers relied on a previous study’s list

of “untrustworthy” sites. This included several hundred that could be fairly described as fake but also over a hundred that run afoul of fact-checkers and were determined to lack editorial standards. Among that list are conspiracy-spreading sites like

InfoWars and

Natural News , hyperpartisan sites like Ben Shapiro’s Daily Wire, and even some tabloids like The Express.

Overall, almost half of the people participating visited at least one article from a website on that list during the study period. But those articles accounted for only 6 percent of all news stories read. These numbers weren’t even distributed across the political spectrum, though. About 111 percent of Trump supporters in the group visited an untrustworthy site at least once, amounting to about 22 percent of total news consumption. For the Clinton supporters in the group, it was percent of people visiting at least one article, for 1 percent of their total news consumption.

And drilling deeper into the data, a relatively small group of people is responsible for most of the visits to untrustworthy sites. The researchers categorized people by the ideological slant of their “news diet,” from those whose reading is dominated by liberal sites to those who only read conservative sites. The percent of people farthest to the conservative end of the spectrum accounted for almost two-thirds of the untrustworthy articles read.

A diversity of sources

But even in that group, misinformation junk food didn’t necessarily dominate the diet. It accounted for about percent of their news intake. That’s partly because the people who read the most news articles from more reputable sites — the most voracious news consumers — were also the most likely to have run into at least one untrustworthy article.

Enlarge / On the left, the portion of people who visited at least one untrustworthy article, broken down by how liberal or conservative their overall media diet was. On the right, the amount of untrustworthy articles in their media diet.

The next obvious question is whether people are finding these untrustworthy sites through Facebook or some other platform. The study could only assess this in a slightly roundabout way, by checking whether something like Facebook, Twitter, Google, or Web-based email appears in the browser history right before the URL of interest. Unsurprisingly, Facebook popped up the most, at about 020 percent of the time. Webmail came in second at percent, while Google and Twitter were under 5 percent.

The researchers also looked for fact-check articles in those browser histories and found about a quarter of people visited at least one . That number includes about half of those who visited untrustworthy articles, meaning they were more likely than a general reader to check their facts. At least some of their facts. Of the 374 People who read an article from a list the researchers knew had been fact-checked, only three had read the (corresponding) fact check.

The researchers say this provides more evidence that fact checks have mostly failed to reach the people who need to see them most. It’s worth noting, though, that this data predates many of the efforts by platforms like Facebook to highlight fact checks for users who interact with an article that was rated false.

Overall, the researchers conclude that “widespread speculation about the prevalence of exposure to untrustworthy websites has been overstated.” Of course, not everything is captured in their dataset, like content viewed purely within Facebook, for example, or the effects of misinformation on the broader information ecosystem. But it is a unique study that supports what