A new Oxford University study found that around 11 percent of content shared on Twitter that was related to the upcoming June 8 U.K. election was “junk news.” The study was based on data from Twitter, which “provides free access to a sample of public tweets posted on the platform.” By contrast, Facebook does not to share data with researchers trying to investigate fake news.
Fake news has become recognized as an international problem following the 2016 U.S. presidential election campaign; according to one analysis, it outperformed real news on Facebook toward the end of the 2016 campaign. Many of the websites that carry fake news have, along with the “alt-right," become part of an ecosystem that is able to mobilize to spread misinformation in election campaigns. Social media and online platforms such as Facebook and Google took some steps after the U.S. election in response to mounting public pressure, including working with fact-checking organizations to try to tamp down the problem for elections in France and Germany, although questions have been raised about the efficacy of those steps.
The Oxford study examined nearly 2.5 million tweets sent from the U.K. between May 27 and June 2. It found that during that time frame, “junk news accounts for over a third of other political news and information and accounts for 11.4% of the relevant content shared.” That amount showed that “UK users were not sharing as much junk news in their political conversations as” was shared during the 2016 U.S. election, “where the level of junk news shared was significantly higher.” But the study also found that it was a “higher percentage of junk news content than social media users who were actively discussing German politics and French politics during election periods” shared. From the June 5 study:
Table 3 explains the distribution of content shared by UK Twitter users and reveals that the largest proportion of content being shared by Twitter users interested in UK politics comes from professional news organizations, which accounts for 53.6% of the relevant content shared. Relevant content is calculated after non-political content, spam, irrelevant social media, language and inaccessible content have been removed.
Junk news accounts for over a third of other political news and information and accounts for 11.4% of the relevant content shared. Within the professional news content that was shared, the BBC was most popular, with 22.7% of professional news coming from this source. This was followed by The Guardian with 17.7% of links directing to the newspaper’s website. A high percentage of other political content that was shared comes from citizen-generated sources like personal blogs or civil society organizations. The number of links to such sources was higher than the number of links to junk news. Like in our earlier UK election study, Russian sources did not feature prominently in the sample, and no content was shared that could be attributed to WikiLeaks. This was in contrast to our project’s previous memos on the US and French elections.
Earlier in the election campaign, UK social media users shared a higher percentage of junk news content than social media users who were actively discussing German politics and French politics during election periods. In the second sampling period, the proportion of relevant content shared on UK social media identified as junk news was 11.4%, compared to 12.6% during the first UK sampling period, 12.5% in Germany and 5.1% and 7.6% respectively in the two election rounds in France. We also found that UK users were not sharing as much junk news in their political conversations as US users in the lead up to the 2016 elections, where the level of junk news shared was significantly higher. In the days leading up to the US election, we did a close study of junk news consumption among Michigan voters and found users were sharing as much junk news as professional news content at around 33% of total content each.
Substantive differences between the qualities of political conversations are evident in other ways. In the US sample, 33.5% of relevant links being shared led to professional news content. In Germany this was 55.3%, and in France this was between 49.4% and 57% of relevant links across both election rounds. Similarly, in the current UK-based study we show that 53.6% of relevant links being shared led to professional news content. In the initial UK sampling period this was almost identical at 53.4%. Having compared the content shared by UK users across two sampling periods, we can show that the quality of information shared did not differ substantially over time. This is different to the other countries we had investigated, where the quality of information shared deteriorated as the election drew closer. We are also able to show that individuals discussing politics over social media in the European countries sampled tend to share more high quality information sources than US users.
Content about the Labour Party strongly dominated traffic on Twitter in the second sampling period, showing a substantial increase from the earlier in the campaign. Social media users in the UK shared five links to professional news and information for every one link to junk news.
CNET, in a write-up of the study, noted that researchers “turned to Twitter, which allows access to 1 percent of its global daily data for free” but that Facebook “doesn't allow for its data to be viewed.” Multiple experts have called on social media organizations to share their data to help understand how fake news spreads about how it can be addressed. That includes one of the Oxford study’s authors, who told CNET that “a good starting part” to fight fake news “would be sharing more data” among academics, nonprofits and tech companies. Yet Facebook has refused to share its data on fake news with experts and analysts, and Facebook’s shareholders and board members, including CEO Mark Zuckerberg, recently rejected a proposal to publish a report on how fake news impacts the social media giant.