The most recent manifestations of populism owe a portion of their rise to social media and the unfettered spread of false and misleading narratives or, as they are sometimes called, ‘alternative facts’. This column makes use of an online experiment conducted among Facebook users in France during the 2019 European Parliament elections to show that fact-checking can staunch the flow of false information, as can the imposition of small costs such as requiring an additional click to confirm a user’s willingness to share news.
The first report in CEPR’s Monitoring International Integration series focused on the decline of trust in Europe’s established political institutions and a surge in support for populist movements and policies (Dustmann et al. 2017). As discussed in VoxEU.org’s Debate on Populism, the recent rise of populism in Europe was driven by different factors including globalisation, automation, the global financial crisis and austerity policies that followed, the rise of immigration, and other forces related to identity and cultural divides. In addition, a potentially important driver of populism is the spread of mobile internet and social media that facilitates the dissemination of populist messages (Zhuravskaya et al. 2020).
In order to limit the dissemination of such ‘alternative facts’ – a term first used by Donald Trump’s advisor, Kellyanne Conway – countries such as Germany and France have introduced laws allowing regulators to block social networking accounts and sites. Many media and independent organisations have started large-scale fact-checking efforts. However, recent research suggests that fact-checking does not always work as expected. First, fact-checking may be too slow, as false news circulates particularly fast (Vosoughi et al. 2018). Second, ex post fact-checking may be too late, since it proves difficult to correct beliefs after an audience’s exposure to false or misleading statements (Swire et al. 2017, Nyhan et al. 2019, Barrera et al. 2020).
In order to slow the sharing of such alternative facts on social networks, one needs to understand the determinants of sharing, and particularly whether fact-checking can have an impact on the sharing of ‘alternative facts’. In our recent paper (Henry et al. 2020), we carried out an online randomised experiment to study the impact of fact-checking on Facebook users’ sharing of alternative facts.
In May 2019, in the context of the European Parliament elections, we used the Qualtrics platform to contact a representative sample of French voters with Facebook accounts. We first showed them misleading statements on the EU (with links to the exact source) made by the leaders of the extreme-right party Rassemblement National (RN). Until June 2018, this party was known as Front National; it was and still is led by Marine Le Pen. The first statement claimed that 87% of French laws come from European directives; the second stated that the EU sought to bring in 50 million immigrants to Europe by 2050. After seeing those statements, a randomly drawn third of the subjects were exposed to fact-checking information related to these statements, compiled from media sources – we refer to this treatment as “Imposed Fact-Check”. Another third was given the choice of viewing or not viewing this fact-checking information – we call this treatment the “Voluntary Fact-Check”. The remaining third (the “Alt-Facts” treatment) was not shown fact-checking information nor given an option to access it.
After being exposed to the ‘alternative facts’ and possibly to the fact-checking, participants had the opportunity to share the ‘alternative facts’ on Facebook. We found that exposing individuals to fact-checking information, or providing them with the opportunity to fact-check themselves, reduces sharing of ‘alternative facts’ by more than 25%. While the sharing rate is 14.7% in the Alt-Fact treatment, it falls to 10.2% in the Imposed Fact-Check treatment (the left-hand side panel of Figure 1). In the Voluntary Fact-Check treatment, where the users can choose whether to view fact-checking or not, the average rate of sharing is 10.8%, a rate not statistically different from the one in the Imposed Fact-Check treatment.
We also show that participants in these two treatments share the fact-checking information at similar rates: in the Imposed Fact-Check treatment, 14.3% participants share fact-checking, while the respective average sharing rate in the Voluntary Fact-Check is 11.5% (see the right-hand side panel of Figure 1). The difference between these two rates of sharing fact-checking information is only marginally statistically significant, and the magnitude of the difference is small. This is striking because in the Voluntary Fact-Check treatment, only 39% of participants chose to view the fact-checking information and thus had an option to share it.
Figure 1 The effect of imposed and voluntary fact-checking on sharing of false news and fact-checking information on social media
Which users choose to view the fact-checking information? Clicking and learning the fact-checking information is costly in terms of time, meaning that those who view the fact-checking (viewers) are different from those who do not (non-viewers). Using the pre-treatment determinants of sharing behaviour in Alt-Facts and Imposed Fact-Check treatments, we predict the ex ante propensity to share ‘alternative facts’ and fact-checking information for each viewer and non-viewer. We find that the ex-ante propensity to share is significantly higher for those who decided to view fact-checking information (viewers) than for those who decided not to view (non-viewers). The viewers are more inclined to share (either alternative facts or fact-checking information), probably because viewing fact-checking information is more valuable for them. Using this predicted propensity to share, we estimate the impact of viewing fact-checking on sharing ‘alternative facts’ and on sharing fact-checking. We find that the exposure of viewers to fact-checking decreases their sharing of ‘alternative facts’ by 67% (relative to their predicted propensity to share) and increases their sharing of fact-checking by 58%. Even more striking is the substantial reduction in sharing ‘alternative facts’ among non-viewers: even though these individuals decide not to view the fact-checking, the very knowledge of its existence reduces sharing of the alternative facts in question by 50%.
Our study identifies another important solution to slowing the dissemination of ‘alternative facts’. In our experiment, in order to share the ‘alternative facts’ via their Facebook account, users have to go through several clicks. We find that each additional click reduces the number of potential sharers by about 75%. This implies that social media sharing is very sensitive to even small non-monetary costs. Therefore, there is a scope for demanding additional clicks; for example, asking to confirm the willingness to share non-fact-checked information.
Figure 2 The effect of additional clicks required to share false news on social media
Taken together, the results of our randomised experiment with real Facebook sharing decisions deliver an optimistic message. Although the previous literature has shown that fact-checking cannot undo initial impressions left by false statements, our paper shows an important role that fact-checking plays in limiting propagation of false news. Fact-checking substantially reduces sharing of false information, whether users choose to view fact-checking information or are forced to see it. Furthermore, the very fact of being offered the choice to view the fact-checking reduces sharing of ‘alternative facts’ even when users do not exercise the option.
Barrera, O, S Guriev, E Henry and E Zhuravskaya (2020), “Facts, alternative facts, and fact checking in times of post-truth politics”, Journal of Public Economics, 182: 104–123.
Dustmann, C, B Eichengreen, S Otten, A Sapir, G Tabellini and G Zoega (2017), “Europe’s Trust Deficit: Causes and Remedies”, London: CEPR Press.
Henry, E and E Zhuravskaya and S Guriev (2020), “Checking and Sharing Alt-Facts” CEPR Discussion Paper 14378.
Nyhan, B, E Porter, J Reifler and T Wood (2019), “Taking Fact-Checks Literally But Not Seriously? The Effects of Journalistic Fact-Checking on Factual Beliefs and Candidate Favorability”, Political Behavior.
Swire, B, A Berinsky, S Lewandowsky and U Ecker (2017), “Processing political misinformation: comprehending the Trump phenomenon”, Royal Society Open Science, 4(3).
Vosoughi, S, D Roy and S Aral (2018), “The spread of true and false information online”, Science, 359: 1146–1151.
Zhuravskaya, E, M Petrova and R Enikolopov (2020), “Political Effects of the Internet and Social Media”, Annual Review of Economics, Forthcoming.