Western-style representative democracy is under pressure. Both US President Biden and EU-Commission President von der Leyen, as well as US Secretary of State Blinken and EU-Commission Vice President Vestager recently declared: “Democracies must deliver.” A key problem is the deteriorating trust of many voters in political institutions and the democratic system itself. Correlational evidence suggests that the information distributed via social media platforms contributes to this problem. Two technology-rooted characteristics of social media appear to be main drivers of the issues: political interest groups have the ability (i) to microtarget news based on individual-level voter data and (ii) to obfuscate their identities, which can be exploited to spread disinformation.

Social media platforms and legislators on both sides of the Atlantic have started to take action in order to mitigate the perceived negative effects. Specifically, discussions are about a ban on microtargeting and mandatory disclosure of politcial interest groups/advertisers’ interests. For instance, the EU’s Digital Services Act (DSA) mandates Very Large Online Platforms to fight disinformation: they have to deliver (public) annual risk-assessment reports and risk-mitigation reports, including annual audits by independent parties.

However, a solid empirical foundation for the suggested interventions is missing.

Together with Freek van Gils and Wieland Müller, in new working paper “Social Media and Democracy: Experimental Results” I develop, theoretically analyze and experimentally test a series of games to study the effects of two interventions, a ban on microtargeting and mandatory disclosure of senders’ interests, on individual voting behavior and election outcomes. The games are implemented in a laboratory setting that follows key features of a social media environment. This approach can both mimic the behavior of political interest groups and voters, in a stylized and framing-free environment, and thereby “look” behind the curtain of proprietary data of social media firms. It also suggests causal relationships that are in line with the evidence, which are necessary to inform policy implications.

We find that mandatory disclosure of interests, with or without a microtargeting ban, increases the efficiency of aggregate voter decision-making. However, only the combination of disclosure of interests and a microtargeting ban counteracts election manipulation. The implementation of a microtargeting ban without disclosure requirements has adverse effects. The latter result, in particular, should be of interest to legislators and regulators.

A short policy briefing is here.