How does media bias contribute to social unrest? Theme leaders from CEPR's ReCIPE programme discuss the links between conflict and social media, as well as current efforts to forecast conflict occurrence.
In this episode of VoxDevTalks, host Tim Phillips is joined by Maria Petrova (Barcelona School of Economics) and Augustin Tapsoba (Toulouse School of Economics) to discuss the links between conflict, hate speech, and media—especially social media. The conversation is part of the Information and Conflict theme within CEPR’s ReCIPE programme, which aims to understand how conflict affects economic growth and public policy.
This episode focuses on how media bias, propaganda, and digital platforms contribute to social unrest, violence, and political polarisation.
What makes social media different from traditional media?
The discussion begins by examining how social media compares with traditional media. While some of the effects are similar—particularly the persuasive nature of content—social media enables faster, more decentralised communication.
“Sometimes internet and social media have persuasive effects on people's behaviour through [an] information channel which is similar to traditional media. For example… sometimes [for] horizontal flows of information and coordination, the role of social media becomes important.” Petrova
Access and quality of data are also barriers.
“Some interactions among users are not open to the general public, so it's difficult to get access to it. Content is more difficult to collect, as compared with traditional media.” Petrova
Why economists should collaborate across disciplines
The theme leaders also explore the role of economists in this evolving area of research. Petrova observes that economists are increasingly collaborating across disciplines.
“More recently, I've seen more and more collaborations with computer scientists and political scientists, who are typically very helpful with the data collection.” Petrova
These interdisciplinary partnerships are especially valuable in accessing and interpreting the vast, often unstructured datasets generated by digital platforms—data that may be difficult to collect or analyse using traditional economic methods alone. But economists also bring unique strengths to this field, particularly through theoretical frameworks and robust analytical techniques.
“Economists are also bringing a lot of theory into discipline... They are also bringing the toolbox, so we have all these causal inference methods that are now very common for economists that are also reaching our sciences.” Tapsoba
Does social media fuel protest and hate?
Social media, once seen as a force for unity and democratic engagement, increasingly appears to be a catalyst for division, protest, and even violence.
“There is evidence that social media promotes protests all over the world.” Petrova
This conclusion draws from a growing body of empirical research, including her own work on Russia, as well as studies covering Africa, US, and Germany.
One reason for this is the platform’s ability to enable real-time coordination and amplify emotionally charged or polarising content. Social media bypasses traditional editorial filters, allowing misinformation, hate speech, and calls to action to spread quickly and widely.
“A number of papers document a relationship between social media and hate crime.” Petrova
Media bubbles, polarisation, and violent ideologies
Another issue explored is whether media-driven polarisation leads to conflict.
“Our paper… actually suggests that theoretically, social media should increase polarisation, and we correspondingly observe an increase in xenophobia and hate crime in Russia.” Petrova
Yet not all media influence is negative.
“There is evidence that traditional media can promote post-conflict reconciliation if they broadcast particular type of content.” Petrova
Predicting outbreaks of conflict using media data
Tapsoba’s focus is on forecasting conflict, particularly through what he terms “the hard forecasting problem.” The aim is to predict violence in otherwise peaceful areas—a task made more feasible through the analysis of media.
“You can leverage all this information to figure out what are the key topics that people have been talking about, and use machine learning techniques to try to tell you whether there is any connection between certain topics and the occurrence of violence or not.” Tapsoba
Tapsoba points out that this research has found both positive and negative correlations between news topics and conflict.
“This has been very successful at identifying certain topics that are positively correlated with the onset of violence, but also topics that are interestingly even negatively correlated.” Tapsoba
He outlines two key policy applications.
You have a lot of policymakers... interested in having early warning systems that would tell them where is a high risk of violence, so that they can prepare and provide relief in a timely manner.... Another type of policymakers that are interested in these forecasting exercises is basically those that are big players on the international diplomacy landscape, that are able to negotiate or talk to different parties that are in a given conflict, so that they avoid escalations that can lead to full blown conflicts.” Tapsoba
Can social media regulation reduce harm?
A recurring question is whether more active moderation by tech companies would reduce harm. Petrova points to empirical studies suggesting it can.
“There is evidence that social media regulation indeed can reduce hate speech online.” Petrova
However, she also warns that excessive control, particularly by authoritarian regimes, can backfire. This was the case in China, where a ban on Instagram has led individuals to install VPNs and learn about the internet outside the firewall.
What research is still needed?
Looking ahead, Petrova and Tapsoba call for more research in new regions and on emerging platforms.
“I expect to see in these years to come, more research on social media… We also need to know more about the mechanisms, especially historical mechanisms, and on the countries outside of traditional focus of conflict literature.” Petrova
Tapsoba outlines key research areas, including the media’s effect on violence, the economic consequences of conflict, and forecasting.
“How these sort of affect their mental wellbeing, their productivity at work, the decisions to seek medical service or education in conflict prone environments, whether the perceived risk and all the information they’re getting on social media lead to them overreacting or being affected more than they should be by the underlying violence risk... I think that's a key thing that would be super useful to have.” Tapsoba
On forecasting, he adds:
“Can we leverage the richness of all this information to improve our ability to forecast or anticipate increases in violence risk in different areas with maybe more spatial granularity, rather than having country-level forecasting? Can we, at a very local level, figure out what’s the underlying risk in these areas?” Tapsoba