This is the 28th in this year’s series of posts by PhD students on the job market.
Over the past two decades, the internet and social media have expanded rapidly to all corners of the world. While these new technologies have liberalized access to information and communication channels, they have also introduced new platforms for surveillance and propaganda. Existing evidence has shown how the expansion of internet access generates more political resistance in some cases (Enikolopov et al. 2020), while impeding these processes in others (Shapiro and Siegel 2015). As such, the internet can be characterized as a “double-edged sword” for society, introducing new freedoms as well as oppressions. In my job market paper, I explore this duality and ask whether we should understand social media as fundamentally a technology of liberation or control.
Reading international headlines from the last 15 years could easily convince the reader of either conclusion. The Arab Spring of the early 2010s was seen as a watershed moment for democracy, and subsequent protest movements in places like Hong Kong and Iran were also widely seen as being fueled by digital dissent on social media. However, media coverage of China’s “Great Firewall,” and stories about election interference, propaganda, and hate speech depict social media as a technology that is more likely to undermine democratic processes.
Internet shutdowns: a natural experiment in digital repression
I explore this question empirically using evidence from Myanmar, a country that was only introduced to widespread internet access within the past decade, after liberalization of the telecommunications sector led to rapid expansion of mobile data networks. In 2021, after Myanmar’s military staged a sudden and unexpected coup, Facebook (the country’s dominant social media platform) became a hub of mass mobilization. Online activists spoke out against the undemocratic takeover, shared information about protest events, and documented military abuses. Six weeks after the coup, when the regime ordered telecom operators to block all mobile data traffic nationwide, much of this digital resistance activity disappeared overnight.
This episode offers a unique opportunity to study the internet’s causal effect on political conflict. By comparing patterns of protest and violence before and after the shutdown, and by exploiting geographic variation in mobile internet access, I estimate how a negative internet shock influenced both the frequency and forms of political conflict. The key insight is that, even though the blackout was nationwide, its intensity of impact varied across space, since areas with higher levels of pre-existing internet access are more exposed to the internet, and thus should be more affected by the outage.
Mapping internet access from the ground up
To measure local internet exposure, I generate a high-resolution map of mobile signal strength across Myanmar using a radio wave propagation algorithm known as the Irregular Terrain Model (see e.g., Gonzalez and Maffioli 2024). By comparing these granular estimates against a standard threshold for data transmission, I can estimate—for each township—the number of people living in areas of reliable internet coverage. I then merge these estimates with data on political conflict from the Armed Conflict Location and Event Data Project (ACLED)—which tracks detailed information on both peaceful protests and violent political conflict—and geo-located data from over 17 million Facebook posts. Using a simple natural language processing (NLP) classifier, I can also identify which of these posts are related to the political crisis.
Technology of liberation
When the internet shutdown first took effect, both Facebook activity and protest frequency dropped sharply. The number of daily Facebook posts fell by more than half, and protest events declined by about 25 percent. Using a Difference-in-differences framework, I estimate that for every additional 10,000 people connected to the internet in a given township, the shutdown reduced the number of Facebook posts by an additional 2-3 percent and the number of protests by an additional 8-9 percent in the weeks that followed. However, I find no corresponding effect on violent events, such as armed clashes or attacks on civilians (see Figure 1). These results suggest that the internet, and social media in particular, plays a critical role in enabling mass mobilization, as it helps citizens share information and organize collective action. This evidence supports the view of the internet as a “liberation technology” (Diamond and Plattner 2012; Howard and Hussain 2013), i.e., a tool that facilitates resistance against authoritarian rule by lowering the cost of mass communication.
Figure 1. Event Study Results – Demonstrations vs. Political Violence
When liberation turns to control
However, the narrative changed as the military began to consolidate control over digital spaces. As the coup and ensuing political crisis unfolded, Myanmar’s military leadership increased the scope of its online surveillance efforts. Citizens were arrested in ever increasing numbers for posting political content on Facebook, and pro-military actors flooded Facebook with propaganda and threats. In the Facebook data, I observe a significant reduction in the proportion of all posts that are political, and an increase in the proportion of political posts that contain pro-military language. The likely explanation is self-censorship: as resistance members recognize the risks of online dissent, they avoid political discussion altogether.
During this time, the military continued to impose periodic internet shutdowns, allowing me to observe whether the effect of internet access changed as the military expanded its control over the network. When I estimate the effects of these later shutdowns (via a similar DiD approach), I still see a strong first stage, i.e., Facebook activity declines more in areas with high access, but the effect on demonstration activity is no longer significant. In other words, as internet freedoms were eroded, the relationship between access and protest disappeared.
This change illustrates the internet’s double-edged nature. In environments of relatively free online expression, the internet empowers civil society, but if the government adapts—by monitoring, censoring, and manipulating online platforms—the pro-democracy function of this technology can be effectively neutralized. As other scholars have noted, information and communication technologies can both lower the costs of collective action and enhance state surveillance capacity (Gohdes 2015; King et al. 2017; Roberts 2018; Weidmann and Rød 2019). The direction of the effect depends critically on the willingness and capacity of the government to control the use of online spaces.
In Figure 2, I illustrate this pattern using cross-country data. Among countries that enjoy relative internet freedom, there is a positive relationship between internet access and protest activity, but I observe no correlation across countries in which internet freedoms have been curtailed.
Figure 2. The Effect of Internet Access – Cross-Country Comparison
Implications for democracy and development
My findings hold several lessons for policymakers in developing economies where internet access is rapidly expanding. First, expanding connectivity can indeed foster civic engagement and democratic mobilization, but only when accompanied by institutional protections of internet freedom—such as the freedom of expression and universal access rights. In fragile or transitional states, new internet users may find that internet access is more of a boon to government surveillance systems than political expression.
Second, intentional internet shutdowns—which have become increasingly common worldwide (see recent examples in Afghanistan and Iran)—carry significant social and political costs. Governments often justify them as security measures to prevent unrest, yet our results show that shutdowns are effective mainly in suppressing peaceful protest, not violence. In the long run, repeated shutdowns can deepen authoritarian control and erode public trust in institutions. According to recent reports, nearly 300 shutdowns were documented across 54 countries in 2024 alone (Rosson et al. 2025), the most ever recorded in a single year. Understanding their true effects is crucial for designing international responses and digital governance norms.
Third, researchers and policymakers should recognize that the internet’s impact is dynamic. Its influence evolves as both citizens and governments learn. The same online platforms that once galvanized resistance in the Arab Spring (Howard and Hussain 2011) can later be weaponized for disinformation and repression. Strengthening digital literacy, transparency, and accountability in platform governance is thus essential to preserve the internet’s democratic potential.
Alexander Fertig is PhD candidate in Economics & Public Policy at the University of Michigan.
Join the Conversation