War of attrition in the streets: Rethinking the dynamics of protest participation

Editors' note: This column is published in collaboration with the International Economic Association’s Women in Leadership in Economics initiative, which aims to enhance the role of women in economics through research, building partnerships, and amplifying voices.

Understanding the dynamics of protest participation has long been a focus of economics and social sciences (Yang and Yuchtman 2024). Seminal work by Granovetter (1978) and Kuran (1989) showed how individual decisions—based on differing thresholds for joining—can trigger cascading participation. More recently, Enikolopov et al. (2023) introduced a model where participation is shaped by social image concerns, assuming government behaviour as fixed, finding that protest participation tends to decline over time.

My research presents a new economic model that advances this literature by jointly modelling the strategic behaviour of both protesters and the government (Correa 2025). Grounded in game theory, it views protests not as static episodes but as dynamic contests with distinct phases—rising tensions and potential concessions. Methodologically, the model resembles a war of attrition (Hendricks et al. 1988, Gieczewski 2025), where both sides endure costs over time, each waiting for the other to yield.

In contrast to earlier models, this framework captures how protests often involve long periods of standoff, grow to a critical point, and then decline, not because the cause has weakened, but because the government strategically concedes and the perceived benefit of continued participation falls.

Protests are costly to everyone, but offer potential benefits for participants

In the model, a government faces a large mass of people who can decide, at any instant, to participate in a protest. Critically, protesting is costly for everyone. For the people, it takes time, energy, and often financial or physical risks. These costs vary widely: students may have more time but less money, wealthier individuals may have resources but less risk tolerance, while geographic distance affects accessibility.

The model assumes that everyone assigns the same value to the potential policy change, although this is a simplification. This allows differences in participation to emerge from differences in cost, effectively modelling each person's ‘propensity to protest’ as their net benefit from participating. Those with higher participation costs (and thus lower net benefits) are less likely to protest.

For the government, protests are disruptive: they can have economic consequences, become a focal point for media and international attention, and may damage a regime’s legitimacy. And these costs accumulate over time: a single protest event may be bearable, but a persistent movement becomes more costly with time, to the point of being unsustainable.

Yet despite these costs, protests also present potential gains. Once the government concedes, the people not only benefit from the policy changes but also derive psychological or social value from being part of something historically significant. In the model, this socially conferred reward is higher for those who participated for longer as they contributed more to the protest’s success.

Technically, the protest unfolds as a ‘war of attrition’—a strategic game in which both the government and protesters incur costs while waiting for the other side to give in. However, in contrast to a classic strategic interaction between two players, it is a game between a government and a large mass of citizens making individually optimal decisions.

The setting considered so far naturally restricts the type of protest events that the model aims to explain, restricting the scope of this paper to sustained, decentralised protest movements rather than one-off demonstrations or highly coordinated uprisings.

Result 1: Protests follow a predictable rhythm

The model identifies three phases in the dynamics of protests:

1. A build up, where people continuously join the movement. Participation increases and the government remains inactive, observant of its trajectory.
2. A peak, which is the highest point, where the protest becomes costly enough for the government to finally feel its pressure. This is the first time we can observe a response by the government.
3. A decay phase where, if the government has not conceded at the peak, the people begin to leave.

Individuals’ differing costs shape the trajectory of protest participation. Those with the lowest participation cost—such as the young or lower-income individuals—are the first ones to join and are potentially the ones who remain the longest. In contrast, the late joiners are the first to leave when the movement wanes. To illustrate, suppose the only cost of a protest is the opportunity cost of time. Under certain assumptions, lower-income individuals have a lower opportunity cost. Initially, then, participation is limited to people with lower incomes. As the movement grows and gains momentum, agents with a higher income join the movement because the expected gains now outweigh their high costs: the government is close to making concessions. At its peak, the protest reaches its most diverse point; people from very different socioeconomic backgrounds unite on the streets, and the magnitude of the protest forces the government to react. At this point, the government may concede to the people’s demands, but, if it resists the pressure, higher-income individuals start to leave, and the protest enters a decay phase where it continuously dies out. The government can still concede, but the likelihood of doing so decreases over time.

Result 2: Government concessions are non-trivial

Governments are most likely to concede during or shortly after the peak of protest participation, where they face high costs, participation is at its most diverse point, and some protesters with a very low propensity to protest have just joined. Since the protest becomes more costly with time, the government must act to stop the buildup. As time goes by, the government can still concede, but in equilibrium, the probability that the government will concede if the protest is still going decreases.

What do we learn about protests?

1. Participation has a single peak. If the underlying conditions remain unchanged, once a protest starts to die out, it will likely continue on this path.
2. The composition and demographics of the people in the streets change as a movement grows and dies out. At the peak, participation reaches the highest level of intensity and diversity. The people with the least propensity to participate join at the highest point for only an instant, but they help to precipitate concessions.
3. Concessions by the government happen late. Governments optimally let a movement grow before reacting, making them less likely to concede as the protest dies out.

In terms of empirical evidence, Cantoni et al. (2024) analyse daily protest data to study the geographic spread of durable protests and provide evidence consistent with my results. They find that in the days leading up to the largest protest event, the proportion of protesting cities rises steadily—from around 15–20% of the peak level 10 days prior to over 40% just before the peak. After the peak, protest activity remains elevated, slowly declining to pre-peak levels only after 20 days. This pattern holds across regime types, though autocracies and weak democracies tend to see a slightly faster post-peak decline. If we think of participation costs as heterogeneous across cities—even if individuals within cities are somehow heterogeneous—this slow build-up and decay closely mirrors the protest trajectory generated by the model.

My research offers academics precise and testable empirical predictions grounded in a game-theoretical model, while equipping policymakers with a deeper understanding of the dynamic incentives that fuel persistent social unrest.

References

Cantoni, D, A Kao, D Y Yang, and N Yuchtman (2024), “Protests”, Annual Review of Economics, 16.

Correa, S (2025), “Persistent protests”, American Economic Journal: Microeconomics (forthcoming).

Enikolopov, R, A Makarin, M Petrova, and L Polishchuk (2023), “Social image, networks, and protest participation”, Working Paper.

Gieczewski, G (2025), “Evolving wars of attrition”, Journal of Economic Theory, 224: 105967.

Granovetter, M (1978), “Threshold models of collective behavior”, American Journal of Sociology, 83(6): 1420–1443.

Hendricks, K, A Weiss, and C Wilson (1988), “The war of attrition in continuous time with complete information”, International Economic Review, 29(4): 663–680.

Kuran, T (1989), “Sparks and prairie fires: A theory of unanticipated political revolution”, Public Choice, 61(1): 41–74.

Yang, D and N Yuchtman (2024), “Why do protests matter? Exploring their causes and lasting impacts”, VoxDev.