Peer-Reviewed Publications
​​
Navigating Violence Between Surface and Below:
Effect of Protester Reaction to Violent Flanks on Mobilization
​
Journal of Global Security Studies
​
How do nonviolent protesters’ responses to violent flanks shape mobilization in unarmed movements? Existing research rarely centers on these responses, instead estimating a single net effect of violent flanks with mixed results. This article argues that the effects of responses hinge on fragmentation discrepancy, the gap between how unified a movement appears to outsiders and how cohesive it is internally. As responses to violent flanks (endorsement or opposition) operate as outward public signals and internal organizing devices, their effects on mobilization vary with the fragmentation discrepancy’s sign and size. Using NAVCO 2.1, ACLED, and an illustrative case of Hong Kong’s 2019 anti-extradition movement, I show that endorsement increases mobilization when movements look unified but are internally divided, whereas opposition is more effective when movements appear fragmented yet are internally cohesive. These results reconcile mixed findings on violent flanks and re-center protesters’ agency in explaining mobilization.
​
​
When Repression Surprises: Contextual Deviations and Protest Mobilization
​​
Research & Politics ​​
​
Many studies examine whether state repression deters or increases protest mobilization, yet empirical findings remain inconsistent. This study argues that mobilization depends less on the absolute level of repression than on how current repression deviates from what people have come to expect in a given context. It introduces discrepancy-based measures that capture these deviations relative to cumulative patterns of state violence, offering a context-sensitive alternative to conventional absolute measures such as fatality counts or arrests. Drawing on event-level data from ACLED, NAVCO, and SCAD, this article analyzes both country and local-level panels using two-way fixed effects regression and cumulative link mixed models. Across multiple datasets and specifications, the discrepancy measures consistently produce a U-shaped relationship: participation increases when repression is either higher or lower than established baselines. Comparisons with absolute measures show that the discrepancy approach more reliably explains variation in mobilization, helping to reconcile fragmented results in the repression–mobilization literature. These findings highlight the importance of expectations and contextual baselines in shaping how repression is interpreted and acted upon, suggesting that studies of contentious politics should move beyond static metrics toward measures that account for how actors evaluate state actions over time.
​​
Riding the Tide: How Online Activists Leverage Repression
Social Science Computer Review
​
How does repression reshape the way online activists engage with their target audiences? While existing research focuses largely on changes in the level of online participation after repression, it overlooks how activists strategically respond to maximize their impact. Addressing this gap, this article argues that repression prompts online activists to increase interactions with different groups, aiming to broaden their support base by signaling openness and inclusivity amid heightened public attention. Using permutation tests and ARIMA models within an Interrupted Time Series (ITS) framework, the study examines assortativity and the proportion of cross-group ties among Twitter users during the 2011 Occupy Wall Street movement. The findings support the argument, revealing that two key repressive events—the Brooklyn Bridge mass arrests on October 1 and the eviction threat of Zuccotti Park on October 13—significantly altered the communication patterns of online activists, leading to greater cross-group engagement in response to state repression.
​​​​​​​​​​​​​
​​​​​
Under Review​
​
Peripheral Urbanity and the Communicative Logic of Protests Violence
​​
This article reframes the puzzle of violent protest escalation, moving beyond the assumption that violence is often counterproductive. It asks: under what conditions does rational calculation lead protesters to adopt violent tactics? It argues that escalation is likely when state repression occurs in peripheral urban areas—regions far from national capitals but with dense communicative infrastructure. Here, emotional and strategic incentives intersect, as protesters seek to publicize marginalized grievances and ensure historical memory. Using ACLED and SCAD datasets, two-way fixed-effects logistic regressions demonstrate that repression in peripheral urban settings significantly increases the likelihood of violent escalation. A case study of the 1980 Gwangju Uprising in South Korea illustrates how protesters used violence to transform repression into a historical focal point for future mobilization. The study highlights the interplay of reason, emotion, and geography in protest dynamics, offering new insights into political communication under informational asymmetries.
​
Works in Progress
​​​​​
Protest Violence and Gender Equality: A Rationalist Approach
How does gender equality affect the prospect of protest violence? This article addresses gaps in the existing literature by emphasizing the rational incentives of women protesters to reduce protest violence rather than attributing their actions to biological or cultural factors. In societies with either low or high levels of gender equality, women’s costs of marginalization due to violence tend to be higher compared to those with intermediate levels. Consequently, in societies with either low or high gender equality, women protesters are more likely to stand at the frontlines to prevent violence, thereby reducing the probability of violence. Thus, protest violence is more likely to occur in societies with intermediate levels of gender equality. The argument is primarily supported by multiple datasets, including Women in Resistance (WiRe), Armed Conflict Location and Event Data (ACLED), and Social Conflict Analysis Data (SCAD).