State development theories posit that fragmentation lets subjects exit to other authorities, spurring jurisdictional competition that drives institutional development, yet behavioral evidence remains scarce. I develop a theory of exit under horizontal and vertical authority and test it in Tokugawa Japan’s distinctive setting. Across nearly 300 domains, villages sought concessions from home rulers by fleeing en masse or appealing through neighboring lords or to the central authority. I use domain conversions to central rule as a natural experiment: they brought central authority closer to neighboring villages. Difference-in-differences analysis finds 3.9 percentage points more exit among nearby villages under other lords, without detectable changes in confrontation. Grid-year panel data show proximity to another ruler’s capital predicted more exit but not confrontation. Proximity to neighboring capitals predicted lower reported tax rates where neighboring domains could absorb more migrants. These findings suggest fragmentation helps translate overlapping jurisdictions into popular checks on elite power.
Generative AI lets influence operations fabricate apparent widespread support cheaply. I develop a theoretical framework for online forums involving politically restricted information to isolate a tradeoff between authenticity and accuracy that generative AI exacerbates. Citizens collectively choose verification, then decide among anonymous posting, verified posting, and silence. When attackers can flood anonymous forums, citizens adopt the least verification that deters them. Because greater access raises knowledge and exposure, verification exposes better-informed sources to greater punishment risk and, unlike in canonical screening models, excludes the best-informed, hence most strongly motivated to contribute, first. Citizens value collective accuracy, but individual contributions remain externalities; private exposure risk therefore makes the forum more authentic but less informative, allowing the influence operation to deplete collective knowledge without deployment or deception. Detection and per-post costs reduce reliance on identity verification. Evaluations of epistemic security should track both the fakes excluded and the composition of who remains.
Modern military technologies, including automated weapons and AI-assisted systems, are thought to strengthen deterrence and preserve peace by enhancing limited military operations without provoking wider war. I challenge this optimism by examining their implications for coercive bargaining rather than their direct military effects. I develop a coercive bargaining model with a subwar capability: access to limited operations short of general war. Such operations can replace war after resistance or settlement after compliance, potentially weakening coercive threats and assurances. The target may concede less, further weakening assurances and the government’s ability to secure concessions. Across broad parameter regions, I show that subwar capability can lower coercive ability and increase war, even when no operation is used. Subwar capability can weaken the effect of military power on coercive ability and reduce social welfare by raising conflict costs. More efficient military technology can thus disrupt coercive bargaining and raise the cost of conflict.
Political actors can fail to secure concessions despite credible threats when targets fear that yielding could invite further harm. We show that this assurance problem can complicate signaling through coercive demands, making it harder for targets to distinguish restraint from plans to attack. Our bargaining model allows concessions to improve the demanding state’s military position before it decides whether to attack. This gives states planning an attack a reason to imitate the modest demands of states seeking peace. Greater military strength can make reassuring demand patterns harder to sustain and permit additional ways to conceal an intended attack. These changes can reduce the concessions the demanding state obtains peacefully and increase the risk of war, imposing collective costs when disputes that could have been settled peacefully instead lead to fighting. Lack of assurance can exacerbate information problems in interstate politics, undermining coercive bargaining and making peace harder to secure.