State violence inflicts obvious direct costs on its victims, but many of its significant consequences may be indirect. Building on the literature on agency problems in authoritarian regimes, we argue that coercion against bureaucrats motivates them to pursue over-zealous goals at the expense of wider social costs. We test this argument by studying how bureaucratic purges in China under Mao impacted the behavior of local bureaucrats during the Great Leap Forward, a campaign that caused over 30 million deaths from mass starvation. Exploiting variations in purge intensity across about 1,400 counties, we find the purge intimidated local bureaucrats into inflating agricultural production and extracting excessive amounts of grain from farmers, which resulted in significantly higher famine mortality. The results highlight the downstream perils of "accountability by violence" in autocratic regimes.

Existing literature shows that political oversight of bureaucrats can improve government responsiveness to citizens. Yet when the state is mainly concerned with extracting resources from society, increasing oversight may result in social loss. I test this argument in the context of the Great Chinese Famine, a state-directed tragedy that killed over 30 million people. Using new county level data, I find that weather shocks, which increased central-local information asymmetry about local grain production levels, allowed local officials to relax the execution of excessive grain extraction orders from above. Despite production decline, weather shocks led to higher grain retention in rural communities. Consequently, the famine mortality rate was significantly lower in counties that experienced weather shocks. This effect was present only when a county party secretary was socially proximate to the residents. The findings highlight the perils of bureaucratic accountability to political leaders in autocratic states.

The case of China has inspired a large body of literature on the meritocratic model of bureaucracy. Drawing on the positive correlation between local officials' promotion outcomes and self-reported economic growth, the literature indicates that a merit-based promotion tournament drives bureaucratic selection in China. Because self-reported GDP growth data are commonly inflated by their producers, however, it remains unclear whether the promotion tournament rewards performance or cheating. We answer this question by developing a novel approach to recalibrating local GDP growth data using machine learning algorithms. Focusing on a panel of city party secretaries between 2000 and 2013, our analysis shows that promotion was positively correlated with overreported GDP growth. No significant correlation is detected between promotion and real GDP growth. Lower-performing officials were more prone to overreport GDP growth and received higher career returns from that behavior. The findings highlight the limits of bureaucratic selection in authoritarian states.

Civilian victimization by armed groups is a prevalent byproduct of war, but does it have wider political consequences beyond the ongoing conflict? We argue that civilian victimization by a foreign military ingrains memories of victimization among the affected population, which facilitates military mobilization in subsequent international conflicts. To test this argument, we examine the impact of atrocities committed by the Japanese army against Chinese civilians during the Sino-Japanese War (1937-1945). Drawing on over 3,300 geo-located cases of aerial bombings and civilian killings, we find that these atrocities by the foreign force facilitated the Chinese state's mobilization of civilians to participate in the later Korean War (1950-1953). In counties where civilians had been targeted with violence by the Japanese army, a higher number of people volunteered to enlist to fight in the Korean War. This effect was larger in counties with higher exposure to propaganda politicizing the collective experiences of victimization. 

While many rebel groups take actions to govern, there has been little research investigating the military consequences of rebel governance. We argue that exclusionary rebel governance, which rewards a narrow population of supporters at the expense of broader civilians, can negatively affect mass military mobilization by rebel groups. We test this claim with archival data from the Chinese Civil War (1946-1949), when the Chinese Communist Party carried out radical land reform in occupied areas. We digitize a new dataset of 260,000 Communist soldiers to estimate how land reform affected the Communist Party's soldier recruitment. Comparing county-level soldier recruitment before and after land reform with a difference-in-differences design, we document a substantively large and negative effect of land reform on soldier recruitment. This effect was likely driven by the extensive victimization of local population in coercive redistribution. Our finding highlights the countervailing impact of exclusionary and coercive rule.

The conventional wisdom says that representation pacifies politics by replacing bullets with ballots. But this argument overlooks the downstream political cost of representation for rulers: the institutionalized congregation enables the representatives to exercise their collective power, making the threat of a revolt more credible and raising the risk of revolution. Using the case of China at the end of the Qing dynasty, we show how territorial representation increased the corporate power of the elite and engendered the revolution. The counties with more powerful elites obtained better representation in the newly formed provincial assemblies in 1909, and later counties with more representation facilitated the fall of the imperial order during the 1911 Revolution. Representative institutions involve a trade-off between effective administration and the risk of political conflict, which may explain why the adoption of such institutions is not universal despite their purported benefits.