Research
My research examines how political elites are selected, how hierarchical institutions shape their careers and behavior, and how political actors perform authority, conflict, and emotion. I study these questions across authoritarian, democratic, and international institutions. A related methodological agenda develops the data and measurement infrastructure required for comparative research, including longitudinal political biographies, auditable AI-assisted data construction, and multimodal analysis.
For datasets, measurement systems, and platform access, see the Data page.
Published work
Publications
“Portraits of Power: Facial Appearances and the Tacit Domain of Political Selection in China.”
Modern bureaucracies are commonly assumed to select personnel through rules, credentials, and performance. We argue that intuitive assessments of candidates' facial appearance can also enter political selection. Using an AI-based measure trained to reproduce human evaluations, we link perceived facial traits to the careers of more than 4,000 mid- and senior-level Chinese officials. Officials perceived as more competent and trustworthy and less aggressive have better promotion prospects and lower purge risk; warmth-related traits are especially valued in higher-level promotions and among male candidates. Conjoint experiments using real officials show that appearance-based impressions can be comparable in magnitude to performance and political connections. The findings identify a tacit domain of bureaucratic selection that complements meritocratic and network-based accounts.
“China and the Liberal International Order: A Pragmatic and Dynamic Approach.”
How should we understand China's relationship with the liberal international order? Drawing on Chinese academic debates and government statements, we argue that China and Western states often understand the postwar order differently: Chinese discourse emphasizes its functional institutions, whereas Western accounts more often connect the order to liberal political values. This perceptual gap fuels suspicion that China seeks to overturn the existing order. In practice, however, China's engagement has remained selective and pragmatic, varying across the order's institutional components. Although China has become more ambitious, domestic constraints, limited capacity, and the absence of a broadly attractive alternative continue to restrict the emergence of a distinct Chinese-led order.
Current submissions
Papers Under Review
“Agentic Framework for Political Biography Extraction.”
Producing large-scale political datasets requires researchers to recover structured facts from fragmented and unevenly documented sources. We develop a two-stage synthesis-and-coding framework in which LLM agents first search, filter, and organize source-linked evidence and then map the curated material into structured records. Across Chinese, American, and OECD political elites, we benchmark the system against human-curated and human-coded baselines. LLM coders match or exceed human performance when given curated inputs, while agentic synthesis improves coverage in open-web settings. Direct coding from long, multilingual corpora performs substantially worse, demonstrating that extraction quality depends on how evidence is selected and represented. The results shift attention from model choice alone to the architecture of the full data-production system.
“What Drives Legislative Anger? Multimodal Evidence from South Korea’s National Assembly.”
Legislative anger is often understood as expressive representation: a signal that legislators passionately defend constituent interests. We argue instead that anger is an instrument of partisan contest, especially when parties can blame one another for scandal or misconduct. Using deep-learning models, we measure vocal and facial anger across 355 hours of plenary debate in South Korea's 20th National Assembly, covering 11,060 speech segments by 247 legislators. Anger is highest in scandal-related debate. A quasi-experiment around the impeachment of President Park Geun-hye further shows that the same legislators become angrier in opposition and calmer in government. These findings challenge constituency-centered accounts and show how partisan blame and institutional position shape emotional expression in legislatures.
Research in progress
Working Papers
“The Disciplined Face of Power: Affective Signaling and Hierarchical Rule in China.”
This project develops a theory of bureaucratic affect: the informal norms that govern how officials display emotion within hierarchical organizations. Such norms may help political principals evaluate effort, reliability, and conformity when direct information is limited. I analyze an original dataset of more than 20,000 video appearances of Chinese officials to test whether political rank and institutional role systematically shape calmness, stress, and other forms of facial affect, and whether conformity to these display norms predicts career trajectories. By focusing on dynamic behavior rather than static credentials or formal rules, the project identifies a behavioral dimension of hierarchical governance and elite control.
“Winners Take All: Performance Targets, Bureaucratic Momentum, and Stratified Economic Growth in China.”
Target-based governance is a central feature of state capacity in China and beyond, yet research has focused mainly on the immediate incentives created by targets. This project examines whether barely meeting a GDP growth target changes a locality's subsequent development trajectory. Using city-level data from 2002 to 2023 and a regression-discontinuity design around the target threshold, I study whether target fulfillment generates material and cognitive momentum through greater investment, more ambitious future targets, and changes in how officials explain success and failure. The project connects performance management to cumulative advantage, asking how small differences in bureaucratic evaluation may produce increasingly unequal local growth paths.