Moving into Risky Floodplains: The Spatial Implications of Federal Disaster Relief Policies (with Xinle Pang) (Sep, 2025)
Journal of the European Economic Association, Revise and Resubmit
Abstract: This paper employs a quantitative dynamic spatial migration model to assess the welfare impacts of U.S. disaster relief policies in the context of flood risk. The model highlights the trade-off induced by mobility in the policy design. While relief policies provide insurance benefits to affected populations, they also inadvertently encourage sorting into flood-prone areas, generating a fiscal externality. Utilizing this model, we investigate the spatial economic consequences of relief efforts following Hurricane Harvey in Texas. To address the computational challenges posed by incorporating uncertainty into spatial framework, which results in a curse of dimensionality, we propose a novel solution method that leverages neural network. Our findings indicate that the current post-Harvey relief efforts have a positive effect on overall U.S. welfare when compared to a hypothetical scenario without such interventions. Furthermore, the paper explores alternative policies that yield even greater welfare benefits, such as floodplain taxation and moving subsidies.
The Philosopher’s Stone for Science – The Catalyst Change of AI for Scientific Creativity (with Qian Chen, Yi-Jen (Ian) Ho, and Dashun Wang) (Apr, 2026)
Abstract: This study examines whether and under what conditions artificial intelligence (AI), including machine learning and algorithms, is associated with scientific creativity. Drawing on the Logical Creative Thinking (LCT) framework, we conceptualize creativity as a structured search process governed by two mechanisms: recombination, reflected in novelty, and replacement, reflected in disruption. Using publication and citation-network data from more than 21 million papers published between 2000 and 2019, we identify AI-in-use research based on abstract/title keywords and compare AI-in-use and non-AI papers using established measures of novelty and disruption. We find that, on average, AI-in-use papers are associated with higher levels of both novelty and disruption. However, these relationships vary across levels of creativity. For novelty, AI-in-use is positively associated with both mid-tier and top-tier novel work, with stronger gains in mid-tier research. For disruption, AI-in-use is primarily positively associated with top-tier disruptive work, suggesting that AI may reinforce conventional trajectories in mid-tier research while enabling paradigm-shifting contributions when deployed by visionary researchers. We further find substantial heterogeneity across fields: the association between AI-in-use and both novelty and disruption is stronger in analytically oriented fields, such as Basic and Applied Science, and weaker—and at times negative—in more interpretive domains, such as Social Science and the Humanities. Finally, different AI tool categories exhibit distinct patterns: machine perception and data retrieval tools are more strongly associated with novelty, whereas pattern recognition and predictive modeling tools are more strongly associated with disruption. Together, these findings advance understanding of AI’s role in scientific knowledge production by showing how AI can support both exploratory recombination and transformative change, with substantial variation across contexts.
Energy Chokepoints and Production Networks: The Distributional Incidence of a Strait of Hormuz Blockade (with Jinhu Jia, Zhen Sun, Luotong Zhang and Houwang Zhao) (Jun, 2026)
Abstract: Energy security has moved to the center of economic policy as geopolitical conflict, supply disruptions, technological transformation, and the climate transition have repeatedly tested the resilience of energy systems. We quantify the global welfare incidence of a complete Strait of Hormuz blockade using a multi-country, multi-sector quantitative trade model with input-output linkages. Under this setting, real GDP changes can be decomposed into three channels: a final-goods channel, an intermediate-goods channel, and an input--output linkage channel. Our simulation results show that Gulf energy exporters bear the largest welfare losses, ranging from 2.09 percent of real GDP for the United Arab Emirates to 7.48 percent for Iraq. For energy-intensive importing economies, the input--output linkage channel is the dominant transmission mechanism: it accounts for nearly 90 percent of Japan's welfare loss and exceeds the total welfare loss of Korea. China and the United States, despite large bilateral Gulf energy flows, experience near-zero real GDP effects because diverse supplier relationships dampen all three channels. These results imply that relying on bilateral energy import shares alone would understate true Hormuz exposure for production-network-embedded economies, and that effective energy-security policy should also address the indirect channel that conventional strategic reserves cannot reach.