报告题目：Connected and Contagious: Unveiling Hidden Bank Vulnerabilities through AI-driven Co-Lending Networks

报 告 人：丘步晖

报告时间： 2025年05月14日（周三）14:00-15:30

报告地点：劝学楼505

主办单位：成人直播

【报告人简介】

丘步晖，悉尼大学商学院商业与金融学教授，于2009年获得辛辛那提大学金融学博士学位，中山大学会计学文学硕士学位、约翰霍普金斯大学-南京大学中美研究中心的研究生文凭，以及复旦大学物理学理学学士学位。在加入悉尼大学商学院之前，在鹿特丹管理学院（鹿特丹伊拉斯谟大学）担任教职。在追求学术生涯之前，在福特、普华永道咨询和IBM担任财务分析师和高级财务顾问。研究兴趣包括公司金融、会计和资本市场，在金融、会计、管理、国际商务和运营管理的主要期刊上发表了30多篇论文，其中许多发表在The Accounting Review, Management Science, Journal of Financial and Quantitative Analysis, Journal of International Business Studies 等FT50/UTD24期刊上。获得了多项国际研究奖项，包括两次芝加哥量化联盟（CQA）学术竞赛奖，研究被包括彭博社、经济学人、MoneyWeek 在内的各种媒体引用。

【内容摘要】

We develop a novel AI-driven Co-Lending Graph Neural Network (CoLGNN) model to capture risk spillovers in syndicated lending markets. Our approach integrates bank characteristics, loan attributes, and network topology to generate a robust spillover risk measure(CLN score) that predicts bank distress and profitability up to two years ahead. Leveraging natural experiments involving credit rating downgrades and the collapse of Lehman Brothers, we provide causal evidence of risk transmission across financial institutions. The CLN score significantly outperforms traditional risk metrics and standard network centrality measures, particularly for smaller, complex, and privately held banks. Further analysis reveals that revolving credit facilities primarily drive risk spillovers due to their intensive monitoring requirements. By uncovering previously hidden bank vulnerabilities, our deep learning framework offers regulators and market participants an effective early-warning tool for managing spillover risks in highly interconnected banking ecosystems.

撰稿：赵鹏辉  审核：史永东  单位：成人直播