学生汇报，汇报人：胡诗云、张一灿、张澄

时间：2025年5月30日（周五）14：00-15：30

地点：北大国发院承泽园131教室

汇报人1：胡诗云 （导师：黄卓）

题目：《The Effects of Human-AI Collaboration on Human-Human Collaboration: A Network Perspective》

摘要： 

While the impact of human-AI interaction on individual and organizational performance has garnered significant attention, the influence of generative AI (GenAI) adoption on human-to-human workplace collaboration remains unexplored. Drawing on archival data from a technology company comprising 8,476,046 human-human collaboration records and detailed human-AI interactions from 2022 to 2023, we examine how GenAI shapes workplace social networks. Difference-in-differences estimates reveal that GenAI adoption significantly enhances users' network position, as measured by centrality and structural holes. These effects are more pronounced among employees with higher educational attainment, initially peripheral network positions, and non-computer-science backgrounds. Furthermore, GenAI adopters develop stronger ties with organizational leaders while simultaneously expanding their weak ties across diverse job roles. Further analyses reveal that GenAI adoption catalyzes both cross-functional collaboration and informal interactions. Our findings demonstrate GenAI's potential to foster a more connected, inclusive, and dynamic collaboration environment within organizations.

汇报人2：张一灿（导师：黄卓）

题目：《银行资产配置与宏观经济增长的反馈机制研究——基于逆周期资本缓冲政策的视角》

摘要：

宏观审慎政策主要关注在不同的经济周期阶段中如何有效控制各重要金融指标，在宏观层面保证经济增长的同时维持金融系统的稳定性。而作为最主要资本类宏观审慎政策的逆周期资本缓冲，我国至今还没有统一的标准。本文将三种不同的逆周期缓冲政策机制引入自创的宏观-金融反馈循环系统模型中，探究资本缓冲政策在系统遭受外生冲击时如何发挥作用，为我国相关的政策制定提供参考。

汇报人3：张澄 （导师：黄卓）

题目：《Inflation Forecasting with Panel Trees》

摘要：

     We use the panel tree (P-Tree) model to extract information from cross-sectional equity prices for inflation forecasting. P-Tree is a tree-based model tailored for analyzing panel data utilizing global split criteria that incorporate economic guidance to guard against overfitting while preserving interpretability. We benchmark its forecasting performance against that of Hong, Pan, and Tian (2023) (HPT), who uses the relative pricing between stocks with high- and low-inflation exposures. We find that P-Tree model's in-sample performance is better than HPT's. P-Tree model's out-of-sample performance is slightly better than or similar to HPT's. Moreover, using macroecnomomics variables as split variables can further improve out-of-sample performance of P-Tree model.