北大经院工作坊第1083场
Marginal Fairness: from Fair Predictions to Fair Decisions
风险、保险与不确定性经济学工作坊
主讲人：Fei Huang (Associate professor at the University of New South Wales (UNSW）
主持老师：
（人大财金）陈泽
（北大经院）贾若
（清华经管）冯润桓
参与老师：
（人大财金）魏丽
（北大经院）郑伟
时间：2025年4月24日（周四）10：00-11：30
线上形式：腾讯会议
会议号：175 376 534
线下地点：中国人民大学立德楼503
主讲人简介：
Short bio: Fei is an Associate Professor in the School of Risk and Actuaries Studies, UNSW Business School. She received her BSc. in Mathematics from Xiamen University, MPhil in Actuarial Science from the University of Hong Kong, and PhD in Actuarial Studies from the Australian National University. Before joining UNSW in 2020, she was a senior lecturer at the Australian National University. Fei is a columnist writing Responsible Data Science series for Actuaries Digital - Actuaries Institute Magazine.
In her research, Fei focuses on responsible AI and data-driven decision-making, especially for the insurance industry, utilizing tools from statistics, machine learning, economics, and marketing. Her research interests include fair and non-discriminatory insurance pricing, algorithmic bias, interpretable machine learning, mortality modelling, and customer relationship management. She particularly explores ways to make insurance equitable, affordable, and sustainable in the contexts of AI and climate change. Her work has been published in leading actuarial journals and received many prestigious research awards, including the ABDC Innovation and Excellence Award for Research, Carol Dolan Actuaries Summit Prize, the ASTIN Colloquium Best Paper Award, the Actuaries Institute's Volunteer of the Year Award, and the UNSW Business School SDG Research Impact Award. Her research has been funded by multiple international and domestic institutions, including the Australian Research Council (Discovery Project: Dealing with Climate Disasters), Society of Actuaries, and Casualty Actuarial Society.
摘要：
This paper introduces Marginal Fairness, a new fairness notion for equitable decision-making in the presence of protected attributes such as gender, race, and religion. Traditional fairness notions primarily focus on Prediction Fairness, ensuring that machine learning models produce unbiased predictions. However, real-world decision-making often extends beyond pure machine learning tasks, particularly in financial and insurance applications. To address this, we model decision-making as a two-step process: (i) a machine learning model predicts a loss random variable, and (ii) a generalized distortion risk measure informs the final decision. We advocate for enforcing fairness in this second step—termed Decision Fairness—to ensure equitable outcomes. Furthermore, we introduce Marginal Fairness as a novel criterion that eliminates the sensitivity of decisions to protected attributes, regardless of whether they are continuous, discrete, univariate, or multivariate. By introducing Cascade Sensitivity, this approach ensures that dependencies between covariates are preserved when perturbing a protected attribute. Marginal fairness is broadly applicable to high-stakes decision-making contexts, including catastrophe insurance pricing, capital allocation, and financial credit and mortgage pricing. By shifting the fairness focus beyond prediction to risk-informed decisions, Marginal Fairness provides a robust and practical solution for achieving fairness in real-world applications.
北大经院工作坊第1084场
A Name is More Than Just a Name: Responses to the Japanese Name Policy in Colonial Korea, 1940-1945
经济史工作坊
主讲人：Duol Kim (Myongji University)
主持老师：
（北大经院）赵一泠
（清华大学）徐志浩
参与老师：
（北大经院）郝煜、管汉晖、周建波
（北大光华）颜色、李波
时间：2025年4月24日（周四）12:00-13:30
地点：北京大学经济学院301会议室
主讲人简介：
Duol Kim is a professor of economics at Myongji University. He received his Ph.D. in economics from UCLA. His research focuses on economic history (Korea and East Asia), and law and economics. He is a trustee of the Asian Historical Economics Society and the president of the Korean Economic History Society. He is also an associate editor of the Asian Journal of Law and Economics and previously served as an editor of the Review of Economic History (a Korean journal) and the guest editor of the Asian-Pacific Economic History Review Special Issue.
摘要：
We investigate people’s response to an assimilation policy by studying the Japanese Name Policy (1940) in colonial Korea. By analyzing the household registry of a rural district, we find that 12 percent of residents registered a new given name, a rate that is considered low. Moreover, males, wealthier individuals, those with higher levels of education, people from higher social classes, and urban dwellers were more likely to comply with the policy. This suggests that, individuals with higher abilities or those expecting greater benefits were more inclined to change their identities to align with the majority, or they may have faced greater pressure to do so. we could not find evidence that the Japanese name contributed to asset growth.
北大经院工作坊第1085场
Public Goods, Social Alternatives, and the Lindahl-VCG Relationship
微观理论经济学工作坊
主讲人：Claudio Mezzetti（Colin Clark Professor of Economics at the University of Queensland）
主持老师：
（北大经院）吴泽南、石凡奇
（北大国发院）胡岠
参与老师：
（北大经院）胡涛
（北大国发院）汪浩、邢亦青
（北大光华）翁翕、刘烁
时间：2025年4月24日（周四）10:00-11:30
地点：北京大学经济学院302会议室
主讲人简介：
Claudio Mezzetti (Oxford DPhil) is the Colin Clark Professor of Economics at the University of Queensland. Previous positions include: Professor of Economics, University of Melbourne, Leverhulme Professor of Industry and Organisation, University of Warwick, Professor of Economics, University of North Carolina, Chapel Hill. Mezzetti is a Fellow of the Econometric Society, and of the Society for the Advancement of Economic Theory. He is a microeconomic theorist whose main interests are in mechanism design, game theory and information economics. He has also worked in industrial organization, law and economics, international trade and public economics.
摘要：
Lindahl prices, set by a fictitious auctioneer with full knowledge of values and costs, are a generalization of Walrasian prices. By making the efficient allocation utility- and profit-maximizing for all participants, they induce an efficient outcome in a decentralized way even in the presence of public goods. We study a collective choice model with quasilinear utility, which encompasses the allocation of public and private goods as special cases. We show that each agent's smallest Lindahl price for the efficient alternative is equal to his VCG transfer while the firm's VCG transfer is equal to the largest sum of Lindahl prices. Thus, the VCG mechanism incurs a deficit if and only if the set of vectors of the agents' Lindahl prices for the efficient alternative is multi-valued. Unlike Walrasian prices, Lindahl prices are not restricted to be anonymous or linear. This is the reason why, when considering the allocation of private goods, the agents' smallest Walrasian payments are at least as large as their smallest Lindahl prices, and thus their VCG transfers. It is also why Lindahl prices always exist while Walrasian prices may not.
北大经院工作坊第1086场
Deep Autoencoders for Nonlinear Factor Models: Theory and Applications (非线性因子模型的深度自编码器：理论与应用)
计量、金融和大数据分析工作坊
主讲人：Dacheng Xiu（University of Chicago）
主持老师：（北大经院）王法
参与老师：（北大经院）王一鸣、王熙、刘蕴霆
时间：2025年4月25日（周五）10:00-11:30
地点：北京大学经济学院107会议室
主讲人简介：
Dacheng Xiu is the Joseph Sondheimer Professor of Econometrics and Statistics at Booth School of Business. His earlier research includes high-frequency financial data and econometric modeling of derivatives, and recently he focuses on machine learning methods for empirical asset pricing. His research has appeared in Econometrica, JPE, JF, RFS, JoE, JASA and AoS, and he serves as Co-Editor/Associate Editor for several journals, including JBES, JF, RFS, JoE and JASA.
摘要：
Autoencoders are neural networks widely used in unsupervised learning tasks such as dimensionality reduction and feature extraction. This paper establishes nonasymptotic guarantees for deep autoencoders within a nonlinear factor model, demonstrating their ability to effectively extract latent components with errors that diminish as dimensionality and sample size increase. The extracted factors are shown to converge to the true latent factors, up to a functional transformation. Furthermore, we extend these results to supervised autoencoders, providing theoretical support for their application in factor-augmented prediction and structured matrix completion. Finally, we showcase the practical utility of supervised autoencoders in macroeconomic forecasting and asset return prediction, and of autoencoders in noise reduction for causal analysis.
北大经院工作坊第1087场
Optimal Charging Infrastructure for Electric Vehicles Market
生态、环境与气候变化经济学工作坊
主讲人：王炳霖（新加坡国立大学助理教授）
主持老师：
（北大国发院）邢剑炜
（北大经院）季曦
时间：2025年4月25日（周五）10:30-12:00
地点： 北京大学国家发展研究院承泽园131教室
主讲人简介：
Binglin Wang is an Assistant Professor in the Department of Real Estate at the National University of Singapore (NUS) Business School. His research interests are at the intersection of urban, transportation, environmental, and energy economics, and empirical industrial organization. He earned his doctorate degree in applied economics at Cornell University.
摘要：
We develop an optimal charging network for electric vehicles (EVs) by estimating a two-sided market model of EV demand and charging station entry. Using granular data on EV sales and charging stations, we conduct policy simulations to find the socially optimal charging network, examine different cost-sharing ratios, and suggest subsidy policies that mimic the social optimum under a budget constraint. Our analysis informs the design of charging infrastructure to effectively promote EV adoption and improve social welfare.
供稿：科研与博士后办公室
美编：闻听
责编：度量、雨禾、雨田
