Skip to main content Skip to secondary navigation
Main content start

Will Cong (Cornell): Building AI Models for Finance

Uncommon Factors and Asset Heterogeneity in the Time Series and Cross Section

Event Details:

Thursday, June 8, 2023
5:00pm - 6:00pm PDT


475 Via Ortega
Room 305
Stanford, CA 94305
United States

Will Cong (Cornell): Building AI Models for Finance

Uncommon Factors and Asset Heterogeneity in the Time Series and Cross Section

I provide a brief overview of how goal-oriented search as a core driver in the success of modern AI can be utilized to answer fundamental questions in finance. I briefly introduce deep reinforcement learning (e.g., Transformer-based RL) as a heuristic search for the solutions to portfolio management and managerial decision-making as stochastic control problems without pre-specified probabilities of state-transitions and rewards. I then focus on a new class of tree-based models as economically guided, goal-oriented greedy search for panel data analysis, with specific applications to clustering/sorting assets, providing basis portfolios, and constructing pricing kernels. In particular, in Cong, Feng, He, and Li (2023), we introduce the Bayesian Clustering Model (BCM), a novel and interpretable framework combining decision tree and Bayesian variable selection, to identify and model grouped heterogeneity in panel data, especially for asset returns. Utilizing marginal likelihood that accounts for parameter and model uncertainties, BCM detects time-series breaks using macroeconomic information and splits the cross section based on high-dimensional characteristics. We find strong evidence of structural breaks linked to market variance and valuation, and differential factor exposure and potential segmentation of assets primarily associated with idiosyncratic volatility, size, and value. We also identify MKTRF and SMB as common factors and multiple uncommon factors in different characteristics-managed clusters and macroeconomic regimes. BCM delivers outstanding asset pricing performance and informs the priceability of assets by well-established factors, achieving out-of-sample cross-sectional R2 exceeding 25% for some clusters. Moreover, a tangency portfolio built from leaf clusters delivers exceptional investment performance, including tripling the out-of-sample Sharpe ratio of that built from the Fama-French double-sorted portfolios.

Lin William Cong is the Rudd Family Professor of Management (endowed faculty chair by the Rudd Family Foundation) and a Tenured Professor of Finance at the Johnson Graduate School of Management at Cornell University SC Johnson College of Business. He is also the founding faculty director for the FinTech Initiative at Cornell and a research associate at the National Bureau of Economic Research. Prior to joining Cornell, he was an assistant professor of Finance and Ph.D. advisor at the University of Chicago Booth School of Business and faculty member at the Center for East Asian Studies. He is formerly a Kauffman Junior Faculty Fellow, a Poets & Quants World Best Business School Professor, a 2022 Top Quant Professor, a doctoral fellow at the Stanford Institute for Innovation in Developing Economies, and the George Shultz Scholar at the Stanford Institute for Economic Policy Research. Cong has served as a Finance Editor for the Management Science, and as associate editors for the Journal of Financial Intermediation, Journal of Corporate Finance, and the Journal of Banking and Finance, among other editorial roles. He is also a member of multiple professional organizations such as the American Economic Association, European Finance Association, and the Econometric Society. 
Cong researches on financial economics, information economics, FinTech and Economic Data Science, Entrepreneurship, and China. His academic interests include financial innovation, mechanism and information design, blockchains, cryptocurrencies, digital economy, real options, financial policy and markets in China, machine learning, AI, and alternative data. His recent work has focused on the intersection of technology, data science, and finance. His research has been featured in top academic journals and media such as Bloomberg, CNN, VOX, and Washington Post, and has been recognized with a number of accolades including over 40 conference best paper prizes and competitive grants. He has also been invited to speak and teach at hundreds of world-renowned universities, venture funds, technology firms, investment and trading shops, and government agencies such as IMF, Ant Financial, SEC, and federal reserve banks.
Cong has advised FinTech organizations and investment firms such as Wall Street Blockchain Alliance, Blackrock, Dfinity (incubator), and Chainlink, as well as government and regulatory agencies such as the New York State Department of Financial Services, Bank of Canada, and Asset Management Association of China. He has also been consulted regarding prominent FinTech litigation cases, by FBI agent and New York State Office of the Attorney General, among others, and was invited to advise regulators and federal prosecutors to fight against manipulations in cryptocurrency markets and frauds. 
Cong earned a Ph.D. in Finance and a MS in Statistics from Stanford University, where he served as the president of Ph.D. students association, received the Asian American Award for Graduate Leadership, and was recognized with the Lieberman Fellowship for outstanding contributions in research, teaching, and university service. He also holds dual degrees from Harvard University where he graduated summa cum laude and top in the Physics department, with an A.M. in Physics, an A.B. in Math & Physics, a minor in Economics.

Link to paper here

Related Topics

Explore More Events