Growing the efficient frontier on panel trees

Lin William Cong; Guanhao Feng; Jingyu He; Xin He

doi:10.1016/j.jfineco.2025.104024

Growing the efficient frontier on panel trees

Lin William Cong, Guanhao Feng^*, Jingyu He, Xin He

^*Corresponding author for this work

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

5 Citations (Scopus)

44 Downloads (CityUHK Scholars)

Abstract

We introduce a new class of tree-based models, P-Trees, for analyzing (unbalanced) panel of individual asset returns, generalizing high-dimensional sorting with economic guidance and interpretability. Under the mean–variance efficient framework, P-Trees construct test assets that significantly advance the efficient frontier compared to commonly used test assets, with alphas unexplained by benchmark pricing models. P-Tree tangency portfolios also constitute traded factors, recovering the pricing kernel and outperforming popular observable and latent factor models for investments and cross-sectional pricing. Finally, P-Trees capture the complexity of asset returns with sparsity, achieving out-of-sample Sharpe ratios close to those attained only by over-parameterized large models. © 2025 The Authors.

Original language	English
Article number	104024
Journal	Journal of Financial Economics
Volume	167
Online published	18 Feb 2025
DOIs	https://doi.org/10.1016/j.jfineco.2025.104024
Publication status	Published - May 2025

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Funding

ow, Ron Kaniel, Chris Malloy, Alberto Rossi (discussant), Gideon Saar, Shrihari Santosh (discussant), Artem Streltsov, Ziying Sun (discussant), Peixuan Yuan (discussant), and Guofu Zhou for detailed comments and suggestions. We also thank Doron Avramov, Jie Cao, Zhanhui Chen, Eric Ghysels, P. Richard Hahn, Dashan Huang, Sophia Zhengzi Li, Asaf Manela, Stefan Nagel, Adam Reed, Artem Streltsov, Yinan Su, Fred Sun, Yaki Tsaig, Mao Ye, Dacheng Xiu, Bobby Yu, and seminar and conference participants at ArrowStreet Capital, 2024 AsianFA Annual Meeting, Baylor University, BUAA, Ca’ Foscari University of Venice, Cambridge University Algorithmic Trading Society Quant Conference, CFTRC 2022, CityUHK, CKGSB, Conference on FinTech, Innovation and Development at Hangzhou (2nd), Cornell University, 2024 EFMA Annual Meeting, Goethe University Frankfurt, GSU-RFS FinTech Conference 2022, HKAIFT-Columbia joint seminar, Hunan University, INFORMS Annual Meeting 2021, 4th International FinTech Research Forum (RUC), 2023 Mid-South DATA Conference (Memphis), KAIST Digital Finance Conference, NFA 2022, University of Oxford SML-Fin Seminar, OSU, PHBS, PKU Guanghua, PKU-NSD, PKU-NUS Annual International Conference on Quantitative Finance and Economics, Qube Research and Technology, Reichman University (IDC Herzliya), Schroders Investments, Shanghai Financial Forefront Symposium (3rd), SHUFE, NUS, SMU, SUSTech, TAMU, Tsinghua SEM, UNC, University of Bath, University of Hawai’i, USC, USTC, University of Macau, World Online Seminars on Machine Learning in Finance, WFA 2022, and 2023 XJTLU AI and Big Data in Accounting and Finance Conference for constructive discussions and feedback. Feng’s research is partly supported by the Hong Kong Research Grants Council (GRF-11502721, GRF-11502023) and the National Natural Science Foundation of China (NSFC-72203190). He J.’s research is partly supported by the Hong Kong Research Grants Council (ECS-21504921, GRF-11507022, GRF11509224) and the National Natural Science Foundation of China (NSFC-72403214).

Research Keywords

Decision tree
Factors
Generative models
Interpretable AI
Test assets

Publisher's Copyright Statement

This full text is made available under CC-BY 4.0. https://creativecommons.org/licenses/by/4.0/

RGC Funding Information

RGC-funded

Access Output

10.1016/j.jfineco.2025.104024

273323288.pdfFinal Published version, 2.39 MBLicence: CC BY

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{7983f9d64d824c7c848632390ebf1394,

title = "Growing the efficient frontier on panel trees",

abstract = "We introduce a new class of tree-based models, P-Trees, for analyzing (unbalanced) panel of individual asset returns, generalizing high-dimensional sorting with economic guidance and interpretability. Under the mean–variance efficient framework, P-Trees construct test assets that significantly advance the efficient frontier compared to commonly used test assets, with alphas unexplained by benchmark pricing models. P-Tree tangency portfolios also constitute traded factors, recovering the pricing kernel and outperforming popular observable and latent factor models for investments and cross-sectional pricing. Finally, P-Trees capture the complexity of asset returns with sparsity, achieving out-of-sample Sharpe ratios close to those attained only by over-parameterized large models. {\textcopyright} 2025 The Authors.",

keywords = "Decision tree, Factors, Generative models, Interpretable AI, Test assets",

author = "Cong, {Lin William} and Guanhao Feng and Jingyu He and Xin He",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.",

year = "2025",

month = may,

doi = "10.1016/j.jfineco.2025.104024",

language = "English",

volume = "167",

journal = "Journal of Financial Economics",

issn = "0304-405X",

publisher = "Elsevier B.V.",

}

TY - JOUR

T1 - Growing the efficient frontier on panel trees

AU - Cong, Lin William

AU - Feng, Guanhao

AU - He, Jingyu

AU - He, Xin

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2025/5

Y1 - 2025/5

N2 - We introduce a new class of tree-based models, P-Trees, for analyzing (unbalanced) panel of individual asset returns, generalizing high-dimensional sorting with economic guidance and interpretability. Under the mean–variance efficient framework, P-Trees construct test assets that significantly advance the efficient frontier compared to commonly used test assets, with alphas unexplained by benchmark pricing models. P-Tree tangency portfolios also constitute traded factors, recovering the pricing kernel and outperforming popular observable and latent factor models for investments and cross-sectional pricing. Finally, P-Trees capture the complexity of asset returns with sparsity, achieving out-of-sample Sharpe ratios close to those attained only by over-parameterized large models. © 2025 The Authors.

AB - We introduce a new class of tree-based models, P-Trees, for analyzing (unbalanced) panel of individual asset returns, generalizing high-dimensional sorting with economic guidance and interpretability. Under the mean–variance efficient framework, P-Trees construct test assets that significantly advance the efficient frontier compared to commonly used test assets, with alphas unexplained by benchmark pricing models. P-Tree tangency portfolios also constitute traded factors, recovering the pricing kernel and outperforming popular observable and latent factor models for investments and cross-sectional pricing. Finally, P-Trees capture the complexity of asset returns with sparsity, achieving out-of-sample Sharpe ratios close to those attained only by over-parameterized large models. © 2025 The Authors.

KW - Decision tree

KW - Factors

KW - Generative models

KW - Interpretable AI

KW - Test assets

UR - http://www.scopus.com/inward/record.url?scp=85217919680&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85217919680&origin=recordpage

U2 - 10.1016/j.jfineco.2025.104024

DO - 10.1016/j.jfineco.2025.104024

M3 - RGC 21 - Publication in refereed journal

SN - 0304-405X

VL - 167

JO - Journal of Financial Economics

JF - Journal of Financial Economics

M1 - 104024

ER -

Growing the efficient frontier on panel trees

Abstract

Bibliographical note

Funding

Research Keywords

Publisher's Copyright Statement

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

GRF: Are Asset Pricing Models Sparse?

GRF: Time-Varying Coefficient Modeling for Factor Selection in Asset Pricing

GRF: Regression Tree for Portfolio Optimization and Imbalanced Data

Cite this

Growing the efficient frontier on panel trees

Abstract

Bibliographical note

Funding

Research Keywords

Publisher's Copyright Statement

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Projects

GRF: Are Asset Pricing Models Sparse?

GRF: Time-Varying Coefficient Modeling for Factor Selection in Asset Pricing

GRF: Regression Tree for Portfolio Optimization and Imbalanced Data

Cite this