@inproceedings{ETFflow,
author = {Petit, Nicolas and Cucuringu, Mihai and Cartea, \'{A}lvaro},
title = {Data-Driven Trade Flow Decomposition for Exchange-Traded Funds and their Constituents},
year = {2025},
isbn = {9798400722202},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3768292.3770434},
doi = {10.1145/3768292.3770434},
abstract = {This paper presents a novel data-driven methodology for decomposing trade flow based on the co-occurrence patterns between Exchange-Traded Funds (ETFs) and their constituent securities. Lu et al. [19, 29] use a fixed set of rules for trade clustering to demonstrate the economic utility of trade flow decomposition. In contrast, we propose a purely data-driven approach. We employ unsupervised learning techniques with a set of features that capture the joint informational content between ETFs and their underlying constituents. Clustering is performed using principal component analysis for dimensionality reduction followed by k-means++ for clustering. We demonstrate the economic utility of this decomposition with an application to portfolio trading. We construct factor-mimicking portfolios based on the conditional order imbalances computed within each cluster, orthogonalised to a baseline set of common risk factors. We use data from SPY and its constituents between January 1, 2018 and December 31, 2021 to test our methodology and find that a subset of factor-mimicking portfolios achieve statistically significant Sharpe ratios. These results demonstrate that our data-driven trade flow decomposition captures economically meaningful trading patterns, and that the conditional order imbalances capture information beyond what is available from unconditional order imbalance.},
booktitle = {Proceedings of the 6th ACM International Conference on AI in Finance},
pages = {569--577},
numpages = {9},
keywords = {Exchange-traded funds, portfolio management, clustering, time series analysis, forecasting.},
location = {
},
series = {ICAIF '25}
}