Whence LASSO? A Rational Interpretation

This paper develops an economic setting to rationalize the use of the “least absolute shrinkage and selection operator” (LASSO) in estimating asset returns. In this setting, multiple traders engage in trading based on information extracted from historical asset prices. Facing model uncertainty in forecasting asset returns, these traders adopt robust-trading strategies. Within this context, the use of LASSO for estimating asset returns emerges endogenously as an equilibrium outcome. We further extend our analysis to rationalize the application of elastic-net estimation. Although LASSO-type strategies enhance traders’ profits by mitigating competition among them, they also introduce biases in trading decisions, which can adversely affect profitability. This dual effect highlights the nuanced tradeoffs associated with employing such estimation techniques in financial markets.

This paper has been This paper was accepted by Will Cong for the Special Issue on AI for Finance and Business Decisions.

Funding: This work was supported by the Social Sciences and Humanities Research Council of Canada [Grant 435-2021-0040] and the Bank of Canada [Fellowship].

Supplemental Material: The data files are available at https://doi.org/10.1287/mnsc.2024.06127.