High-Dimensional Portfolio Selection with Cardinality Constraints

The expanding number of assets offers more opportunities for investors but poses new challenges modern portfolio management (PM). As a central plank PM, selection by expected utility maximization (EUM) faces uncontrollable estimation and optimization errors in ultrahigh-dimensional scenarios. Past strategies high-dimensional PM mainly concern only large-cap companies select many stocks, making impractical. We propose sample-average-approximation-based strategy to tackle the difficulties above with cardinality constraints. Our bypasses mean covariance, Chinese walls Empirical results on S&P 500 Russell 2000 show that an appropriate carefully chosen leads better out-of-sample mean-variance efficiency. On 2000, our best profits as much equally weighted reduces maximum drawdown average 10% 90%, respectively. flexibility stability incorporating factor signals augmenting performances are also demonstrated. balances tradeoff among return, risk, Therefore, we provide theoretically sound computationally efficient make practical growing global financial market. Supplementary materials this article available online.