Abstract:We extend the classical mean-variance (MV) framework and propose a robust and sparse portfolio selection model incorporating an ellipsoidal uncertainty set to reduce the impact of estimation errors and fixed transaction costs to penalize over-diversification. In the literature, the MV model under fixed transaction costs is referred to as the sparse or cardinality-constrained MV optimization, which is a mixed integer problem and is challenging to solve when the number of assets is large. We develop an efficient semismooth Newton-based proximal difference-of-convex algorithm to solve the proposed model and prove its convergence to at least a local minimizer with a locally linear convergence rate. We explore properties of the robust and sparse portfolio both analytically and numerically. In particular, we show that the MV optimization is indeed a robust procedure as long as an investor makes the proper choice on the risk-aversion coefficient. We contribute to the literature by proving that there is a one-to-one correspondence between the risk-aversion coefficient and the level of robustness. Moreover, we characterize how the number of traded assets changes with respect to the interaction between the level of uncertainty on model parameters and the magnitude of transaction cost.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the limitations of the classical mean - variance (MV) framework in practical applications, especially when dealing with estimation errors and fixed transaction costs. Specifically: 1. **Impact of Estimation Errors**: The classical MV model is very sensitive to the expected estimation of asset returns. Even a tiny estimation error may lead to significant changes in portfolio weights. To reduce this impact, this paper introduces an ellipsoidal uncertainty set to improve the robustness of the model. 2. **Problem of Over - Diversification**: The classical MV model tends to over - diversify portfolios, which may lead to the inclusion of some low - return or unnecessary assets, thus sacrificing the overall performance of the portfolio. To solve this problem, this paper punishes over - diversification by introducing fixed transaction costs, encouraging the construction of more concentrated and lower - management - cost portfolios. 3. **Computational Complexity**: When the number of assets is large, the MV optimization problem considering fixed transaction costs (i.e., sparse or cardinality - constrained MV optimization) is a mixed - integer problem and difficult to solve. For this reason, this paper proposes an efficient semi - smooth Newton - based proximal difference - of - convex algorithm (SN - pDCA) to solve large - scale robust sparse MV models and proves its convergence. ### Specific Problems and Solutions - **Combination of Robustness and Sparsity**: This paper proposes an extended classical MV framework - the Robust Sparse MV (RSMV) model. By introducing an ellipsoidal uncertainty set and fixed transaction costs, it reduces the impact of estimation errors and avoids over - diversification. \[ \text{RSMV} := \begin{cases} \min_{x \in C} \max_{r} \kappa x^T \Sigma x - r^T x + \phi^T 1(x) \\ \text{subject to } (r - \bar{r})^T \Omega^{-1}_{\bar{r}} (r - \bar{r}) \leq \varepsilon \end{cases} \] where: - \(x\) is the portfolio weight vector; - \(r\) is the worst - case return vector of assets; - \(\kappa\geq0\) is the risk - aversion coefficient; - \(\varepsilon\geq0\) is the uncertainty level; - \(\bar{r}\) and \(\Sigma\) are the estimated mean vector and covariance matrix of asset returns respectively; - \(\Omega_{\bar{r}}\) is the estimated error covariance matrix of \(\bar{r}\); - \(\phi\) is the fixed transaction cost vector; - \(1(x)\) is the indicator function vector, indicating whether a certain asset is held; - The constraint set \(C\) is defined as \(\{x \in \mathbb{R}^n : e^T x = 1\}\). - **Algorithm Development**: For the NP - hard problem of the RSMV model, this paper develops an efficient semi - smooth Newton - based proximal difference - of - convex algorithm (SN - pDCA) and proves its global convergence and local linear convergence rate through theoretical analysis. - **Numerical Experiments**: The effectiveness and computational efficiency of the SN - pDCA algorithm are verified through the Fama - French data set. The results show that this algorithm can provide high - quality solutions within a reasonable time and has better performance compared to other benchmark methods. In conclusion, this paper aims to improve the classical MV model by introducing robustness and sparsity, solve the challenges brought by estimation errors and fixed transaction costs, and provide effective algorithm support.

Robust and Sparse Portfolio Selection: Quantitative Insights and Efficient Algorithms

Asymptotic Behavior of Mean-Cvar Portfolio Selection Model under Nonparametric Framework

Portfolio selection under possibilistic mean-variance utility and a SMO algorithm

Sparse portfolio selection with uncertain probability distribution

Robust mean-risk portfolio optimization using machine learning-based trade-off parameter

An Empirical Study of Robust Mean-Variance Portfolios with Short Selling

Portfolio diversification and model uncertainty: a robust dynamic mean-variance approach

Portfolio Selection with Parameter and Model Uncertainty

A Robust Optimization Model For Project Portfolio Selection With Information Uncertainty

Robust equilibrium strategy for mean-variance-skewness portfolio selection problem

Robust portfolio asset allocation and risk measures

Robust Optimization Approaches for Portfolio Selection: A Computational and Comparative Analysis

Robust Portfolio Rebalancing with Cardinality and Diversification Constraints

Data-driven robust mean-CVaR portfolio selection under distribution ambiguity

Portfolio Optimization with Sparse Multivariate Modelling

Portfolio selection under model uncertainty: a penalized moment-based optimization approach

Autonomous Sparse Mean-CVaR Portfolio Optimization

Distributionally Robust Mean-Variance Portfolio Selection with Wasserstein Distances

Robust Portfolio Optimization with Multi-Factor Stochastic Volatility

Robust Portfolio Selection with Distributional Uncertainty and Integer Constraints

Robust portfolio selection based on asymmetric measures of variability of stock returns