On Reference-Regulated Multiperiod Mean-Variance Portfolio Optimization in High Dimensions

Battle Count: 73

Highly relevant to quantitative trading. The paper directly addresses multiperiod portfolio construction, estimation risk mitigation, and out-of-sample performance in high-dimensional settings (p=280 stocks with n=120 observations). The RRMV framework provides actionable portfolio policies with closed-form solutions, practical regularization mechanisms, and demonstrated improvements in Sharpe ratios across industry and stock-level datasets. The cross-regularization insight (Q and w jointly affecting both mean and covariance estimation errors in dynamic settings) is novel and practically important for portfolio managers.

Moderate to high complexity. The core algorithm involves recursive computation of parameters (a_k, b_k, c_k, D_k) via Equation (4), which is straightforward. However, the high-dimensional asymptotic analysis requires solving fixed-point equations for s(rho) and tilde-s(rho) from random matrix theory. The rolling validation procedure for selecting rho adds computational overhead. The feedback policy structure requires wealth-dependent rebalancing at each period. For practitioners, the closed-form policy (Theorem 2) is implementable, but tuning Q_k and w_k requires domain expertise.

The paper provides detailed recursive algorithms (Equation 4), closed-form optimal policies (Theorems 1-2), and uses publicly available datasets (Ken French's data library for industry portfolios, CRSP for S&P 500 constituents). Rolling estimation scheme with n=120 months is clearly specified. However, no code repository is provided, and some numerical procedures (solving for s(rho), tilde-s(rho)) require custom implementation.