An eigenvalue distribution derived ‘Stability Measure’ for evaluating Minimum Variance portfolios
提出一种基于Marchenko-Pastur随机相关矩阵的理论稳定性度量，用于评估最小方差组合的稳定性，并在S&P 400等指数上验证其有效性，帮助研究者衡量估计风险变化并管理再平衡策略。
The Minimum Variance portfolio is subject to varying degrees of stability and robustness. We, therefore, propose a theoretical measure of its stability relative to a Marchenko–Pastur derived random correlation matrix. We demonstrate its practical use on the S&P 400, the S&P 500, the S&P 600 and the Russell 1000. Using historic market data from 2002 to 2021, we perform an optimisation on the empirical correlation matrix eigenvalue distribution to determine the implied variance ν(t) for the underlying data-generating process. Through monitoring its change over time Δν(t), we provide a Stability Measure for the Minimum Variance portfolio and thereby help researchers measure changes to estimation risk and manage rebalancing regimes.