学海拾珠系列之二百六十七：多因子视角下的波动率管理组合

Quantitative Models and Construction Methods 1. Model Name: Conditional Mean-Variance Multi-Factor Portfolio (CMV) - Model Construction Idea: The model dynamically adjusts factor weights based on market volatility, optimizing post-cost mean-variance utility[4][16][30] - Model Construction Process: - The return of the k-th volatility-managed factor is calculated as: $r_{k,t+1}^{\sigma}=\frac{c}{\sigma_{k,t}^{2}}r_{k,t+1}$ where $r_{k,t+1}$ is the unmanaged factor return, $\sigma_{k,t}^{2}$ is the realized variance of the factor in month $t$, and $c$ is a scaling parameter ensuring the volatility of the managed factor matches the unmanaged factor[30][31] - Factor weights are parameterized as an affine function of the inverse of market volatility: $\theta_{k,t}=a_{k}+b_{k}\frac{1}{\sigma_{t}}$ where $a_k$ and $b_k$ are parameters, and $\sigma_t$ is the realized market volatility in month $t$[33][34] - The portfolio return is expressed as: $r_{p,t+1}(\theta_{t})=\sum_{k=1}^{K}r_{k,t+1}\left(a_{k}+b_{k}\,\frac{1}{\sigma_{t}}\right)$ where $K$ is the number of factors[34] - The optimization problem maximizes the mean-variance utility of the extended factor weight vector $\eta$, accounting for transaction costs: $\max_{\eta\geq0}\widehat{\mu_{\rm ext}}\eta-{\rm TC}(\eta)-\frac{\gamma}{2}\eta^{\prime}\widehat{\Sigma_{\rm ext}}\eta$ where $\widehat{\mu_{\rm ext}}$ and $\widehat{\Sigma_{\rm ext}}$ are the sample mean and covariance matrix of extended factor returns, and ${\rm TC}(\eta)$ represents transaction costs[37][38] - Model Evaluation: The CMV model demonstrates superior performance compared to the unconditional mean-variance portfolio (UMV), even after accounting for transaction costs, due to its ability to dynamically adjust factor weights and incorporate transaction cost optimization[16][62][67] --- Model Backtesting Results 1. Conditional Mean-Variance Multi-Factor Portfolio (CMV) - Mean Annualized Return: 0.477[61] - Annualized Standard Deviation: 0.449[61] - Sharpe Ratio: 1.062[61] - Sharpe Ratio Improvement over UMV: 13% (statistically significant at 1% level)[62] - Annualized Alpha: 0.066 (Newey-West t-statistic: 3.637)[61] - Transaction Costs: 0.213[61] 2. Unconditional Mean-Variance Multi-Factor Portfolio (UMV) - Mean Annualized Return: 0.430[61] - Annualized Standard Deviation: 0.458[61] - Sharpe Ratio: 0.940[61] - Transaction Costs: 0.163[61] --- Quantitative Factors and Construction Methods 1. Factor Name: Volatility-Managed Factors - Factor Construction Idea: Adjust factor exposures inversely proportional to their realized variance to enhance Sharpe ratios during high-volatility periods[30][31] - Factor Construction Process: - The return of the k-th volatility-managed factor is calculated as: $r_{k,t+1}^{\sigma}=\frac{c}{\sigma_{k,t}^{2}}r_{k,t+1}$ where $r_{k,t+1}$ is the unmanaged factor return, $\sigma_{k,t}^{2}$ is the realized variance of the factor in month $t$, and $c$ is a scaling parameter ensuring the volatility of the managed factor matches the unmanaged factor[30][31] - Factor Evaluation: Volatility-managed factors show improved Sharpe ratios in-sample but face challenges from transaction costs and estimation errors out-of-sample[49][51][53] --- Factor Backtesting Results 1. Volatility-Managed Factors (Out-of-Sample with Transaction Costs and Trading Diversification) - Market Factor (MKT): Sharpe Ratio 0.433[54] - Size Factor (SMB): Sharpe Ratio 0.035[54] - Value Factor (HML): Sharpe Ratio 0.089[54] - Profitability Factor (RMW): Sharpe Ratio 0.226[54] - Investment Factor (CMA): Sharpe Ratio 0.153[54] - Momentum Factor (UMD): Sharpe Ratio 0.209[54] - Return on Equity Factor (ROE): Sharpe Ratio 0.324[54] - Investment Activity Factor (IA): Sharpe Ratio 0.193[54] - Betting Against Beta Factor (BAB): Sharpe Ratio 0.746[54]