This study introduces a novel multi-factor portfolio optimisation framework that extends the traditional Markowitz meanvariance model by integrating forward-looking factor signals (price momentum and analysts earnings forecasts), robust risk measures (Conditional Value-at-Risk and higher-moment constraints), and Bayesian shrinkage to mitigate estimation risk (papers.ssrn.com. Prior research finds that analysts forecasts and momentum signals often dominate multi-factor expected return models, jacobslevycenter.wharton.upenn.edu. Which conduct empirical backtests on a diversified set of U.S. equities (e.g., the S&P500 index and major financial stocks such as JPMorgan Chase, Morgan Stanley, and U.S.Bancorp) and evaluate risk-adjusted performance using multiple criteria, including Sharpe ratio, maximum drawdown, volatility, and cumulative return. Compared to the classical meanvariance optimizer, the multi-factor model consistently performs well across all dimensions. Portfolios constructed with the proposed approach exhibit higher Sharpe ratios, cumulative returns, and lower volatility and drawdowns. This outperformance is especially pronounced during market crises, as the combination of predictive signals and robust risk constraints helps limit losses when markets tumble. These results demonstrate that incorporating forward-looking factors and advanced risk modelling (e.g. CVaR and higher moments) with Bayesian estimation yields more resilient portfolios than meanvariance optimisation alone, with important implications for portfolio management practice and asset allocation theory.
Tian Qiu (Fri,) studied this question.