Hierarchical Bayes Methods for Multifactor Model Estimation and Portfolio Selection
提出一种层次贝叶斯方法，通过引入横截面信息提高多因子模型参数估计精度，进而改善投资组合表现。基于纽交所数据的实证表明，该方法在样本外组合表现上优于其他估计方法。
The factor model is an important construct for both portfolio managers and researchers in modern finance. For practitioners, factor model coefficients are used to guide the construction of optimal portfolios. For academicians, factor model parameters play a fundamental role in explaining equilibrium asset prices and other market phenomena. This paper presents a hierarchical modeling procedure that can substantially improve the accuracy of factor model parameter estimates through incorporation of cross-sectional information. It is shown that this improvement in parameter estimation accuracy translates into substantial improvement in portfolio performance. Expressions are derived that characterize the sensitivity of portfolio performance to parameter estimation error. Evidence with NYSE data suggests that the hierarchical estimation technique leads to superior out-of-sample portfolio performance when compared to alternative estimation approaches.