讲座简介: | In this paper, we utilize the ridge regression to develop a novel set of characteristics-based factors, referred to as "ridge factors". We demonstrate that the usage of the ridge factors enables the creation of a Bayesian average SDF, which can tackle with model uncertainty in accordance with the theory of asset pricing: shrinking the relative contribution of low variance principal portfolios. Then, through the application the RP-PCA method proposed by Lettau and Pelger (2020a,b), we show that the ridge factor principal portfolios can exhibit greater sparsity compared to the IPCA while maintaining the equivalent empirical prediction accuracy. Furthermore, our Bayesian average SDF yields higher Sharpe ratio for the tangency portfolio compared to other models including IPCA. Additionally, the characteristic selection can be carried out judiciously through the Bayesian approach, leading to reliable results. |