Modeling the Cross Section of Stock Returns Using Sensible Models in a Model Pool

An increase in the number of asset pricing models intensifies model uncertainties in assetpricing. While a pure "model selection" (singling out a best model) can result in a loss of usefulinformation, a full “model pooling” may increase the risk of including noisy information.We make a trade-off between the two methods and develop a new two-step trimming-then-poolingmethod to forecast the joint distributions of asset returns using a large pool of assetpricing models. Our method allows investors to focus on certain regions of the distributions.In the first step, we trim the uninformative models from a pool of candidates, andin the second step, we pool the forecasts of the surviving models. We find that our methodsignificantly enhances portfolio performance and predicts downside risk precisely, and theimprovements are mainly due to trimming. The pool of sensible models becomes largerwhen focusing on extreme events, responds rapidly to rising uncertainty, and reflects themagnitude of factor premiums. These findings provide new insights into asset pricing modelevaluation