Abstract
This paper demonstrates that building characteristics-managed factors using firms’ asset returns greatly reduces the number of factors necessary to explain the cross-section. A 5-factor model based on asset returns explains 62.4% of the variation in 100 factors, whereas an 88-factor model using equity returns explains only 38.6%. Out-of-sample, the asset-based implied mean-variance-efficient (MVE) portfolio achieves a Sharpe ratio of 1.2, compared with 0.75 for its equity-based counterpart. The parsimonious asset-based model explains equity returns better than the equity-based model, as it reduces the number of equity anomalies to 15 compared with 23 for the latter. The nonlinear transformation of returns caused by leverage increases the loadings of firms with high leverage on the equity-based factors, exposes these factors to firm-level systematic risks that would not arise in asset-based factors, and contributes to the factor zoo.