Multifactor Attribution for long-only investors

This paper provides a comparative analysis of techniques used to improve the quality of multifactor attributions on constrained long-only portfolios. Our research includes a Weighted Least Squares (WLS) approach, the derivation of constrained factor mimick­ing portfolios as well as a time series- and a non-linear adjustment approach. Based on the multiperiod attribution of a representative portfolio over the timeframe from 1997 to 2019, we find that significant improvements can be achieved by applying both adjust­ment methods since these better align the attribution with the constrained investment process. In addition, using a shrinkage estimator in the WLS approach enhances the widely used market cap weighting scheme by adding further statistical robustness.