Seeking the Best Fundamental Risk Factors: A Clinical Approach to Fama-French Portfolio Decomposition

This paper performs a thorough analysis of competing construction methods for the design of size (SMB) and value (HML) spread portfolios à la Fama-French. This quasi-clinical investigation of methodological choices uncovers substantial differences in the capacity of estimated premiums to translate stock characteristics into returns. A sequential sort of stocks into long and short portfolios conditioned on control variables (“pre-conditioning”) produces factors that best reflect the corresponding fundamental attributes. Our results are stronger when using the whole firm sample to define breakpoints and a triple sort, which ensures the same diversification (in terms of number of firms) across the characteristic-sorted portfolios forming the long and short legs of the factor. Our results are robust to the inclusion of the momentum dimension in the multiple sorting. The best method produces a volatile and insignificant size premium, but a high and stable value premium.