Footnote:
Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments April 20, 2022 erstellt
Description:
Using Compustat data for US stocks from 1962 to 2021, we confirm others’ findings that the capacity of several variables/factors to predict relative returns diminishes considerably in the second half of the period, after the factors’ discovery and public dissemination.We further find that the portion of annual cross-sectional returns related any factor is exceedingly small, with the median R-square only about 1%. The preponderance of most periods’ overall returns are thus unrelated to any factor.We confirm through an accidental discovery that the statistical techniques common to factor-testing, even techniques designed to be more robust, can be heavily influenced by very few observations. One of our preliminary inferences regarding return momentum continuity is significantly altered when its sample changes by only 11%.We also document that BBB-rated corporate bonds’ excess returns to Treasury bonds do not meet conventional levels of statistical significance, despite being due to a pervasive risk factor. This contrasts with equity factors, several of which have statistically significant effects in the first half of the 1962-2021 period despite having no risk-based rationale for those effects.Between the near-zero R-squares, unduly sensitive cross-sectional estimates and statistically insignificant corporate bond excess returns, we ponder whether the tools used to measure factors’ statistical significance are up to the task of discerning economic relevance.Disclosure: The views expressed herein are those of the author and not necessarily those of Frontier Financial