A Protocol for Factor Identification

Abstract

Several hundred factor candidates have been suggested in the literature. We propose a protocol for determining which factor candidates are related to risks and which candidates are related to mean returns. Factor candidates could be related to both risk and returns, to neither, or to one but not the other.

A characteristic cannot be a factor. Time variation in both risk premiums and covariances is a challenge, but manageable with recently developed statistical procedures. We illustrate those techniques and also propose a new instrumental variables method to resolve the errors-in-variables problem in estimating factor exposures (betas) for individual assets.