A Protocol for Factor Identification

Abstract

We propose a protocol for identifying genuine risk factors. The underlying premise is that a risk factor must be related to the covariance matrix of returns, must be priced in the cross-section of returns, and should yield a reward-to-risk ratio that is reasonable enough to be consistent with risk pricing. A market factor, a profitability factor, and traded versions of macroeconomic factors pass our protocol, but many characteristic-based factors do not. Several of the underlying characteristics, however, do command premiums in the cross-section.

Keywords: Factors, Factor Identification, Principle Components, Cross-Sectional Regression, Return Variation, Error-In-Variable Problem

JEL Classification: G10, G11, G12