Working Paper: NBER ID: w20682
Authors: Kewei Hou; Haitao Mo; Chen Xue; Lu Zhang
Abstract: Many recently proposed, seemingly different factor models are closely related. In spanning tests, the q-factor model largely subsumes the Fama-French (2015, 2018) 5-and 6-factor models, and the q⁵ model captures the Stambaugh-Yuan (2017) model. The Stambaugh-Yuan factors are sensitive to their construction, and once replicated via the standard approach, are close to the q-factors, with correlations of 0.8 and 0.84. Finally, it seems difficult to motivate the Fama-French 5-factor model from valuation theory, which predicts a positive relation between the expected investment and the expected return.
Keywords: factor models; asset pricing; investment; expected returns
JEL Codes: E22; E44; G11; G12; G14
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Stambaugh-Yuan factors (C29) | performance differences (D29) |
| replicated factors (C59) | q-factors (C38) |
| theoretical basis of Fama-French 5-factor model (G41) | empirical relationships (C59) |
| q-factor model (C38) | Fama-French 5-factor model (G12) |
| q5 model (C52) | Fama-French 6-factor model (G12) |
| Fama-French models (C58) | q-factor premiums (G22) |