Abstract
The coskewness puzzle has occupied the empirical asset pricing research since the third-moment asset pricing model was introduced by Kraus and Litzenberger (1976) and Friend and Westerfield (1980). Using the Fama-French 49 US industry portfolios this paper empirically shows that the coskewness and cokurtosis components of higher moment asset pricing models are not robust across the conditional return distribution. The result has a major impact on the modeling, estimation and testing approaches of asset pricing models and is important for a correct interpretation of the higher-moment effects on required returns. This further implies that separation of bull and bear markets and/or applying discrete regime switching approaches may not be sufficient to fully account for the asymmetric higher-moment effects. This also partly explains previous puzzling empirical findings and instability over time of higher-moment factor loadings.
Keywords: Asset pricing, Return distribution, Co- moments, Asymmetry,
JEL classification: G11; G12