A conditional higher-moment CAPM

• Published in: International review of financial analysis Vol. 86; p. 102524
• Main Authors: Vendrame, Vasco, Guermat, Cherif, Tucker, Jon
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.03.2023
• Subjects: CAPM; Conditional models; Higher-moments; Regime switching; CAPM; Higher-moments; Regime switching; Conditional models
• ISSN: 1057-5219; 1873-8079; 1873-8079
• DOI: 10.1016/j.irfa.2023.102524

This paper investigates whether dynamic and moment extensions to the traditional CAPM can improve its empirical performance and offer some alternative explanation to the cross-section of average returns on portfolios of stocks double sorted on book-to-market ratios and size. We consider three extensions. First, we introduce time-varying factor loadings obtained from a multivariate GARCH and dynamic conditional correlations. Second, we extend the model to a four-moment CAPM, which incorporates coskewness and cokurtosis. Finally, we allow for time-varying risk premia, based on a Markov-switching process. Our results confirm that the higher-moment CAPM does not perform well in its unconditional version, but its performance is significantly improved when we introduce a conditional version that accounts for both time-varying factor loadings and time-varying risk premia. The four-moment CAPM tests lead to a positive total risk premium estimate of 0.67% per month over the period 1926–2021, with all risk premia (beta, coskewness, and cokurtosis) exhibiting the expected theoretical signs. •Combining higher moments and conditional models offers significant improvement over the static CAPM.•We compute time-varying risk premia based on Markov-switching process.•We introduce time-varying factor loadings and extend the standard model to incorporate coskewness and cokurtosis.•The conditional four-moment CAPM provides a positive and significant total risk premium.•Beta and cokurtosis exhibit a positive risk premium, while coskewness shows a negative premium as expected.