Testing the Conditional CAPM using Cross-sectional Regressions

A Multi-task Learning Approach

authored by

Joachim Grammig, Constantin Hanenberg, Christian Schlag, Jantje Sönksen

Abstract

In this paper, we introduce a novel representation of the conditional CAPM that allows us to express both the beta and the market premium as functions of option prices. To test our model, we conduct cross-sectional regressions that include the implied beta and other stock characteristics as regressors. We contribute to the existing literature by 1) systematically selecting stock characteristics with a combination of ℓ1- and ℓ2-regularization, known as the multi-task Lasso, and 2) addressing the problem of post-selection inference via repeated sample splitting. Empirically, we find that while variants of the momentum effect lead to a rejection of our model, the implied beta is by far the most important predictor of cross-sectional return variation. The framework is suitable to test other implementations of the conditional CAPM or, more generally, conditional linear factor models with time-varying parameters.

External Organisation(s)

Goethe University Frankfurt
University of Tübingen
University of Cologne
Leibniz Institute for Financial Research SAFE

Type

Working paper/Discussion paper

No. of pages

67

Publication date

16.04.2024

Publication status

E-pub ahead of print

Electronic version(s)

https://doi.org/10.2139/ssrn.4788066 (Access: Open)

 

Details in the research portal "Research@Leibniz University"