The Efficacy of the Conditional CAPM: Improved Tests in an International Context

Abstract

Using a machine learning model known as a Long-Short Term Memory model to overcome high dimensionality obstacles, I jointly predict the conditional second moments of eight international indices and test the conditional Capital Asset Pricing Model (CAPM). My results indicate that the world price of covariance risk is equal across eight world equity markets according to the conditional CAPM. Strengths and weaknesses of the estimation process are studied. All results are assessed and reported using out-of-sample tests.

Keywords: Asset Pricing, International Conditional Asset Pricing, Machine Learning, Neural Networks, LSTM, GARCH, Conditional Covariances

JEL Classification: G10, G11, G12, G15, F15, C10, C45