The term structures of co-entropy in international financial markets

We propose a new entropy-based correlation measure (coentropy) to evaluate the performance of international asset pricing models. Coentropy captures the codependence of two random variables beyond normality. We document that the coentropy of international stochastic discount factors (SDFs) can be decomposed into a series of entropy-based correlations of permanent and transitory components of the SDFs. We employ the cross section of G-10 countries to obtain model-free estimates of all the components of coentropy at various horizons and show that a model featuring two predictable components of consumption growth rates, global disasters, and recursive preferences is needed to account for the composition of codependence at all horizons