A model-free characterization of causality

This paper introduces a nonparametric characterization of causality relying on information-theoretic statistics and proposes a causality testing procedure robust to the presence of potential nonlinearities. Monte Carlo simulations suggest that the test performs well in a nonlinear context.