Testing for spurious causality in exchange rates

Abstract

In this paper, we distinguish between the ‘true’ and ‘spurious’ components of observed causalities. ‘True’ refers to the underlying continuous-time model while ‘spurious’ designates a term arising from the discretization of the model. We provide a test for making the difference between these two components, and thus testing for spurious causality. An application to exchange rates emphasizes the importance of the phenomenon. Indeed, it puts forward that the causalities observed at finite frequencies from the German mark to the Swiss franc are actually spurious.