Nonlinearity in financial series: transitory or permanent?

In this article we present evidence that nonlinearity episodes in financial series are more permanent than transitory. At the same time, these episodes show different behaviors depending on the market analyzed, which would indicate that they are not completely synchronized. On the other hand, the size of the window for detecting nonlinear episodes has an impact on the number of nonlinear windows found, as well as the percentage of nonlinear windows with respect to the total number of windows, confirming a window size effect. The results strongly invalidate the efficient markets hypothesis and forcefully explain the incapability to predict its future values.