Testing for (non)linearity in economic time series: a Monte Carlo comparison

In recent years interest has been growing in testing for (non)linearity in time series. Several tests are available in literature, some of them are designed to distinguish linearity from a well specied parametric nonlinear model, while others have been developed without a parametric nonlinear alternative in mind. In this paper we review the issue of testing for (non)linearity and examine, via Monte Carlo experiments, the power and size properties of the major linearity tests applied to dierent nonlinear time series models.