Linearity Testing in Time-Varying Smooth Transition Autoregressive Models under Unknown Degree of Persistency

Building upon the work of Vogelsang (1998) and Harvey and Leybourne (2007) we derive tests that are invariant to the order of integration when the null hypothesis of linearity is tested in time-varying smooth transition models. As heteroscedasticity may lead to spurious rejections of the null hypothesis, a White correction is also considered. The asymptotic properties of the tests are studied. Our Monte Carlo simulations suggest that the newly proposed tests exhibit good size and competitive power properties. An empirical application to US inflation data from the Post-Bretton Woods period underlines the empirical usefulness of our tests.