The properties of asymmetric unit root tests in the presence of mis-specified asymmetry

The seminal analysis of Enders and Granger (1998) is extended to examine the properties of asymmetric unit root tests when the nature of the actual asymmetric adjustment process underlying the observed data is unknown. The analysis is further extended by considering joint testing for asymmetric stationarity in addition to unit root testing. It is shown that the momentum-threshold autoregressive (MTAR) test outperforms the threshold autoregressive (TAR) test. The results indicate that when employing asymmetric unit root tests, practitioners will tend to detect asymmetry of an MTAR rather TAR nature, irrespective of the form of asymmetry actually present in the data.