Testing for Unit Roots: Threshold Autoregression and Asymmetric Trend Stationarity

In recent research Enders and Granger (1998) have extended the Augmented Dickey Fuller statistic to allow the unit root hypothesis to be tested against an alternative of stationarity with asymmetric adjustment. In this paper the threshold autoregressive (TAR) and momentum threshold autoregressive (MTAR) unit root tests proposed by Enders and Granger are applied to data on UK investment. It is found that the TAR, MTAR and standard Augmented Dickey-Fuller tests fail to reject the unit root hypothesis. In contrast, an alternative modified TAR unit root test is proposed which rejects non stationarity using specifically derived Monte Carlo critical values. Further testing rejects the null hypothesis of symmetric adjustment, with the form of asymmetry detected supporting recent theoretical predictions for investment behaviour presented by Gale (1996). The results obtained show the form of asymmetry incorporated within the Enders-Granger test to critically influence the resulting inferences drawn.