Unit-root testing strategies are unnecessarily complicated because they do not exploit prior knowledge of the growth status of the time series, they worry about unrealistic outcomes, and they double- or triple-test for unit roots. The authors provide a testing strategy that cuts through these...

F tests which test jointly for a unit root and a zero intercept, and so compete against Dickey-Fuller t tests, are shown not to enhance power because they are invariant to the intercept value in the absence of a unit root. Monte Carlo results in the literature that indicate otherwise are shown...