More Powerful Unit Root Tests with Non-Normal Errors

This paper proposes new unit root tests that are more powerful when the error term follows a non-normal distribution. The improved power is gained by utilizing the additional moment conditions embodied in non-normal errors. Specifi cally, we follow the work of Im and Schmidt (2008), using the framework of generalized methods of moments (GMM), and adopt a simple two-step procedure based on the "residual augmented least squares" (RALS) methodology. Our RALS-based unit root tests make use of non-linear moment conditions through a computationally simple procedure. Our Monte Carlo simulation results show that the RALS-based unit root tests have good size and power properties, and they show significant efficiency gains when utilizing the additional information contained in non-normal errors information that is ignored in traditional unit root tests