In this paper we show that panel unit root tests based on OLS detrending have inferior power relative to tests based on GLS detrending when the deviations of the initial observations from the deterministic components of the series are small. This ranking, however, is reversed for larger deviations. We propose a hybrid panel unit root test that captures the desirable power features of both approaches across the range of initial conditions.
Download the paper in PDF format
David I. Harvey, Stephen J. Leybourne and Nikolaos D. Sakkas
View all Granger Centre discussion papers | View all School of Economics featured discussion papers
School of EconomicsUniversity of Nottingham University Park Nottingham, NG7 2RD
lorenzo.trapani@nottingham.ac.uk