An instrumental variable approach for panel unit root tests under cross-sectional dependence

Abstract

For dynamic panel models with cross-sectional dependence, several unit root tests are constructed using a Huber-type instrument, whose null asymptotics are standard Gaussian and do not depend on nuisance parameters. A Monte-Carlo simulation shows that the proposed tests have better sizes and comparable powers relative to other two existing tests developed for cross-sectionally dependent dynamic panel models. © 2005 Elsevier B.V. All rights reserved.