New Panel Unit Root Tests under Cross Section Dependence for Practitioners

This paper studies the principle of common recursive mean adjustment and proposes a new detrending method in dynamic panel models. By utilizing recursive mean adjustment, this paper provides three unit root tests: a recursive mean adjusted (RMA) unit root test, a covariate RMA and a pooled RMA-feasible generalized least squares tests. The first two tests are designed for testing the cross sectional average of panel time series data to examine if the common factors in a panel are stationary or not. The third test is designed to test if the idiosyncratic errors are stationary or not. The proposed panel unit root test under cross section dependence is precise and powerful especially when T is larger than N