The Effect of Time-Series and Cross-Sectional Heterogeneity on Panel Unit Root Test Power

Panel unit root tests represent a significant advancement in addressing the low power of unit root tests by exploiting cross-sectional and time-series information. In this article we employ Monte Carlo techniques to quantify the power improvements due to cross-sectional information and assess test sensitivity to heterogeneous data. Pooling the data alleviates negative effects of slowly adjusting equilibrium relations as well as persistence in the forcing variable. However, if the panel contains a mixture of unit root and stationary series, the power of the test decreases substantially and the interpretation of the results becomes tenuous