Cross-sectional and serial correlation in a small-sample homogeneous panel data unit root test

Abstract

In this paper, response surface parameters are provided that can be used to obtain critical values for an augmentation of an existing homogeneous panel data unit root test. The augmentation is performed to account for serial correlation in the disturbances. As the existing panel data unit root test is robust against cross-sectional correlation, the augmented test is robust against both cross-sectional and serial correlation. By running a Monte Carlo simulation study, the small-sample properties of the augmented test are shown to be good.

Acknowledgements

The author would like to thank Johan Lyhagen for pointing out the problem studied in this paper. The paper has benefited from the discussions and comments of Tommy Andersson, David Edgerton and Klas Fregert. All simulations in the current paper were performed on the LUNARC cluster. The author would like to thank Thomas Elger, Ulf Erlandsson and Per-Anders Wernberg for technical advice. Finally, financial support from The Crafoord Foundation is gratefully acknowledged.

Notes

¹ It can be noted that one observation for each of the cross-sections is lost when constructing the lag of y. (T − 1) is the effective sample, while T denotes the number of observations available before lagging.