Abstract
In this paper, we introduce new type of within groups transformations so called the double within groups (DWG) and the long within groups (LWG) transformations. Both transformations eliminate individual effects and heterogeneous deterministic time trends jointly. Asymptotic analysis shows that both the DWG and LWG estimators are consistent when T is large. Simulation results show that both estimators have smaller root mean squared errors than the modified within groups estimator.
Acknowledgements
The authors are deeply grateful to Taku Yamamoto and Hiroaki Chigira for helpful comments. The first author also acknowledges financial support from the JSPS Fellowship. All remaining errors are our own.
Notes
1 Since is inconsistent when T is large, we do not consider
.