Unit root tests, seeking mean or trend reversion, are frequently applied to panel data. We show that more powerful variants of commonly applied tests are readily available. Moreover, power gains persist when the modifications are applied to bootstrap procedures that may be employed when cross-correlation of a rather general sort among individual panel members is suspected.
|Number of pages||24|
|Journal||Journal of Applied Econometrics|
|Publication status||Published - 2004 Mar|
All Science Journal Classification (ASJC) codes
- Social Sciences (miscellaneous)
- Economics and Econometrics