On the Local Power of Fixed T Panel Unit Root Tests with Serially Correlated Errors

Analytical asymptotic local power functions are employed to study the effects of general form short term serial correlation on �fixed-T panel data unit root tests. Two models are considered, one that has only individual intercepts and one that has both individual intercepts and individual trends. It is shown that tests based on IV estimators are more powerful in all cases examined. Even more, for the model with individual trends an IV based test is shown to have non-trivial local power at the natural root-N rate.