Jörg Breitung analyzes the local power of unit root tests for panel data. He shows that the local power of these tests depends on two terms: one related to the detrending method and another to the limiting distribution under local alternatives. He argues that these terms can offset each other, leading to tests with no power against local alternatives. To address this, he proposes a class of t-statistics that do not require bias correction. These tests are based on simple least-squares regressions and have a standard normal limiting distribution. Monte Carlo experiments suggest that avoiding bias correction can significantly improve test power. Breitung also compares the performance of different tests, including the LL and IPS tests, and finds that the LL test is sensitive to the choice of augmentation lag. He concludes that tests without bias correction, such as the UB test, are more powerful and robust to changes in augmentation lag. The paper highlights the importance of considering the specification of deterministic terms and the need for bias correction in unit root testing.Jörg Breitung analyzes the local power of unit root tests for panel data. He shows that the local power of these tests depends on two terms: one related to the detrending method and another to the limiting distribution under local alternatives. He argues that these terms can offset each other, leading to tests with no power against local alternatives. To address this, he proposes a class of t-statistics that do not require bias correction. These tests are based on simple least-squares regressions and have a standard normal limiting distribution. Monte Carlo experiments suggest that avoiding bias correction can significantly improve test power. Breitung also compares the performance of different tests, including the LL and IPS tests, and finds that the LL test is sensitive to the choice of augmentation lag. He concludes that tests without bias correction, such as the UB test, are more powerful and robust to changes in augmentation lag. The paper highlights the importance of considering the specification of deterministic terms and the need for bias correction in unit root testing.