Testing for Stationarity in Panel Data Models when Disturbances are Cross-Sectionally Correlated

In this paper, we investigate the effects of cross-sectional disturbance correlation on a previously suggested panel data stationarity test. We find that the previously suggested test has a serious size distortion if the disturbances to different cross sections are correlated. We suggest a new panel data test procedure that also tests the null hypothesis of stationarity. However, the test procedure that we suggest is robust against the presence of cross-sectional disturbance correlation, as well as serial correlation. Furthermore, the test has an approximate normal distribution and which makes p-values and critical values easy to obtain. By applying our test to investigate output convergence, we illustrate the adverse effects that can occur when neglecting to account for cross-sectional correlation when testing for stationarity in panel data models.