Abstract: | Ford et al. (this issue) point out that the SURADF panel unit root test may be sensitive to panel composition. This reply shows that they overstate the case since they focus on a short time series of 44 observations. Type II errors are much more likely in this environment so that inconsistent conclusions may arise for individual panel members across differently composed panels. We demonstrate that this problem becomes much less likely when the number of time‐series observations increases. |