This paper examines the testing of unit roots in large $n$ and $T$ panels with correlated cross-sectional units. The authors model the cross-sectional correlation using an approximate linear dynamic factor model, where the panel data is generated by both idiosyncratic shocks and unobservable common factors. They propose unit root tests in this environment and derive their asymptotic distributions under the null hypothesis of a unit root and local alternatives. The tests are shown to have significant asymptotic power when there are no incidental trends, but they lack power against local alternatives when incidental trends are present and deterministic components need to be estimated. Through Monte Carlo simulations, the finite sample properties of the tests are evaluated. The paper also discusses the estimation of the number of factors and extends the analysis to a model with incidental trends.This paper examines the testing of unit roots in large $n$ and $T$ panels with correlated cross-sectional units. The authors model the cross-sectional correlation using an approximate linear dynamic factor model, where the panel data is generated by both idiosyncratic shocks and unobservable common factors. They propose unit root tests in this environment and derive their asymptotic distributions under the null hypothesis of a unit root and local alternatives. The tests are shown to have significant asymptotic power when there are no incidental trends, but they lack power against local alternatives when incidental trends are present and deterministic components need to be estimated. Through Monte Carlo simulations, the finite sample properties of the tests are evaluated. The paper also discusses the estimation of the number of factors and extends the analysis to a model with incidental trends.