The authors address the issue of estimation and inference in dependent non-stationary panels of small cross-section dimensions. The main conclusion is that the best results are obtained applying bootstrap inference to single-equation estimators, such as fully modified ordinary least squares and...