A0253
Title: Asymptotically UMP tests for unit roots in cross-sectionally dependent panels
Authors: Gaia Becheri - Delft University of Technology (Netherlands) [presenting]
Ramon Van den Akker - Tilburg University (Netherlands)
Abstract: Testing for unit roots is a crucial aspect of (time series and) panel data analysis. The presence of unit roots not only determines how to proceed for a correct statistical inference but may also have relevant policy implications. For a correct analysis of cross-country data, the presence of heterogeneity and correlation between units should be taken into account. We consider large panels with cross-sectional dependence generated by dynamic common factors and derive the (asymptotic) power envelope for testing for a unit root. We demonstrate that some of the most popular second-generation tests are asymptotically equivalent and only attain the power envelope in case the long-run variances of the idiosyncratic components of the innovations are homogeneous. We propose new tests that are asymptotically UMP, i.e. whose power functions attain the envelope irrespective of possible heterogeneity in the long-run variances. We consider an asymptotic scheme in which $n$ (the number of units) and $T$ (the time series dimension) tend to infinity jointly, which is the common when studying macroeconomic panels with large $n$ and $T$. Finally, we assess the finite-sample performances of the new tests via a Monte Carlo study.
