George Kapetanios ,
Queen Mary, University of London
February 1, 2004
Recently, considerable emphasis has been placed on the problems arising out of cross-sectional dependence in panel unit root tests. This paper adopts the factor based cross-sectional dependence paradigm of Bai and Ng (2004) but suggests alternative factor extraction methods. Some theoretical results for these methods are provided. Further, a detailed Monte Carlo study of these methods for multiple and persistent factors is undertaken. It is found that results are radically different to the serially uncorrelated single factor case. Tests perform much worse and in some cases it is preferable not to correct at all for cross-sectional dependence.
J.E.L classification codes: C32, C33
Keywords:Panel unit root tests, Factor models, Subspace algorithms