A sequential test for pair-wise convergence in Chinese provincial income

In this paper a recently proposed method to sequentially determine the proportion of stationary units in a panel is employed to investigate per-capita GDP convergence in China. The main advantage of this method, in addition to being able to accommodate virtually any serial and cross-section dependence, is its ability to identify arbitrarily small convergence clubs, thereby avoiding potential pre-classification bias inherent in many existing methods. Our main conclusion is that the evidence of convergence is basically nonexistent, and that the little evidence that does exist indicates the existence of a small number of very small convergence clubs.