In this paper, we examine beta-convergence of real per-capita income of Chinese counties. We account for both the spatial dependences between counties and the possibility of different convergence regimes. The first feature is captured by the spatial error term, whereas the second one is modeled using the spatial logit smooth transition approach. Two groups of counties can be identified: 1) counties, which have relatively poor neighbors and tend to grow faster and converge, and 2) counties, which have relatively rich neighbors and tend to grow slower and hence fail to converge. The counties belonging to the first group are concentrated mainly in western interior provinces, such as Qinghai, Sichuan, Yunnan, western part of Xinjiang Uygur. The counties of the second group are located mainly in coastal regions. Whereas in the benchmark model the estimated convergence rate is 0.8% for unconditional regression and 1.7% for condtional regression, the alternative models produce the convergence rate of 1.3-1.5% for unconditional regressions and 2.3-2.6% for conditional regressions, which is quite close to the estimates reported typically in the literature.
chinese counties income convergence LSTR spatial effects