Spatial-Temporal Dynamics of Jiangsu Regional Convergence with Spatial Markov Chains Approach
PU Yingxia,MA Ronghua,GE Ying,HUANG Xingyuan
DOI: https://doi.org/10.3321/j.issn:0375-5444.2005.05.013
2005-01-01
Acta Geographica Sinica
Abstract:Based on the per capita GDP dataset at the county level in Jiangsu province from1978 to 2000, this paper attempts to apply spatial Markov chains to investigate the spatial andtemporal characteristics of regional in Jiangsu. Firstly, all the per capita GDPdata in Jiangsu are classified into 4 different classes by annual provincial average. Due to thechanges in the regional development strategies over time, the whole period is then dividedinto two sub-periods (1978-1990 and 1990-2000), and two Markov transition probabilitymatrices for these two periods are estimated respectively for comparison. Secondly, two kindsof maps are accordingly made in order to visualize spatial patterns of class transitions, one forregion, and the other for region and its neighbors. Finally, conditioning on each region'sspatial lag at the beginning of each year, spatial Markov matrices for the two different periodsare constructed. The conclusions are drawn as follows: (1) The process of regionalconvergence in Jiangsu has been globally characterized by convergence clubs since 1978,but this trend in the 1990s is sharpened and statistically different from that of the period from1978 to 1990. (2) Those regions and their neighbors that both experience upward mobility arelocated in southern Jiangsu, while the regions or their neighbors that move downwardly aremostly found in northern Jiangsu. (3) Regional per capita GDP class transitions in Jiangsu arehighly constrained by their geographical neighbors. If a poor region is surrounded by poorregions, the probabilities of moving upward for the periods from 1978 to 1990 and 1990 to2000 decrease to 0.148 and 0.025 respectively, while they average 0.2 and 0.042 in traditionalMarkov matrices. It suggests that poor regions are negatively affected when surrounded byother poor regions. Conversely, if a rich region is surrounded by rich neighbors, theprobabilities of moving upward for those two sub-periods increase to 1 and 0.991, while theyare 0.987 and 0.984 on average in traditional Markov matrices. It suggests rich regions arepositively influenced by other rich regions being surrounded. These empirical analyses providea spatial explanation to convergence clubs detected in traditional Markov method.