Investigating regional income convergence in China: an exploratory spatio-temporal perspective

Abstract

Uneven regional development has become an issue in China since the Open and Reform Policy in the 1980s. The imbalance can be observed as the “coastal-west divide” in previous convergence studies. This paper aims to investigate the convergence of income and its determinants across Chinese provinces from 1999 to 2017 by including the role of space. Our findings from a new exploratory technique in spatial analysis show convergence in income and investment but not in human capital, which is in line with results from non-spatial classical convergence models. We confirm the investment-driven economy in China and suggest that investment mainly facilitates the convergence trend of regional income and gradually narrows the coastal-wide divide. Targeting investment and spatial connectivity in low-income regions could be a key policy direction to achieve balanced regional development.

Appendices

Appendix A: constructing a queen contiguity weight matrix based on the Thiessen polygon

To facilitate the use of the queen contiguity matrix concerning Hainan island, we employ the Thiessen polygon approach to construct an entirely contiguous polygon structure (Anselin et al. 2009). We take the centroids of the original administrative boundaries in Fig. 

Fig. 7

Connectivity map of queen contiguity for Chinese provinces and Thiessen polygon

7a to generate the new polygon structure as shown in Fig. 7b. The constructed Thiessen polygon structure allows us to create a spatial weight matrix with common-boundary criteria. The constructed spatial connectivity based on the Thiessen polygon in Fig. 7b can also be displayed on the original polygon map of Fig. 7a.

Appendix B: ESTDA results using the 5-KNN weight matrix

For robustness and comparison of our results, we apply ESTDA using $k$-nearest neighbor weights, an alternative type of distance-based spatial weights that avoids the problem of isolates. The results of the robustness check support our main findings. From plots (a) in Figs.

Fig. 8

Directional LISA scatter plot of GDPpc; 1999–2017 (using 5-KNN matrix)

8,

Fig. 9

Directional LISA scatter plot of INVpc; 1999–2017 (using 5-KNN matrix)

9, and

Fig. 10

Directional LISA scatter plot of HC; 1999–2017 (using 5-KNN matrix)

10, we can observe the role of spatial dependence in the regional convergence of GDP and investment, but it is unclear for human capital. Provinces with lower initial income and investment levels tend to move upward in their relative positions with their neighbors, whereas richer provinces gradually lose their relative positions. From panels (b) in Figs.8, 9, and 10, the spatio-temporal patterns of GDP and investment show the reduction trend in the coastal-wide divide, whereas human capital demonstrates a vague conclusion. Finally, the statistical inference in Table

Table 5 Inference of directional LISA using 5-KNN matrix.

5 reports a similar distribution pattern in all quadrants as shown in Table 4. In addition, the null hypothesis of spatial independence is rejected for all variables, but not in all quadrants. Although the spatial dynamics of regional investment is primarily characterized by the negative spatial co-movements as shown in quadrants 2 and 4, the role of spatial dependence has shaped the observed convergence trend in GDP and investment.