Time series analysis of income convergence in China

Abstract

In this article, we examine both pre- and post-reform economic convergence behaviour in China using the time-series procedures developed in the Hobijn and Franses (2000). Our results are striking: First, we find a number of extremely small convergence clubs for both the pre- and post-reform periods, and these clubs cannot be characterized by geography or the extent of policy preference level. Second, our results show that economic reform led to more divergence.

Notes

¹ As shown in Hobjin and Franses (2000), the convergence of growth rates implies infinite cross-sectional variance of log per capita GDP; that is, the cross-province income distribution will not settle down to a stable nondegenerate distribution. Hence, we focus on asymptotically perfect and relative convergence clubs.

² Actually, Rawski (2001) and Chow (2002) assert that despite some problems, even official estimates are accurate enough for econometric analysis (Henderson et al., 2006)

³ In the interest of space, summary statistics are not provided here, but available upon request.

⁴ Notice that there is no statistical significance attached to the correlation measure. Hence, the interpretation of these numbers itself requires care as well.

Table 1. Cluster correlation between the results for difference bandwidth (Pre-reform)

Table 2. Cluster correlation between the results for difference bandwidth (Post-reform)

⁵ Our results are reasonably robust to the choice of bandwidth parameters. Although the actual composition of the cluster may vary, the main conclusions for other bandwidth choices are very similar to those for bandwidth being three. In the interest of space, only results for bandwidth being three are provided, but the remaining results available upon request.

⁶ Note that more clubs means more divergence.