Measuring inter-provincial income inequality in China: a sensitivity analysis

Abstract

China's regional inequality is a hotly debated topic in the literature, but the fundamental issue of measurement has somehow escaped from the deliberations. In this paper, we select several commonly used income indicators, inequality indices and concepts to measure China's inter-provincial inequality and compare the results. We find China's inter-provincial inequality is very sensitive to different measurements. These sensitivity analyses help us to understand the ongoing debates on China's regional income inequality. We also conduct a cross-country comparison to illustrate the broad ranking of China's regional inequality in the world. We find that China's regional inequality appears not to be as high as usually perceived, despite approaching an alarming level.

Keywords:

• inter-provincial inequality
• measurement
• sensitivity
• China

JEL Classifications:

• O15
• O53
• R10

Acknowledgements

The authors would like to thank Pascale Combes Motel, Barry Naughton, Jean-Claude Berthélemy, Xiaming Liu and an anonymous referee for helpful comments and suggestions.

Notes

Notes

1. In the expenditure approach, GDP is composed of capital formation, household consumption expenditure, governmental consumption and net exports.

2. Since y is in logarithmic form, σ ²(y) is a standard inequality measure known as the logarithmic variance (Cowell 1995).

3. According to the previous definition, permanent population includes temporary residents who have stayed within the jurisdiction for more than one year. However, the period of staying is altered to be half of a year in the new definition reassigned in the 2000 population census. As a result, the magnitude of provincial population has been systematically enlarged since 2000 (Tsui 2007).

4. For example, fiscal decentralization weakened the financial capacity of the central government and its ability to redistribute resources among provinces for an equity purpose (Zhao and Zhang 1999).

5. For simplicity, only the weighted formula is presented. Obviously, the unweighted is merely a special case of the weighted formula where w_i = 1 and N = n.

6. In this paper, we focus on studying regional inequality from the perspective of production rather than consumption or redistribution. Selecting GDPPC and GDPPW as income indicators can also help avoid serious problems in the quality of the data on rural and urban populations.

7. Note that the shrinking inequality in terms of GDPPC in 2000 could result from the definition change in population statistics.

8. In principle, Concept 2 inequality can be used as a lower bound of Concept 3 inequality. However, this is only valid when personal income is used as an income indicator. When using GDP per capita to proxy for an income indicator, the value of Concept 2 inequality cannot be reasonably taken as the lowest bound of Concept 3 inequality because GDP per capita excludes the possible transfers among provinces (Milanovic 2005, 442).

9. The computed values of these inequality measures are available upon request.