The Resource Curse and Its Transmission Channels: An Empirical Investigation of Chinese Cities’ Panel Data

ABSTRACT

This article re-examines the resource curse hypothesis on the city level in China using data from 273 cities during the period 2001–2010. The system GMM dynamic panel estimator is applied to address the potential endogeneity problems. Our empirical analysis suggests that natural resource dependence has a small and insignificant impact on economic output when we control for the negative indirect impacts. If the indirect impacts of the transmission channels through which the resources hinder economic output are included, the total effect of natural resource dependence on economic output increases to 10 times the direct effect. Moreover, the capital investment channel is shown to be the most important of these transmission channels.

KEY WORDS:

• economic output
• natural resource dependence
• resource curse
• transmission channels

Funding

The authors appreciate the financial support from the “Beijing Higher Education Young Elite Teacher Project” (YETP0201), National Science Foundation of China (71273269), and funds of Renmin University of China (11XNL009).

Notes

^1. http://www.china.org.cn/china/news/2009-03/05/content_17384699.htm

^2. Mining covers metal and non-metal mining, timber harvesting, and water production and supply.

^3. Some variables are missing in certain cities. In the estimation, we apply the largest available sample.

^4. China had experienced rocketing prices of almost all of the natural resources in this period of 2000–2010 due to the striking economic growth, and the resource sector in China benefited most from it. However, this trend has been reversed since 2011. For example, the coal price had tripled during 2004-2010 but fell more than 40% since 2011. Unfortunately, we cannot test the resource curse hypothesis using more years’ data since several key variables at the city level are not available.

^5. The aggregate effect of natural resource abundance on economic growth can also be estimated by running the equation $ln{y}_{i,t}-ln{y}_{i,t-1}={\alpha }_0+{\alpha }_{1}ln{y}_{i,t-1}+{\alpha }_2{R}_{it}+yea{r}_t+{\eta }_i+{\epsilon }_{it}$. The estimated system GMM coefficient is 0.0007 with a marginal significance level, which is the same as summing up ${\alpha }_2$ and ${\alpha }_3{\rho }_2$of each transmission channel.

Additional information

Funding

The authors appreciate the financial support from the “Beijing Higher Education Young Elite Teacher Project” (YETP0201), National Science Foundation of China (71273269), and funds of Renmin University of China (11XNL009).