Driving effect of CO2 emissions on economic growth—application of empirical likelihood for generalized method of moments

Abstract

In this paper, the panel data of 19 out of 21 APEC members from 1993 to 2018 is selected as a research sample. Firstly, based on the extended Cobb-Douglas production function, the relationship between CO2 emissions and economic growth is explored via the panel data methods, which fully consider the endogeneity between variables and the unobserved individual heterogeneity. Secondly, to examine the driving effect of CO2 emissions on economic growth, the Empirical Likelihood for Generalized method of moments with nonparametric estimation characteristics is utilized to calculate the returns on economic scale and the corresponding confidence interval. The results show that CO2 emissions have a significant positive impact on economic growth. Moreover, the confidence interval of returns to economic scale is very small, indicating current levels of CO2 emissions have reached the limit to drive economic growth when the technical level, labor employment, capital, urbanization, trade, and other factors tend to stabilize. Finally, some recommendations were presented to achieve sustainable development goals. Highlights In combination with the extended Cobb-Douglas production function theory, employs 2SLS and GMM method that takes unobserved the heterogeneity between regions and endogeneity between variables in consideration to explore the relationship between CO2 emissions for economic growth. The extended model of the GMM-Empirical Likelihood for Generalized Moment Method (GMEL) is used to investigate whether the Driving effect of CO2 emissions on Economic Growth reaching its limit or not? Selected the panel data of APEC countries or regions as the research samples, which has never been studied by scholars. The GMEL is a non-parametric statistical inference method that has higher estimation accuracy than the traditional parametric estimation method. The confidence interval constructed by the GMEL method has the advantages of domain retention, transformation invariance.