The Decoupling of China's Economy-Carbon Emission and Its Driving Factors

: 13 Decoupling between economic growth and carbon emission is a global hot topic. This paper studies China’s economic-carbon decoupling and its driving factors. By 15 using the panel data of 30 Chinese provinces from 2000 to 2016, the decoupling index 16 of each province is calculated with the Tapio model. It is found that many provinces 17 have progressed from no decoupling to weak decoupling and then strong decoupling. 18 Then, the econometric models are used to explore the driving factors. Results show that 19 energy structure is the most important factor, followed by GDP per capita and energy 20 intensity, which all increase CO 2 emission significantly. The results are robust when 21 tested with GMM, PCSE and FGLS estimation and LMDI decomposition. Further, we 22 conduct a comparative analysis regarding the temporal and spatial characteristics of the 23 above three driving factors to identify their relationship with decoupling, four groups of regions that represent different economic features are selected for the analysis. Heterogeneity effects of the factors among the regions has been observed, based on this we provide targeted strategies for different regions.

conduct a comparative analysis regarding the temporal and spatial characteristics of the 23 above three driving factors to identify their relationship with decoupling, four groups 24 of regions that represent different economic features are selected for the analysis. Declarations: 30 (1) Note of preprint server 31 I have not submitted my manuscript to a preprint server before submitting it to 32 Environmental Science and Pollution Research. 33 (2) Ethics approval and consent to participate 34 Not applicable.

35
(3) Consent for publication 36 Not applicable. 37 (4) Availability of data and materials 38 The datasets used and/or analysed during the current study are available from the 39 corresponding author on reasonable request. 40 (5) Competing interests 41 The authors declare that they have no competing interests.  (Grand, 2016). By comparing the economic-carbon emissions decoupling 74 relationships across regions and periods, we can determine whether the developing is 75 following a sustainable path. 76 However, a more important question is what drives the decoupling? Although 77 there are many studies expressed concerns to this problem, they mainly focus on two 78 aspects: one is to obtain the decoupling state through direct calculation, such as however, this step is neglected in many studies. In this paper, besides identifying factors 85 affecting CO2 emissions, we will also analyze the dynamic relationship between the 86 factors and the decoupling based on the economic characteristics of typical regions, so 87 as to provide more targeted strategies. 88 Regarding identification of factors affecting CO2 emissions, econometric models 89 will be used in this paper. However, it is worth noting that the results of related studies thus resulting in inaccurate estimation (Li, 2008). There are two approaches which may 99 have potential, one is the Kaya identity approach (Leal et al., 2019), and the other is the 100 SPIRPAT (Stochastic Impacts by Regression on Population, Affluence and Technology) 101 approach (Shuai et al., 2017), both use chain multiplication to decompose the factors 102 that affect carbon emissions, so as to obtain a direct theoretical relationship between 103 each factor and carbon emissions. This paper tries to employ the above two approaches 104 for the selection of explanatory variables to make an improvement. In addition, we also 105 use a variety of methods for robustness analysis; especially, we compare the regression 106 results with that of the LMDI decomposition (this method is also widely used in related 107 studies (Chang et al., 2019)). 108 In summary, this article intends to supplement the current research in the following 109 aspects. First, we conduct a comparative analysis regarding the temporal and spatial

117
The rest part of the study includes: Chapter 2 is about literature review; we focus 118 on the factors affecting CO2 emissions and the link between them and decoupling.

119
Chapter 3 is about the overview of China's economy-carbon decoupling. Chapter 4 is 120 method and data, which explains econometric model and related indexes and data;

121
Chapter 5 is result analysis, which interprets the results revealed by the econometric 122 model (including robustness analyses)；and a further analysis is conducted to identify 123 the consistency between the key factors and decoupling; the last chapter is conclusion 124 and suggestion.   Another important question should be concerned is that only obtaining the factors 168 affecting CO2 emissions (many studies stop here) is not enough to explain decoupling, 169 the economic situation also needs to be concerned. Because decoupling is exactly the 170 ratio between growth in emissions relative to growth in GDP (Mikayilov et al., 2018).

171
A more immediate piece of evidence is that China's carbon emission intensity has been 172 declining in recent years, however, our calculations shows that not every region can be 173 stably decoupled (see Table 1). The underlying logic of this heterogeneity is that: 174 decoupling, or rather, the carbon emission elasticity of GDP growth will be affected by  Obviously, as a member of many developing countries, China's carbon reduction 211 experience is of great demonstration significance. Therefore, taking China as the case 212 can not only provide more scientific basis for domestic carbon reduction practices, but 213 also provide a reference for other developing countries' carbon reduction policies.    will be, thus leading to more increase in carbon emissions. Based on the above 287 viewpoints, we further take per capita GDP and population size into consideration and 288 expand the basic model to: Where, gdppr and p represent per capita GDP and population size respectively.        and several indicators such as intensity of energy consumption, GDP per capita, et al.

387
The results in Table 5 show that: the sign of inte is generally positive, indicating 388 that as the intensity of energy consumption increases, carbon emissions will also 389 increase. The sign of stru is positive, indicating that the higher the proportion of fossil 390 energy consumption in the total energy consumption, the greater the carbon emissions.  Note: * p < 0.1, ** p < 0.05, *** p < 0.01；Standard errors in brackets.

503
The above regression results show that the signs of the coefficients in the 4 504 regressions are consistent (except for lnrd and L.lnrd in column (4)), and they are also 505 consistent with the results in Table 5. In particular, the sign of L.lnrd is negative, 506 indicating that there is a lag effect in R&D. From the perspective of statistical 507 significance, the significance levels of the variables in the three GMM regressions are 508 basically consistent. Compared with the results in Table 5, stu and gdppr are still 509 significant, rd is still insignificant, but inte changes from significant to insignificant, p 510 and iep from insignificant to significant. Judging from the ranking of the effects of each 511 explanatory variable, stru is still the most prominent factor, followed by gdppr. In  Table 5 are consistent. Therefore, it can be concluded 515 that the results of main explanatory variables such as stru, gdppr are robust.

516
It should be noted that dynamic panel regression needs to satisfy the assumption 517 that the error term sequence is not correlated, and meanwhile the IVs are not over-518 identified. In this paper, the Arrellano-Bond method is used to test the hypothesis of 519 serial uncorrelation (H0: the disturbance term of the differential model is not serially 520 correlated), and the Hanson method is used to test the validity of IV (H0: IV is not 521 correlated with the error term), results are shown in   Note: * p < 0.1, ** p < 0.05, *** p < 0.01；Standard errors in brackets；Column (2) and (4) correspond to the first 564 random sampling. The ids of selected provinces and cities include: id1, id4, id8, id11, id14, id16, id17, id20, id24, 565 id27, and id28; Column (2) and (4) correspond to the second random sampling, The ids of selected provinces and 566 cities include: id2, id3, id4, id8, id11, id15, id17, id18, id23 and id25.

567
Compared with the results in Table 5, the regression results in Table 9 show that 568 rd has changed from insignificant to significant in several regressions (Column (4) and 569 (5)), but the sign remains the same; iep changes from a positive sign to a negative sign, 570 but the significance remains the same; the symbol of inte has changed in several 571 regressions (Column (4) and (5)), but the significance remains unchanged. The possible 572 reasons for the above-mentioned changes include: using new method of PCSE and 573 FGLS to correct cross-sectional correlation, heteroscedasticity, and serial correlation of 574 the error term, meanwhile, random sampling leads to a reduction in sample size. In 575 general, the results in Table 9 are consistent with that in identify the relationship between these factors and decoupling, we will analyze their 606 trends over time & region to observe whether they are consistent with the decoupling 607 changes; regional economic development will also be concerned during the analyses.    Based on the economic features and Figure 4~6, the regions are divided into 4 645 groups. The first group includes economically developed regions such as Beijing, 646 Shanghai and Tianjin. They possess high per capita GDP growth rate, at the same time, 647 they are actively optimizing the energy consumption structure and reducing the energy 648 intensity, thus a well decoupling status is obtained.

649
The second group includes the moderately developed regions such as Shandong 650 and Hunan, which have faster economic growth and higher per capita GDP growth rate, 651 their energy consumption intensity has also been reduced to a low level. However, as 652 the adjustment of energy structure of Shandong is small but of Hunan is larger, Hunan 653 has achieved a better decoupling status, but Shandong has not.

654
The third group includes Jilin and Heilongjiang, both belong to traditional 655 resource-based provinces. Although the economic foundation was good, the economic 656 growth rate has been declining in recent years (especially for Heilongjiang). Regarding 657 the relatively high GDP per capita, the reason may not be economic growth but 658 population loss. At the same time, adjustment of their energy structure is relatively 659 small due to high dependence on fossil energy. Although energy consumption intensity 660 has experienced a significant reduction, it is mainly due to the rapid economic recession, 661 which reduces CO2 emission and make Jilin decouples between economic growth and 662 carbon emission (but this doesn't mean sustainable). Comparatively, Heilongjiang 663 performs even poorer because it experiences much faster economic recession and slight 664 increment in energy structure, which prevent it from decoupling.