What Is the Regional Differences in Carbon Emission Performance: Evidences From Energy-intensive Industries in China?

5 As the major energy consumers, energy-intensive industries are the key players in achieving carbon emission 6 reduction targets. Grasping the carbon emission reduction potential has a direct impact on the implementation 7 of the carbon emission reduction policies of China. The paper builds a super-Slack Based Model(SBM) 8 considering this undesirable output, and calculates the carbon emission efficiency. Then, the Meta-Frontier 9 Malmquist-Luenberger productivity index (MF-MLPI) is constructed to dynamically analyse the growth rate 10 changes of the carbon emission efficiency and the regional differences in energy-intensive industries. 11 Furthermore, the carbon emission reduction potential of the energy-intensive industries in various economic 12 regions of China is discussed and the conclusions are as follows: there is a big difference in the carbon emission 13 Technology Gap Ratios (TGRs) of the energy-intensive industries in different economic regions; the growth rate 14 of the carbon emission efficiency of energy-intensive industries shows a trend of first declining and then slowly 15 recovering while the carbon reduction potential generally shows a trend of decreasing and then rising; and the 16 carbon emission reduction potential in the eastern region keeps decreasing. The following is recommended: the 17 government should rationally distribute energy-intensive industries, promote industrial structure adjustment, 18 optimize the energy structure according to the regional industrial advantages; increase investment in R&D, 19 promote energy technology innovation in energy-intensive industries; prioritize the promotion of carbon peaks 20 on key emission industries and regional, formulate differentiated plans for the regions and industries with 21 different carbon emission reduction potentials. 22 The paper describes the CO 2 emissions of the energy-intensive industries of China, builds a Super-SBM model 2 considering undesirable outputs, and calculates the carbon emission efficiency. Then, Meta-Frontier Malmquist- 3 Luenberger Productivity Index (MF-MLPI) is constructed to dynamically analyse the growth rate changes of the 4 carbon emission efficiency and the regional differences among energy-intensive industries; furthermore, the 5 carbon emission reduction potential of the energy-intensive industries in various economic regions of China are 6 discussed. Conclusions are drawn as follows.


Introduction
The establishment of the National Independent Contribution (INPC) mechanism in the Paris Agreement has 26 opened a new stage of global climate change governance. Green and low carbon have become the core concepts 27 of global climate governance in the future. With the largest carbon emissions in the world, China is facing huge 28 pressure to conduct carbon emission reductions, which has restricted the industrial green development of China. 29 The Chinese government promises to reach its peak carbon dioxide (CO2) emissions at approximately 2030 and 2010 National Economic and Social Development Statistical Report" identified six energy-intensive industries 36 including the following: chemical raw materials and chemical products manufacturing, the petroleum processing 37 coking and nuclear fuel processing industry, the non-metallic mineral products industry, the ferrous metal 38 smelting and rolling processing industry, the non-ferrous metal smelting and rolling processing industry, and the 39 electric power heat production and supply industry. 40 As shown in Figure 1, from 2005 to 2017, the energy consumption of energy-intensive industries rose from 1.12 1 to 2.17 billion tons of standard coal, increased by 94.84%; and energy consumption of energy-intensive 2 industries accounted for about 72.95% of industrial energy consumption. Here, the proportion was significantly 3 reduced in year 2005 and year 2010 due to a relatively large increase in industrial energy consumption compared 4 to other years. During the research period, the proportion of energy consumption in energy-intensive industries 5 gradually increased from 42.94% to 48.74%, with a growth of 11.11%. As shown in Figure 2, the proportion of CO2 emissions in energy-intensive industries with respect to all industry 9 rose from 59.34% to 76.51%, which increased by 28.93% (it strongly fluctuated in the Years 2005 and 2010). In 10 terms of specific industries, the CO2 emissions of the ferrous metal smelting and rolling processing industry 11 were relatively large, accounting for approximately 35% of the CO2 emissions of energy-intensive industries. 12 The CO2 emissions of the ferrous metal smelting and rolling processing industry increased by 89.36% from 2005 13 to 2014 (with the largest increase) while it steadily decreased after 2015. The change of the CO2 emissions in 14 the non-metallic mineral products industry is similar to that in the ferrous metal smelting and rolling processing 15 industry. The CO2 emissions of the chemical raw materials and chemical products manufacturing industry, the 16 non-ferrous metal smelting and rolling processing industry, the petroleum processing coking and nuclear fuel 17 processing industry, and the electric power thermal production and supply industry all showed gradual upward 18 trends. The increase of the CO2 emissions in the chemical raw materials and chemical product manufacturing 19 industry was the largest, rising by 1.17 times; meanwhile, after 2015, the CO2 emissions of chemical raw 20 materials and chemical product manufacturing stabilized and had a slight downward trend. Year chemical raw materials and chemical products manufacturing petroleum processing coking and nuclear fuel processing industry non-metallic mineral products industry ferrous metal smelting and rolling Processing industry non-ferrous metal smelting and rolling processing industry electric power heat production and supply industry Proportion of Energy-intensive Industries reduction targets. At the end of 2017, the National Development and Reform Commission officially launched 1 the national carbon emissions trading system for the power industry; however, it is difficult to calculate the 2 quotas for other key industries such as petrochemicals, building materials, and iron and steel due to the lack of 3 basic carbon emissions data for most companies. In addition, the adjustment of the industrial distribution has led 4 to energy-intensive industries gradually shifting from the southeast coast to the central and western regions; 5 therefore, establishing the carbon market can not only effectively reduce the "pollution paradise" effect, but it 6 can also improve the economic growth of the industrialized transition regions in the central and western regions 7 (Tang et al. 2016). It can be seen that as an important part of the carbon market, the data of energy-intensive 8 industries directly affect the improvement of the national carbon emissions trading system, and the carbon 9 productivity of energy-intensive industries directly affects the carbon productivity of downstream industries and 10 the entire economic system. The carbon emission reduction potential of their industries also directly influences 11 the achievement of the carbon emission peak control goals of China. Therefore, the green development dynamics 12 and carbon emission reduction potential of enterprises in energy-intensive industries need to be widely studied 13 by academics and departmental decision makers.  -output table and the structural  26 decomposition model, Shao and Li (2018) found that after 2007, the declines in export contributions and 27 technological changes led to a decline in the demand for energy-intensive products, which ultimately led to the 28 industry's overcapacity and the value-added rate continued to decline. 29 Xie et al. (2017) calculated the environmental efficiency and emission reduction costs of the industrial sectors 30 and considered that it was necessary to consider the heterogeneity of the industry when proposing carbon 31 emission reduction strategies. Guo (2014)  countries implemented large-scale CCS projects in energy-intensive industries such as steel, power, 7 petrochemicals, etc., and effectively achieved carbon emission reduction targets (IEA 2017).

8
A series of studies on energy-intensive industries have been conducted to assess their carbon emissions efficiency 9 and carbon emissions reduction potential. The carbon dioxide emissions of energy-intensive industries account 10 for 64% of the total EU industrial emissions. To achieve the emission reduction targets of the EU by 2050, the 11 scientific community and industry have conducted in-depth cross-industry technical decarbonisation research in 12 key areas (Gerres et al. 2019). Research found that the proportion of energy-intensive industries in Korean 13 manufacturing was large, which meant that the energy efficiency was relatively low; and it also found that R&D Different from the existing literatures, which mainly discuss the carbon emissions of energy-intensive industries 29 as a whole or a single carbon-intensive industry, the main contribution of the paper is to analyse the regional 30 heterogeneity of the CO2 emissions of energy-intensive industries of China, evaluate the carbon efficiency and 31 further calculate the carbon emission reduction potential of energy-intensive industries. Considering regional 32 heterogeneity factors such as the level of economic development and resource endowments among different 33 provinces in China, we can avoid deviations in the assessment of the overall carbon emission efficiency of 34 energy-intensive industries and dynamically predict the state of the carbon emission reduction potentials of 35 different provinces. The rest of this paper is organized as follows. The third part states the main methods adopted 36 in this paper, including the data processing and data resources. In the fourth part, we show our results and the 37 related discussion. Finally, we conclude this paper and put forward some policy recommendations. 38

Super-SBM model considering undesirable output 41
The multi-objective decision model is used to evaluate the relative efficiency of the decision-making unit in the 42 Data Envelopment Analysis (DEA) model. Since the model does not need to consider the input and output 1 function relationships, it has been widely used in the field of environmental and energy efficiency assessment in 2 recent years. 3 Tone (2001) proposed a slack-based measure (SBM) model, which solves the problem that the traditional radial 4 DEA model does not include the slack variable to measure the inefficiency, and makes up for the deviation 5 caused by the selection difference between the radial and angle. In actual industrial production, a "good" output 6 is usually accompanied by "bad" outputs such as sewage and exhaust gas. Tone (2014) defined the SBM-DEA 7 model including the undesired output and better solved the problem of slack variables and the effect of undesired 8 output on efficiency values in efficiency evaluation. In the analysis results of DEA, there are usually cases where 9 multiple decision-making units are invalid, and it is unable to further distinguish decision-making units with the 10 same efficiency value. To solve this problem, Andersen and Petersen (1993) proposed the Super-Efficiency 11 model, which can further evaluate and rank the results of SBM-DEA. Thence, based on an undesired output, the 12 paper builds an SBM super-efficiency model (Super-SBM). Compared with other DEA models, it can more truly 13 reflect the essence of the carbon emission efficiency evaluation of energy-intensive industries. 14 Suppose that the low-carbon economic production system has "n" decision-making units (DMUi, where i=1, 15 2...n), and each decision-making unit has 3 vectors of input, desirable output, and undesired output. Each vector 16 is represented as x ∈ × , ∈ 1 × , and ∈ 2 × . Then, define the input-output matrix as follows: 17 Under the assumption of constant returns to scale, the possible production sets of production units are as follows: 19 where λ is the weight vector. The Super-SBM model considering undesired output is as follows: 21 In the model, is the target efficiency value, k is the decision-making unit being evaluated, refers to the 24 input, refers to the undesirable output, represents the shortage of the desirable output of the decision-25 making unit, represents the excess of the undesired output of the decision-making unit, − represents the 26 input relaxation variable, represents the relaxation variable of the desirable output, and represents the 27 undesirable output The resulting relaxation variable is strictly monotonically decreasing with respect to 28 − , , and , and it satisfies 0 ≤ ≤ 1 ; and when = 1 and − , , and = 0, the decision-29 making unit is effective. If < 1, it means that the decision-making unit is ineffective and the input-output 30 variables of the model needs to be optimized. 31

Meta-Frontier Malmquist-Luenberger Productivity Index (MF-MLPL) 32
In addition to measuring the carbon emission efficiency of energy-intensive industries, this paper will also 1 dynamically analyse the changes in the carbon emission performance of energy-intensive industries. Considering 2 the regional heterogeneity of the production technology of different decision-making units, combined with the 3 Directional Distance Function (DDF), this paper proposes the MF-MLPL considering undesirable output. 4 Hayami (1969) proposed the meta-frontier technology to solve the problems that may arise from a single 5 production frontier. It divides the decision-making units into different groups as needed, and each group 6 constitutes a production frontier as a group frontier. Afterwards, the envelope curves formed by different groups 7 of production frontiers obtain a common frontier (Meta-Frontier), and the technological gap between the group 8 frontier and the common frontier is the Technology Gap Ratio (TGR). To distinguish different decision-making 9 unit groups containing undesirable outputs, based on the environmental technology (the Environmental 10 Technology) of Fare et al. (2007), the group production technology is set as follows ( )： = 11 {( , , ): produce ( , )}. Assuming that the production frontier is the optimal production frontier of 12 the evaluated decision-making unit, its production technology can be expressed as all groups' Union of 13 production technology sets ( Directional Distance Function, the Directional Distance Function (DDF) for the group frontier production 15 technology and the common frontier production technology is as follows: 16 The Malmquist index is usually decomposed into the efficiency change (EC) and technology change (TC). 20 Learning from the meta-Frontier approach of Oh and Lee (2010), the Group-Frontier Malmquist-Luenberger 21 Productivity Index (GF-MLPL) is decomposed into the following: 22 1 + ⃗ ⃗ 0 ( +1 , ( +1) , ( +1) ; g +1 ) = EC × 23 1 + ⃗ ⃗ 0 ( +1) ( +1 , ( +1) , ( +1) ; g +1 ) 24 The Meta-Frontier Malmquist-Luenberger Productivity Index (MF-MLPL) is decomposed into the following: 26 1 + ⃗ ⃗ 0 ( +1 , ( +1) , ( +1) ; g +1 ) = 1 Among them, the BPC (Best Practice Change) is the gap between the current frontier and the global frontier 5 within the group from period t to t+1, that is, the global Malmquist technology change (TC); and the TGC is the 6 change in the technical gap ratio from period t to t+1 (TGR). 7 year. Based on the theoretical model analysis, this paper selects the capital stock (Capital), the fossil energy 16 consumption of energy-intensive industries (FE), and industrial employees (Labour) as the input variables; the 17 industrial value-added (Yield) as the desirable output; and the CO2 emissions (E-CO2) of energy-intensive 18

Indicators and data
industries as the undesired output. The industrial value-added and energy consumption data of the six energy-19 intensive industries come from the "China Statistical Yearbook", the "China Industrial Economic Statistical 20 Yearbook", the "China Energy Statistical Yearbook" and the statistical yearbooks of the provinces over the years. 21 The data characteristics of each input-output variable are shown in Table 1.

22
Capital stock. Drawing on the estimation method of the capital stocks in various provinces of Shan (2008), the 23 capital stock is adjusted using constant 2005 prices with the economic depreciation rate of 9.6%. The calculation 24 formula is as follows: 25 refers to the fixed asset investment in the current year; and , refers to the depreciation rate of fixed assets. 28 Energy consumption. The fossil energy consumption of the energy-intensive industries of various regions is 29 taken as the input element, including raw coal, coke, natural gas, crude oil, gasoline, kerosene, fuel oil, liquefied 30 petroleum gas, diesel and other fossil energy sources. Since the data of Shanghai, Jiangsu, Zhejiang, and Sichuan 31 provinces cannot be directly obtained, we treat these provinces using data from those with similar energy 32 consumption structures by following Lin and Tan (2016): the energy consumption structures of Shanghai and 1 Sichuan refer to Beijing and the energy consumption structures of Chongqing, Jiangsu and Zhejiang refer to 2 Guangdong. 3 labour force. Considering the availability of data and the methods of most studies, this paper uses the number of 4 industrial employees to measure the labour input. 5 Industrial added-value. The industrial value-added of the energy-intensive industries during the research period 6 is used as the desirable output, and the figures are adjusted using 2005 prices. 7 CO2 emissions. Based on the types of fossil energy consumption in energy-intensive industries, the average low 8 calorific values and carbon oxidation rates of different fossil energy sources using the "2006 IPCC Guidelines 9 for national greenhouse gas emission Inventories" are adopted to calculate CO2 emissions (IPCC 2006). 10  Mid-western region, resulting in a large gap between the meta-frontier and group-frontier of the energy-intensive 32

Results and Discussion
industries. Especially for Hunan, Hubei, Sichuan, Chongqing, Guangxi, Shanxi and other provinces, most of 33 them belong to the Yangtze River Economic Belt and have regional advantages and development potential. 34 Driven by industrial transformation, the economic growth rates of these provinces are higher than the average 1 level; furthermore, loose policy also provides wide energy and carbon emission spaces for them. 2 After 2000, due to international industrial transfer, resource and environmental constraints, and increased labour 3 costs, the spatial distribution of the energy-intensive industries in China has experienced the characteristic of 4 "agglomeration to diffusion to aggregation" (Liu et al. 2019a). By transferring industry from eastern coastal 5 regions to the central and western regions, industrial transfer brings economic development and production 6 technology, but it also brings a lot of carbon flow, stimulating the further development and growth of energy-7 intensive industries in the central and western regions. In the "Eleventh Five-Year Plan" and "Twelfth Five-Year 8 Plan", the government has accelerated the energy conservation and emission reduction processes, and has 9 controlled energy consumption and carbon emissions from experiencing excessively rapid growth by setting 10 energy consumption intensity targets and carbon emission intensity targets, respectively. During the period of

Meta-Frontier Malmquist-Luenberger Productivity Index (MF-MLPL)
3 Based on the meta-frontier, figure 5 presents the changes in the carbon emission performance in energy-intensive 4 industries. The growth rate of the carbon emission efficiency of energy-intensive industries in different economic 5 regions shows a trend of first declining and then slowly recovering. During the "Eleventh Five-Year Plan" period, 6 the Chinese government attached great importance to energy conservation and emission reduction and regarded 7 energy conservation and emission reduction as an important tool for adjusting the economic structure, thus 8 changing the development mode, responding to climate change, and promoting scientific development. They put 9 forward the constraint targets of reducing energy consumption per GDP by 20% and reducing the total pollutant 10 emissions by 10%. Therefore, the growth rate of the carbon emission efficiency of energy-intensive industries Year Northeast East Medium West Average significantly from 1.0074 to 1.016 while the carbon emission efficiency of the energy-intensive industries in the 1 northeast region dropped to 0.9819. The BPC in different economic regions has shown a downward trend, and 2 the technological innovation capacity in the eastern region is relatively high. The BPC in the western region has 3 decreased from 1.0491 to 0.9986 since 2010. The possible reason is that the continuous increase in R&D 4 innovation investment in the eastern region has been effective. The energy-intensive industries in the western 5 region are mostly heavy industries with low energy efficiency, resulting in an insufficient technological 6 innovation capability.

20
(1) Carbon emission reduction potential in energy-intensive industries (relative value) 21 As shown in Figure 6, the carbon reduction potential of energy-intensive industries generally shows a trend of 22 decreasing and then rising, with an average score of 29.74%. The carbon emission reduction potential of energy-23 intensive industries in the eastern region keeps decreasing, falling from 34.05% to 15.31%. After 2000, energy 24 conservation and emission reduction have been constraints in the "Eleventh Five-Year Plan", and the government 25 has proposed specific requirements for controlling greenhouse gas emissions. Figure 5 shows that the carbon reduction potential of energy-intensive industries in the western region was relatively high, with an average score 31 of 37.73%; the reduction potential of energy-intensive industries in the eastern region was relatively low, with 32 an average score of 27.16%; and the reduction potentials in the northeast and central regions are both close to 33 the national level. After 2011, energy conservation and emission reduction policies were adjusted accordingly 1 by emphasizing industrial optimization, upgrading energy-intensive industries and adjusting the energy structure; 2 establishing and improving the economic structure adjustment guarantee mechanism; and promoting the 3 accountability mechanism for energy conservation and emission reduction. After 2012, except for the eastern 4 region, the carbon emission reduction potentials of energy-intensive industries greatly increased, and the 5 reduction potential in the northeast region rose from 26.94% to 39.51%. The carbon emission reduction 6 potentials in the central and west regions increased by 27.92% and 26.17%, respectively. The results indicate 7 that the reconstruction plans of the old industrial bases in the Midwest and Revitalizing the Northeast Strategy 8 have been effectively implemented. With the urbanization and industrialization in the Midwest, more emphasis 9 has been placed on upgrading, transforming and adjusting the energy structure of energy-intensive industries. 10 From the perspective of provinces, the carbon emission reduction potentials of energy-intensive industries in the 11 eastern region are polarized, those in the central region is more even, and the reduction potential in the western 12 region is highly fluctuant. The carbon emission reduction potentials of energy-intensive industries in Tianjin and 13 Guangdong are relatively low, all within 10%; and the provinces such as Hainan, Guizhou, Gansu, and Ningxia 14 have high reduction potentials, approximately 40%. 15 Under the group-frontier technology in Figure 7, the trends of the carbon emission reduction potentials of energy-16 intensive industries in different economic regions are similar to that of the meta-frontier. Except for the eastern 17 region, the carbon emission reduction potentials of energy-intensive industries are lower than that of the meta-18 frontier. The average carbon emission reduction potential of energy-intensive industries nationwide is 23.94%.

19
The carbon emission reduction potential of energy-intensive industries in the central region is significantly lower Tons, 35.16%). Liaoning is the city with the highest potential value in the northeast region. Since the "Eleventh 12 Five-Year Plan", the growth rate of the energy-intensive industries in Liaoning has slowed down, and the 13 potential value of energy-intensive industries has increased year by year. This indicates that Liaoning Province 14 has achieved significant effectiveness in the compression of energy-intensive industries and the adjustment of 15 economic structure. 16 During the research period, the carbon emission reduction potential values of the energy-intensive industries in 17 the eastern region showed a tiered fluctuation trend, and the potential values in Hebei and Shandong are similar 18 and high as they are both provinces with relatively concentrated energy-intensive industries. Hebei is an 19 industrial province with steel, cement, and glass as the economic mainstays while Shandong is dominated by 20 petrochemicals and electricity. The total output values of the energy-intensive industries in Hebei and Shandong 21 account for approximately 40% and 50%, respectively, and the proportion of the energy consumption of the 22 energy-intensive industries with respect to the total industrial energy is up to 75%; therefore, the carbon reduction 23 trends of the two provinces are similar in the implementation of the energy saving and emission reduction policy. 24 Shanghai, Jiangsu, Zhejiang and Guangdong in the eastern region are located in the Yangtze River Economic 25 Belt with the strongest comprehensive strength and are pioneers of the ecological civilization construction of 26 China. Meanwhile, Shanghai, Jiangsu and Zhejiang rank as the top three in the green development of the Yangtze 27 River Economic Belt. With the gradual increase in the energy consumption, the carbon emission reduction 28 potential in the eastern region is decreasing yearly, and the actual potential value of each province fluctuates less. 29 The innovation-driven effect in Guangdong has been improved the fastest, and the potential value of the energy-30 intensive industries in Guangdong has changed significantly.

31
The  Conclusions and policy implications 1 The paper describes the CO2 emissions of the energy-intensive industries of China, builds a Super-SBM model 2 considering undesirable outputs, and calculates the carbon emission efficiency. Then, Meta-Frontier Malmquist-3 Luenberger Productivity Index (MF-MLPI) is constructed to dynamically analyse the growth rate changes of the 4 carbon emission efficiency and the regional differences among energy-intensive industries; furthermore, the 5 carbon emission reduction potential of the energy-intensive industries in various economic regions of China are 6 discussed. Conclusions are drawn as follows.

7
(1) From the perspective of the province, there is a big difference in the carbon emission Technology Gap Ratio 8 (TGR) of energy-intensive industries in different economic regions. The technology gap between the meta-9 frontier and the group-frontier of carbon emissions in Heilongjiang is large. The technology gap ratio of 10 each province in the eastern region is close to 1. The TGR of the meta-frontier and group-frontier of energy-11 intensive industries in central regions is around 0.88, and the technical gap between the meta-frontier and 12 group-frontier of energy-intensive industries in the western regions fluctuates slightly, among 0.83 to 0.95. is constantly narrowing, while the gap in the Northeast region has gradually expanded.

17
(2) The growth rate of carbon emission efficiency of energy-intensive industries in different economic regions 18 shows a trend of first declining and then slowly recovering. The growth rate of carbon emission efficiency 19 of energy-intensive industries fell sharply before 2010. and the growth rate in western regions is relatively 20 high. After 2014, the MF-MLPL has been gradually improved, especially for the eastern region, the growth 21 rate of carbon emission efficiency of energy-intensive industries has rebounded since 2012, increased by 22 7.15%. Compared with other economic regions, the efficiency of energy-intensive industries in the central 23 and northeastern regions increased rapidly before 2010. After 2010, with the advancement of energy saving 24 and emission reduction policies, the carbon emission efficiency in the eastern region increased significantly, 25 while that in the northeast region dropped to 0.9819.

26
(3) Carbon reduction potential of energy-intensive industries generally shows a trend of decreasing and then 27 rising, with an average score of 29.74%. The carbon emission reduction potential in the eastern region keeps 28 decreasing. From 2005 to 2012, the carbon emission reduction potential of energy-intensive industries in 29 the western region was relatively higher than that in the eastern region was relatively low. After 2012, except 30 for the eastern region, the carbon emission reduction potentials of energy-intensive industries in other 31 economic regions have greatly increased. From the perspective of provinces, the carbon emission reduction 32 potentials of energy-intensive industries in the eastern region are polarized, the central region is more even, 33 while the reduction potentials in the western region are highly fluctuant. The carbon emission reduction 34 potential in Tianjin and Guangdong is relatively low, all within 10%; provinces such as Hainan, Guizhou, 1 Gansu, and Ningxia have high carbon emission reduction potential, around 40%. 2 (4) During the research period, the carbon emission reduction potential values of energy-intensive industries in 3 the central regions differ in each province, among of which the potential values in Henan, Shanxi, and Anhui 4 are relatively high and rising slowly, while the potential values in Hunan, Hubei, and Jiangxi show a 5 downward trend. The carbon emission reduction potential values of energy-intensive industries in the 6 western region are relatively concentrated, and the potential values in Xinjiang, Inner Mongolia, and Shaanxi 7 provinces have changed significantly. 8 Based on the research conclusions, policy recommendations are drawn as follows: 9 (1) According to the industrial advantages of different regions, the government should rationally distribute 10 energy-intensive industries, promote industrial structure adjustment and optimize the energy structure. As 11 the agglomeration region of energy-intensive industries, the eastern region has experienced a gradually 12 weakened resource carrying capacity. It is necessary to further improve environmental regulations, 13 implement various environmental standards, and strengthen the application of energy-saving and 14 environmental protection technologies in energy-intensive industries. During the rapid industrialization and 15 urbanization process, the central and western regions have transferred some energy-intensive industries. 16 Hence, we should pay attention to the harmonious development of industry, regional resources and the 17 environment; strengthen the energy efficiency to avoid the "pollution paradise" effect in the central and 18 western regions. Furthermore, the growth of the value-added of the energy-intensive industries in the central 19 and western regions should be controlled to avoid overcapacity. 20 (2) The government should increase R&D investment and promote energy technology innovation in energy-21 intensive industries. It is necessary to strengthen low-carbon technological innovation and achieve 22 transformation in the eastern region, effectively support the development of high-tech industries and modern 23 service industries, and increase the value added of energy-intensive industries. The energy efficiency of the 24 energy-intensive industries in the central and western regions is relatively low, and the government should 25 actively eliminate outdated industries and promote the upgrading of the industrial structure. Through the 26 development of energy technology, we can help to improve the energy-consumption efficiency, thereby 27 promoting the transformation of energy-intensive industries from extensive to intensive.

28
(3) The government should formulate differentiated emission reduction plans according to the characteristics of 29 regions and industries. Energy-intensive industries in China are still currently dominated by coal-based 30 energy structures, which contain huge potential for energy structure optimization to promote carbon 31 emission reduction. The scope of resource tax collection for fossil energy such as coal should be 32 appropriately expanded, complete environmental protection tax system could be established, and the 33 competitiveness of the market prices of renewable energy should be enhanced in the meantime. The 34 government should give priory to promoting carbon peaks of key emission industries and regional, formulate 35 differentiated carbon emission reduction plans for the regions and industries with different reduction 36 potentials, and gradually complete scientific and efficient carbon emission reduction targets for energy-37 intensive industries. 38

39
Ethics approval and consent to participate：All studies did not involve human or animal ethics. 40 Consent for publication：This manuscript hasn't contained any individual person's data in any form.

41
Availability of data and materials ： The data and materials used in the study are available from the 42 corresponding author by request. 43