Economic Structure Transformation and Low-Carbon Development in Energy-Rich Cities: The Case of the Contiguous Area of Shanxi and Shaanxi Provinces, and Inner Mongolia Autonomous Region of China

: Energy-rich cities tend to rely on resource-based industries for economic growth, which leads to a great challenge for its low-carbon and sustainable economic development. The contiguous area of Shanxi and Shaanxi Provinces, and the Inner Mongolia Autonomous Region (SSIM) is one of the most important national energy bases in China. Its development pattern, dominated by the coal industry, has led to increasingly prominent structural problems along with di ﬃ cult low-carbon transition. Taking energy-rich cities in the contiguous area of SSIM as examples, this study analyzes the main drivers of CO 2 emissions and explores the role of economic structure transformation in carbon emission reduction during 2002–2012 based on structural decomposition analysis (SDA). The results show that CO 2 emissions increase signiﬁcantly with the coal industry expansion in energy-rich cities. Economic growth and structure are the main drivers of CO 2 emission increments. An energy structure dominated by coal and improper product allocation structure can also cause CO 2 emission increases. Energy consumption intensity is the main factor curbing CO 2 emission growth in energy-rich cities. The decline of agriculture and services contributes to carbon emission reduction, while the expansion of mining and primary energy processing industries has far greater e ﬀ ects on CO 2 emission growth. Finally, we propose that energy-rich cities must make more e ﬀ orts to transform energy-driven economic growth patterns, cultivate new pillar industries by developing high-end manufacturing, improve energy e ﬃ ciency through more investment in key technologies and the market-oriented reform of energy pricing and develop natural gas and renewable energy to accelerate low-carbon transition. data


Introduction
With the development of industrialization and the expansion of cities, climate change with global warming as the main characteristic has attracted worldwide attention. Increasing anthropogenic greenhouse gas emissions are the main cause of global warming. According to the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report, CO 2 emissions account for more than 70% of anthropogenic greenhouse gas emissions. Fossil fuel combustion is the main source of CO 2 emissions. With the depletion of fossil fuels and the threat of climate change, developing the low-carbon economy has become a global trend. Cities are the key areas in which to implement carbon emission popular alternative method. Index decomposition analysis (IDA) [30][31][32][33][34][35] and structural decomposition analysis (SDA) [36][37][38][39][40][41] are mainly employed. IDA can only measure the direct production effect, while SDA can capture the direct and indirect production effects by taking inter-industrial interactions into consideration [6].
Taking energy-rich cities in the contiguous area of SSIM as examples, this study analyzes the main drivers of CO 2 emissions and the role of economic structure transformation during 2002-2012 based on the SDA method. Different from previous studies, this study makes new contributions in three aspects. First, there are few studies on carbon emission reduction in energy-rich cities, and this study fills this gap by focusing on Xinzhou, Lvliang, Ordos and Yulin in the contiguous area of SSIM. Second, in light of the limited availability of input-output tables for cities in China, which is significant for structural decomposition analysis, this study compiles CO 2 emission input-output tables for the contiguous area of SSIM. Third, although CO 2 emission inventories for some cities in China have been provided in several studies [1,3,42,43] and China High Resolution Emission Database (CHRED) has been constructed [44], CO 2 emission data at the city level are of lower quality and availability compared with the national and provincial data. Moreover, there is no CO 2 emission data covering detailed industries for the contiguous area of SSIM. This study estimates CO 2 emissions by 12 energy types and 17 industries for the contiguous area of SSIM. This study helps to better understand the drivers of CO 2 emissions in energy-rich cities and provide references for policymaking for low-carbon transition.
The remainder of this paper is organized as follows: the methodology and data are presented in Section 2, the empirical results are discussed in Section 3 and the main conclusions and policy implications are provided in Section 4.

CO 2 Emission Accounts
This study accounts for the CO 2 emissions of Xinzhou, Lvliang, Ordos and Yulin in the contiguous area of SSIM based on the IPCC emission factor approach. The formula is as follows: where TC is CO 2 emissions from energy consumption; E m,i is the consumption of energy type m by industry i; δ m is the conversion factor of energy type m from physical unit to coal equivalent; CV is the net calorific value; CEF m represents CO 2 emission factor of energy type m; COF m represents oxidization ratio of energy type m.  [1,3,42,45], this study estimates energy consumption of agriculture, construction and tertiary industries in Xinzhou, Lvliang, Ordos and Yulin by scaling down the corresponding provincial energy balance tables. It is assumed that these cities have the same energy intensity of agriculture, construction and tertiary industry as their provinces. Industrial energy consumption in these cities mainly comes from energy consumption and the transformation tables of industrial enterprises. Zhang (2010) adopted SDA from primary supply-side based on the Ghosh input-output model [14,38,46]. This study decomposes CO 2 emissions in Xinzhou, Lvliang, Ordos and Yulin from the supply-side, and the SDA combined with the Ghosh input-output model is used. CO 2 emission input-output tables for Xinzhou, Lvliang, Ordos and Yulin are compiled in this study. There are mainly two steps: compiling the city-level input-output table and placing CO 2 emissions by industries at the bottom of city-level input-output table. Input-output surveys are generally time-and labor-consuming. Therefore, non-survey methods have attracted widespread attention [47][48][49][50]. Following Miller and Blair (2009), , Lenzen et al. (2014) and Flegg et al. (2016) [49,[51][52][53], this study uses the augmentation of Flegg's location quotient method (AFLQ) to compile input-output tables for Xinzhou, Lvliang, Ordos and Yulin (see Appendix A for more details).

Structural Decomposition Analysis
Based on these cities' input-output tables, the Ghosh input-output model is constructed as follows: where e is the 1 × n dentity matrix; Z is the n × n intermediate input matrix; V and X represent the primary input matrix and the total input matrix, respectively. The direct output coefficient matrix A is defined by Formula (3). Then, the total input matrix can also be calculated by Formula (4).
where (I − A ) −1 is the Ghosh inverse matrix. Denote the Ghosh inverse matrix as G. Its element g ij represents the total output of industry j enabled by the unitary input of industry i. Then, CO 2 emissions of Xinzhou, Lvliang, Ordos and Yulin can be estimated as follows: where f is the CO 2 emission intensity matrix and represents CO 2 emissions per unit of output. Its element f i can be decomposed into: where TC m,i represents CO 2 emissions from energy m consumption in industry i; TC i is the total CO 2 emissions in industry i; X i is the total output of industry i; E m,i represents energy m consumption in industry i; E i is the total energy consumption in industry i. Therefore, Formula (5) can also be expressed as: where POP is the population; PGDP is the per capita GDP; IS is the 1 × n economic structure matrix; EF is the n × n energy consumption intensity matrix, and represents energy consumption per unit of output; ES is the n × m energy consumption structure matrix; CE represents CO 2 emissions per unit of energy consumption. Changes in CO 2 emissions from year t 0 to year t 1 can be calculated as follows: C∆POP, C∆PGDP, C∆IS, C∆G, C∆EF, C∆ES and C∆CE represent the contributions of population, economic growth, economic structure, production output structure, energy consumption intensity, energy consumption structure and CO 2 emissions per unit of energy consumption to CO 2 emission changes, respectively.
Superscripts one and zero represent years t 1 and t 0 , respectively. CO 2 emissions per unit of energy consumption are constant during the study period. Therefore, the value of C∆CE is zero. CO 2 emission changes can be attributed to scale effect (C∆PGDP and C∆POP), structural effect (C∆IS, C∆G and C∆ES) and intensity effect (C∆EF).  [5,17,36,54], this study uses the pole decomposition method to estimate the contributions of factors to CO 2 emissions. The formula is as follows:

Data Sources
Input Energy consumption data are derived from the above-mentioned statistical yearbooks and China Energy Statistical Yearbooks (2000)(2001)(2002)2013). This study accounts for CO 2 emissions by 12 energy types and 17 industries for Xinzhou, Lvliang, Ordos and Yulin. The energy types, industrial classification and their codes are presented in Appendix B Tables A1 and A2. The conversion factors of energy consumption from physical units to the coal equivalent are from China Energy Statistical Yearbooks (2013). CO 2 emission factors are from the IPCC Guidelines for National Greenhouse Gas Inventories (2006). The CO 2 emission factor of heat is estimated by dividing CO 2 emissions from energy consumption in the heat production process by the heat production amount. The CO 2 emission factor of electricity is obtained from the Baseline Emission Factors for Regional Power Grids in China. The oxidization ratio of fuel combustion is derived from the Provincial Greenhouse Gas Inventory Compilation Guidelines issued by the National Development and Reform Commission of the People's Republic of China.

Economic Structure and CO 2 Emissions Characteristics
The proportion of coal mining and washing in economic growth in the contiguous area of SSIM was much higher than the national level. The most prominent change in economic structure in the contiguous area was the expansion of coal mining and washing and the decline of agriculture. As shown in Figure 1, coal mining and washing (S2) in Xinzhou accounted for 19.21% of GDP in 2012, rising by 16% compared with 2002. Metal ore mining and processing (S3) also showed an obvious upward trend. The proportions of coal mining and washing (S2) in Lvliang and Ordos increased by 27% and 14%, respectively, reaching 45.42% and 32.04% in 2012. The proportion of coal mining and washing (S2) in Yulin was 38.47% in 2012, with an increase of 20% during the study period. Mining and processing of nonmetal ores and other ores (S4) also expanded. However, farming, forestry, animal husbandry and fishery (S1) shrank greatly. Manufacturing development in the contiguous area was still weak. The proportions of manufacturing (S5-S11) in GDP in Xinzhou, Lvliang, Ordos and Yulin were 10.24%, 22.76%, 9.56% and 9.55%, respectively, in 2012, lower than the national level (31.10%). Meanwhile, the proportion of other services (S17) in these cities was 10%-20% lower than the national level (29.45%).
Sustainability 2020, 12, x FOR PEER REVIEW 6 of 15 contiguous area was the expansion of coal mining and washing and the decline of agriculture. As shown in Figure 1, coal mining and washing (S2) in Xinzhou accounted for 19.21% of GDP in 2012, rising by 16% compared with 2002. Metal ore mining and processing (S3) also showed an obvious upward trend. The proportions of coal mining and washing (S2) in Lvliang and Ordos increased by 27% and 14%, respectively, reaching 45.42% and 32.04% in 2012. The proportion of coal mining and washing (S2) in Yulin was 38.47% in 2012, with an increase of 20% during the study period. Mining and processing of nonmetal ores and other ores (S4) also expanded. However, farming, forestry, animal husbandry and fishery (S1) shrank greatly. Manufacturing development in the contiguous area was still weak. The proportions of manufacturing (S5-S11) in GDP in Xinzhou, Lvliang, Ordos and Yulin were 10.24%, 22.76%, 9.56% and 9.55%, respectively, in 2012, lower than the national level (31.10%). Meanwhile, the proportion of other services (S17) in these cities was 10%-20% lower than the national level (29.45%).

The main Drivers Of CO2 Emission Increments
Economic growth is the main driver of CO2 emission increases in energy-rich cities in the contiguous area of SSIM. As shown in Figure 3

The Main Drivers Of CO 2 Emission Increments
Economic growth is the main driver of CO 2 emission increases in energy-rich cities in the contiguous area of SSIM. As shown in Figure [13,20,27]. During the study period, the coal market was in a golden development period in China. The average annual growth rate of the economy was more than 10%, even up to about 20% in the contiguous area of SSIM. This rapid economic growth inevitably leads to a large amount of CO 2 emissions.  [33,55,56], who found economic structure has an inhibition effect on CO2 emission growth. This difference is mainly attributed to the different characteristics of economic structure changes. During the study period, the coal industry in the contiguous area of SSIM expanded significantly and theeconomic structure became more carbon intensive.
Energy structure has a positive effect on CO2 emission growth in most cities in the contiguous area of SSIM. An energy consumption structure dominated by coal led to 9.88, 5.08 and 3.85 Mt CO2 emission increments in Ordos, Lvliang and Xinzhou, respectively, from 2002 to 2012. However, energy structure improvement had the opposite effect on CO2 emissions in Yulin. This is mainly due to the decline in coal consumption proportion and the significant increase in natural gas consumption in Yulin.
Energy consumption intensity is the main factor reducing CO2 emissions in the contiguous area of SSIM. Energy consumption intensity reduced CO2 emissions by 18.  [20,22,54]. Governments in Xinzhou, Lvliang, Ordos and Yulin have clearly set up constraint targets of energy conservation in different stages since 2005. As a result, energy efficiency has improved significantly.
The production output structure has different effects on CO2 emissions in energy-rich cities. The improvement of the production output structure reduced CO2 emissions in Xinzhou (-1.85) and Ordos (-3.11 Mt). However, the production output structure still positively affected CO2 emissions in Lvliang and Yulin. Therefore, the product supply chain should be taken into consideration in promoting carbon emission reduction. Population has slight promotion effects on CO2 emissions in Xinzhou (1.44), Lvliang (4.21), Ordos (7.76) and Yulin (5.09 Mt). The population in the contiguous area of SSIM grew slowly during the study period. In recent years, population loss has even occurred.  [33,55,56], who found economic structure has an inhibition effect on CO 2 emission growth. This difference is mainly attributed to the different characteristics of economic structure changes. During the study period, the coal industry in the contiguous area of SSIM expanded significantly and theeconomic structure became more carbon intensive.
Energy structure has a positive effect on CO 2 emission growth in most cities in the contiguous area of SSIM. An energy consumption structure dominated by coal led to 9.88, 5.08 and 3.85 Mt CO 2 emission increments in Ordos, Lvliang and Xinzhou, respectively, from 2002 to 2012. However, energy structure improvement had the opposite effect on CO 2 emissions in Yulin. This is mainly due to the decline in coal consumption proportion and the significant increase in natural gas consumption in Yulin.
Energy consumption intensity is the main factor reducing CO 2 emissions in the contiguous area of SSIM. Energy consumption intensity reduced CO 2 emissions by 18. The production output structure has different effects on CO 2 emissions in energy-rich cities. The improvement of the production output structure reduced CO 2 emissions in Xinzhou (−1.85) and Ordos (−3.11 Mt). However, the production output structure still positively affected CO 2 emissions in Lvliang and Yulin. Therefore, the product supply chain should be taken into consideration in promoting carbon emission reduction. Population has slight promotion effects on CO 2 emissions in Xinzhou (1.44), Lvliang (4.21), Ordos (7.76) and Yulin (5.09 Mt). The population in the contiguous area of SSIM grew slowly during the study period. In recent years, population loss has even occurred.
The scale effect contributes most to CO 2 emission growth, followed by the structural effect, while the intensity effect offsets the growth to some extent. The scale effect drove CO 2 emissions to increase by 11.54, 29.86, 67.40 and 41.53 Mt, respectively, in Xinzhou, Lvliang, Ordos and Yulin, and the structural effect also caused an increase of 6.06, 19.61, 41.79 and 19.28 Mt in these cities, respectively. However, the inhibition effect of energy intensity on CO 2 emissions was significantly smaller than the promotion effect of scale and structure.

Contribution of Different Industries to CO 2 Emissions
Economic structure plays an important role in the high carbonization of Xinzhou, Lvliang, Ordos and Yulin. This study further analyzes the contribution of changes in the proportion of different industries in economic growth to CO 2 emissions. As shown in Figure 4, coal mining and washing (S2) and metal ore mining and processing (S3) were the main industries driving CO 2 emission growth in Xinzhou. The remarkable expansion of S2 and S3 caused 2.10 and 1.71 Mt CO 2 emission increments, respectively, from 2002 to 2012. On the contrary, farming, forestry, animal husbandry and fishery (S1) and services (S16 and S17) contributed most to carbon emission reduction (−1.54 Mt). The scale effect contributes most to CO2 emission growth, followed by the structural effect, while the intensity effect offsets the growth to some extent. The scale effect drove CO2 emissions to increase by 11.54, 29.86, 67.40 and 41.53 Mt, respectively, in Xinzhou, Lvliang, Ordos and Yulin, and the structural effect also caused an increase of 6.06, 19.61, 41.79 and 19.28 Mt in these cities, respectively. However, the inhibition effect of energy intensity on CO2 emissions was significantly smaller than the promotion effect of scale and structure.

Contribution of Different Industries to CO2 Emissions
Economic structure plays an important role in the high carbonization of Xinzhou, Lvliang, Ordos and Yulin. This study further analyzes the contribution of changes in the proportion of different industries in economic growth to CO2 emissions. As shown in Figure 4, coal mining and washing (S2) and metal ore mining and processing (S3) were the main industries driving CO2 emission growth in Xinzhou. The remarkable expansion of S2 and S3 caused 2.10 and 1.71 Mt CO2 emission increments, respectively, from 2002 to 2012. On the contrary, farming, forestry, animal husbandry and fishery (S1) and services (S16 and S17) contributed most to carbon emission reduction (−1.54 Mt). Farming, forestry, animal husbandry and fishery (S1), the processing of petroleum, coking, processing of nuclear fuel (S6) and the tertiary industry contributed to reducing CO2 emissions by 9.96 Mt. Coal mining and washing (S2) and the processing of petroleum, coking, processing of nuclear fuel (S6) had the largest positive effects on CO2 emission increments in Ordos (32.96 and 8.36 Mt, respectively), which was offset by farming, forestry, animal husbandry and fishery (S1) and other manufacture (S11) to some extent (−5.51 Mt).
Mining and processing industries are the largest contributors to CO2 emission growth in Yulin. Coal mining and washing (S2) and the mining and processing of nonmetal ores and other ores (S4) had great positive effects on CO2 emissions from 2002 to 2012, which caused 24.59 and 8.09 Mt CO2 emission increments, respectively. Farming, forestry, animal husbandry and fishery (S1) and services showed certain carbon emission reduction effects (−10.41 Mt). The positive effects of economic structure on CO 2 emissions in Lvliang and Ordos are determined by coal mining and washing. Coal mining and washing (S2) led to an increase in CO 2 emissions by 18.30 Mt from 2002 to 2012 in Lvliang. Farming, forestry, animal husbandry and fishery (S1), the processing of petroleum, coking, processing of nuclear fuel (S6) and the tertiary industry contributed to reducing CO 2 emissions by 9.96 Mt. Coal mining and washing (S2) and the processing of petroleum, coking, processing of nuclear fuel (S6) had the largest positive effects on CO 2 emission increments in Ordos (32.96 and 8.36 Mt, respectively), which was offset by farming, forestry, animal husbandry and fishery (S1) and other manufacture (S11) to some extent (−5.51 Mt).

Conclusions and Policy Implications
Mining and processing industries are the largest contributors to CO 2 emission growth in Yulin. Coal mining and washing (S2) and the mining and processing of nonmetal ores and other ores (S4) had great positive effects on CO 2 emissions from 2002 to 2012, which caused 24.59 and 8.09 Mt CO 2 emission increments, respectively. Farming, forestry, animal husbandry and fishery (S1) and services showed certain carbon emission reduction effects (−10.41 Mt).

Conclusions and Policy Implications
Taking Xinzhou, Lvliang, Ordos and Yulin in the contiguous area of SSIM as examples, this study explores the main drivers of CO 2 emissions and the role of economic structure transformation in carbon emission reduction. The CO 2 emission input-output tables of these cities for 2002 and 2012 are compiled. Then, the SDA combined with the Ghosh input-output model is constructed to decompose CO 2 emission increments in these cities from 2002 to 2012. The main conclusions can be drawn as follows: (1) CO 2 emissions from energy consumption in energy-rich cities increase significantly with the expansion of the coal industry. Ordos had the largest increase in CO 2 emissions (90.72), followed by Yulin (50.27) and Lvliang (42.65 Mt). Xinzhou had the smallest CO 2 emission increments (12.90 Mt); (2) Economic growth and structure changes are the main factors driving CO 2 emission growth, and energy consumption intensity contributes most to carbon emission reduction in energy-rich cities. An energy structure dominated by coal resources has a small promotion effect on CO 2 emission increments. Production output structure could also cause CO 2 emission increases if more products flow to carbon-intensive industries; (3) Mining and primary energy processing industries, including coal and nonmetal ore mining, and petroleum, coking, and nuclear fuel processing, play the most significant role in CO 2 emission increments. Agriculture and services have the largest effects on carbon emission reduction. However, mining and primary energy processing industries are carbon-intensive, and their effects on CO 2 emissions are far greater than the emission reduction effects of agriculture and services.
Based on these findings, the following policy implications are proposed. First, energy-driven economic growth patterns must be transformed in energy-rich cities. With the depression of the coal market, the economic growth rate in the contiguous area of SSIM has declined sharply since 2012. Therefore, energy-rich cities must make efforts to transform the economy from energy-driven to innovation-driven in order to achieve sustainable development.
Second, energy-rich cities must accelerate economic structure transformation for carbon emission reduction. Government departments should strengthen the upgrading of the traditional coal industry. Small plants with high emissions should be closed or merged. Meanwhile, backward production capacity and its related facilities should be phased out gradually. Energy performance contracting can be carried out in high energy-consuming enterprises. More efforts should be made to cultivate new pillar industries by developing high-end manufacturing.
Third, investment in science and technology should be increased to improve energy efficiency. During the study period, energy efficiency in the contiguous area of SSIM was obviously improved by strengthening the application of energy-saving technologies. However, the energy consumption intensity of coal mining, the processing of petroleum, coking, processing of nuclear fuel and the production and supply of electricity was still high. Therefore, more funds should be invested in the research and development of key technologies for the efficient and clean use of coal. For example, the low-temperature power generation of coal gangue-fired circulating sulfurization bed technology can be developed. In addition, the market-oriented reform of energy pricing should be implemented. Increasing fossil fuel prices moderately will provide incentives for energy-intensive industries to improve energy efficiency.
Finally, governments should vigorously support the improvement of energy structure and product allocation structure in energy-rich cities. Fiscal policies, such as investment subsidies, financial discounts, tax preferences and price subsidies can be implemented to develop natural gas and renewable energy (wind power, photovoltaic power). Additionally, energy structure adjustment has been paid more attention by strictly controlling the total consumption of coal and promoting new energy power projects since 2012. Moreover, higher income taxes should be levied to prevent enterprises from selling too many products to carbon-intensive enterprises.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
The key to compiling the city-level input-output table is to estimate the intermediate input matrix. The calculation method in early studies is to use simple location quotient (SLQ) to regionalize the national input-output table [57,58]. The formula of SLQ is as follows: where X r i represents the output of industry i in region (city) r; X r is the total output in region (city) r; X n i represents the output of industry i in the country (province); X n is the total output in the country (province).
Based on SLQ, the cross-industry quotient (CIQ) is developed by taking the relative importance of supply industries and purchase industries into consideration. CIQ is defined as: where i and j represent supply industries and purchase industries, respectively. A large number of studies have found that both SLQ and CIQ underestimate the interregional economic interactions [49,59]. Flegg et al. (1995) modified the location quotients and proposed FLQ formula as follows [60]: The range of δ is defined as [0,1). With δ rising, the import in region (city) r will also increase. McCann and Dewhurst (1998) argued that regional specialization might lead to a regional intermediate input coefficient higher than the national coefficient [61]. Thus, Flegg and Webber (2000) adjusted the FLQ [50]. The augmentation of Flegg's location quotient (AFLQ) is defined as: Based on AFLQ, intermediate input coefficients in Xinzhou, Lvliang, Ordos and Yulin can be calculated as follows: a r ij = AFLQ r ij × a n ij (A7) where a r ij is the intermediate input coefficient in city r. a n ij represents the intermediate input coefficient in the province which the city r belongs to.
Intermediate input matrices in Xinzhou, Lvliang, Ordos and Yulin are calculated by multiplying intermediate input coefficients by the total output. According to statistical data of the total intermediate input in Xinzhou, Lvliang, Ordos and Yulin, this study chooses δ with the smallest error. Table A1. Industrial classification and codes.

S1
Farming, forestry, animal husbandry and fishery S2 Coal mining and washing S3 Metal ore mining and processing S4 Mining and processing of nonmetal ores and other ores S5 Manufacture of foods and tobacco S6 Processing of petroleum, coking, processing of nuclear fuel S7 Chemical industry S8 Manufacture of nonmetallic mineral products S9 Smelting and pressing of metals S10 General and special purpose machinery S11 Other manufacture S12 Production and supply of electricity and steam S13 Production and supply of gas and water S14 Construction S15 Transportation, storage, and post services S16 Wholesale, retail trade and hotel, restaurants S17 Others