Primary energy consumption structure and the influencing factors in China: an income decomposition and post-economic crisis era perspective

China’s coal-based energy structure makes its carbon peak and neutrality goals very challenging. As a result, optimizing the energy structure has become an important means, and researching its influencing factors and trends has become the foundation and prerequisite for policy formulation related to energy structure optimization. Especially after the severe economic crisis, the economic structure has undergone profound changes, and the impact of related factors on the energy structure has also changed. This study adopts regression methods considering heteroskedasticity and cross-section correlation to study the panel data of 30 provinces in China and obtains the changes in the relationship between fossil and non-fossil fuel consumption in different regions and their influencing factors after two global financial crises. Research results show that China’s energy consumption tends to decouple from GDP. Income structure changes, especially trade changes, are important factors in influencing energy consumption and energy structure. The deviation between energy resource distribution and consumption distribution tends to increase. Regional heteroskedasticity is evident in the impact of urbanization development models, energy prices, and efficiency.


Introduction
With the development of a low-carbon economy, governments are paying more and more attention to the optimization of energy structure. Due to the characteristics of resource endowment, China's energy consumption has long been dominated by coal. Therefore, the optimization of energy consumption structure has attracted much attention today due to the increase in environmental pressure. The Chinese government has continuously introduced policies to encourage the development and utilization of new energy sources and to optimize energy structure and reduce pollutant emissions through technologies and policies. However, due to the characteristics of the development stages of urbanization and industrialization, such as the economic characteristics of investment-driven and administrative guidance, may offset the effects of these policies. On the other hand, there are significant regional differences in economic development in China, and there is a certain imbalance between the distribution of energy resources and energy consumption. Figures 1, 2, and 3 show the changes in China's primary energy structure, the primary energy consumption structure of 30 provinces in China, and the resource endowments in each region. The combination of the three figures shows that there are huge differences in energy consumption and resource endowment in different regions, and it is necessary to optimize relevant energy consumption policies. As Zhang and Lahr (2014) said, as a spatially large and heterogeneous country, it is crucial to identify the key driving forces of the growth in energy use for each region to design appropriate regional energy conservation policies.

Responsible Editor: Roula Inglesi-Lotz
What is more, economic crises have had a profound impact on the global economy. After the severe economic crisis, the economic structure has undergone profound changes, which also makes the impact of related factors on the energy structure likely to change. Some scholars have also paid attention to the relationship between the economic crisis and energy consumption. However, little research has been done in China on this issue.
Based on existing research and understanding of China's energy consumption and regional differences, this study will comprehensively consider macro-level factors, including income, income structure, urbanization, energy price, and efficiency, from the perspective of geographic location and economic development similarity, using regional panel data and regression models to study the relationship between these factors and the main primary energy consumption, the differences before and after the economic crisis, and then draw some policy implications.
The other parts of the paper are arranged as follows: The second part is a literature review. Based on the literature review, the variables, research methods, and data used in the empirical tests are introduced in the third part. The empirical research results and discussion are given in the fourth part, and the fifth part encapsulates the main points of the study and gives some policy implications.

Literature review
There are numerous empirical studies on energy consumption structure. A literature review for this study would examine the literature which focused more on income perspective and financial crisis.

A review of the literature on the structure of energy consumption and the factors that influence it
There is relatively little direct research on the structure of energy consumption in the academic community. There are also some studies involving energy consumption in different categories (Wang and Lin 2017;Wandji 2013;Bilgen 2014;Park and Hong 2013;Tugcu and Topcu 2018;Saboori and Sulaiman 2013). Many studies have focused on the impact of energy structure on energy efficiency and carbon emission efficiency (Chen and Chen 2016; Wang and Xie 2015;Liu et al. 2016;Zhang et al. 2016;Chen et al. 2018). The above-mentioned studies mostly focus on the relationship between energy structure and energy efficiency, carbon emissions, etc., and seldom directly study energy structure and its influencing factors. The research on China's energy consumption is mainly about primary energy consumption, terminal energy consumption, or a specific energy consumption pattern, rarely involving changes in energy structure. As Liao and Wei (2010) said, most previous studies directly used the aggregate energy data provided by China's National Bureau of Statistics or International Energy Agency, and this approach sometimes may not be scientific when precisely investigating the energy-economy interactions is needed. Research involving the structure of energy consumption, either using a gray model (Li et al. 2006;Tsai 2016;Yuan et al. 2010) or studying the energy consumption structure of certain production sectors Zhou et al. 2009). Some researchers also studied energy consumption or energy intensity in different categories (Chang et al. 2003;Fan et al. 2007;Yuan et al. 2008;Fan and Xia 2012;Fisher-Vanden et al. 2004;Feng et al. 2009). However, they have Coal Crude Oil Natural Gas Primary Electricity and Other Energy rarely considered the issue of energy consumption structure, especially the structure of fossil energy consumption.

A literature review on energy consumption structure and the financial crisis
In recent decades, two major economic crises, including the 1998 economic crisis and the 2008 economic crisis, have had a profound impact on the global economy. Some scholars have also paid attention to the relationship between the economic crisis and energy consumption. By using a multi-scale integrated analysis of societal and ecosystem metabolism, Andreoni (2020) investigated energy requirements, economic trends, and population for the years 2008 and 2015 and discussed how the deceleration of economic growth that followed the financial crisis of 2008 influenced the energy efficiency, allocation, and use of 18 European countries. Andriosopoulos et al. (2017) investigated whether and to what extent events in financially troubled EU markets (Greece, Ireland, and Portugal) affected energy prices during the EU financial crisis. Hofman and Huisman (2012) presented a research project that examined whether the preferences of investors in Europe and the USA changed due to the financial crisis and found that the popularity of 11 out of 12 policies decreased. Selcuk and Durusoy (2019) pointed out that the transportation sector is an energy-intensive sector that needs more attention from policymakers due to its contribution to the increase in energy intensity after the crisis years as well as its increasing contribution to overall energy intensity. Some scholars  (2010) have discussed the influence of the Global Financial Crisis on energy consumption and economic growth.
As Timma et al. (2016) said, there is a need to evaluate the effect of the economic recession in other regions and countries. Moreover, decomposition analysis can be further expanded to include the aggregate of energy carried in order to account for the shift toward renewable energy at a sectorial level. Thus, the structure of the income is also important. In China, like the energy consumption structure, important changes are also taking place in the structure of various parts of the GDP in the macroeconomic body, especially after the 2008 global financial crisis. The changes in the overall environment have caused major changes in China's foreign trade environment, and exports have suffered more with strong resistance, while imports have grown rapidly, as shown in Fig. 4. In addition, in order to promote economic development, various regions have increased their investment. From the perspective of the GDP structure of expenditure method, as shown in Fig. 5, the proportion of total capital formation in GDP has increased. However, few scholars have so far discussed the relationship between the changes in the proportion of expenditures of various sectors in the GDP using the expenditure method and energy consumption, especially the energy structure.
In this study, the primary energy structure in China is considered, and the relationship between coal, oil, natural gas, and GDP will be studied, respectively. At the same time, this paper also added the income structure as an influencing factor from the perspective of the expenditure method of GDP accounting to discuss the relationship between it and energy consumption and supplement the existing research.

Variables and data
From the method point of view, the panel data model is widely used in the study of energy consumption in different countries or regions (for example, Narayan and Popp 2012;Li and Lin 2015;Li et al. 2011;Wang and Lin 2019). Based on the literature, from the perspective of macroeconomics, the influencing factors of primary energy consumption generally include income, price, economic structure, population, and technical progress. Considering the representation and availability of data, the dependent variables and data selected in this paper are as follows: per capita total primary energy consumption (TOT), per capita coal consumption (COA), per capita oil consumption (OIL), per capita natural gas consumption (GAS), and per capita other fuel consumption (OTH), which includes primary electricity and other non-fossil fuels. The energy consumption data are from NBSC (2020).
Per capita GDP is also included. GDP represented by expenditure includes household final consumption expenditure, capital formation, general government final consumption expenditure, and exports and imports of goods and services. As the levels and shares of these variables in GDP reflect the region's economy and social structure    (2020). Note that the fossil resource endowment coefficient here is a simple indicator of the energy supply capacity of the fossils in the region and equals to fossil energy production/fossil energy consumption. Fossil fuels are made from decomposing plants and animals.
These fuels are found in the Earth's crust and contain carbon and hydrogen, which can be burned for energy. Coal, oil, and natural gas are examples of fossil fuels. Non-fossil fuels are alternative sources of energy that do not rely on burning up limited supplies of coal, oil, or natural gas. Examples of these fuels include nuclear energy, wind or water generated energy, and solar power characteristics, this paper uses these variables to analyze the influence of income structure on energy consumption for better comparisons and understanding. These variables include per capita real GDP (GDP), household final consumption expenditure share in GDP (SCON), gross capital formation share in GDP (SINV), general government final consumer expenditure share in GDP (SGOV), exports of goods and services in GDP (SEXP), and imports of goods and services in GDP (SIMP), which are calculated by the expenditure method and can reflect economic structure. And the data are from NBSC (2021b) (constant 2008 RMB). The urbanization rate (URR ) is used to capture sociodemographic structure, and the data is from NBSC (2021a). The fuel Retail Price Index is used as a price variable (EPR) (with 2008 as the base year), and the data is from NBSC (2021b). The fossil resource endowment coefficient, which equals fossil energy production/fossil energy consumption (REE) and is introduced as a simple indicator of the energy supply capacity of the fossils in the region, is also introduced as a simple indicator of the energy supply capacity of the fossils in the region. The data are from NBSC (2020). The ratio of real GDP to electricity consumption is used as a technology variable (EEF) or efficient variable (Chong et al. (2017)), and the data are from NBSC (2021b). The definition of the specific variables used in this study, the unit measurements, and the sources of the data are listed in Table 1. In order to avoid the heteroscedasticity problem, the figures are transformed into logarithmic form. The logarithmic form is expressed as follows: LnTOT, LnCOA, LnOIL, LnGAS, LnOTH, LnGDP, LnSCON, LnSINV, LnSGOE, LnSEXP, LnSIMP, LnURR , LnREE, LnEPR, and LnEEF. And the time scale of the data is from 1998 to 2017. Table 2 provides the statistical review of the variables.

Correlation study on energy consumption structure and influencing factors
First, the results of a national study would be presented and discussed. Then, heterogeneity analysis would be done at the provincial level, as the effects of influential factors on energy consumption structure could be heterogeneous based on location.
(1) Only the results of the Hausman test and the overidentifying test are reported in order to determine the selection of either a random or fixed effects model. This study would use the fixed effects model as the results are affirmed in Table 3. Furthermore, heteroscedasticity, serial correlation, and cross-sectional dependence tests are needed. In this paper, a modified Wald statistic for groupwise heteroskedasticity in the residuals is implemented, following Greene (2000), and the results are shown in Table 4. The test results for the hypothesis of serial dependence correlation are shown in Table 5, and this method has been discussed by Wooldridge (2002). The hypothesis of cross-sectional independence in panel data models with small T and large N can be implemented with two semi-parametric tests proposed by Friedman (1937) and Frees (1995Frees ( , 2004, as well as the parametric testing procedure proposed by Pesaran (2004). The results are shown in Table 6. It can be seen that there are different test results for different energy consumption equations, and the data of some equations may have the problems of heteroscedasticity, autocorrelation, and cross-sectional correlation. To solve these problems, regression with Driscoll-Kraay standard method (Driscoll and Kraay, 1998) has been adopted. In order to test whether there is a structural change in this time period, especially after the Asian financial crisis in 1998 and the global financial crisis in 2008 (Wen et al. 2012; Dungey and Zhumabekova 2001), this paper will introduce dummy variables and test the joint significance of all dummy variables and their cross-term coefficients with explanatory variables. That is to say, the dummy variable "d" is introduced in the basic Eqs.

Results of a national study
(1)-(5). After 2008, d = 1 ; conversely, d = 0 , and the interaction terms between the dummy variable "d" and each influencing variables are also introduced: dLnGDP = d * LnGDP ,  Table 7 shows that there was indeed a structural change after 2008. The coefficient estimates obtained by regressing the data from the two time periods, respectively, are shown in Table 8. The regression results of the data from 1998 to 2017 are also presented in Table 8.
According to the results of Table 8, from the national situation, the role of income (GDP) in energy consumption has shown a decline after 2008.
The increase in household final consumption expenditure in the income (SCON) is inhibiting total energy consumption, while there is a significant positive role in natural gas consumption, and the effect is stronger after 2008.
The increase in the proportion of gross capital formation in GDP (SINV) has a certain forward pulling effect on total energy consumption, but the role is smaller, and after 2008, the role is not significant. Before 2008, an increase in this proportion played a forward role in other energy consumption, but after 2008, its role became negative and the role of natural gas consumption changed to a significant forward direction.  (1) Equation (2) Equation (3) Equation (4) Equation (5 (1) Equation (2) Equation (3) Equation (4) Equation (5

Equation
Equation (1) Equation (2) Equation (3) Equation (4) Equation (5 The proportional change in general government final consumption expenditure (SGOV) and the change in total energy consumption have reversed relationships, and this effect is no longer significant since 2008. The reverse relationship between the proportional change in government consumption and coal consumption has become more remarkable after 2008, and after 2008, it has had a forward relationship with natural gas consumption.
The change in exports (SEXP) has shown a significant reverse relationship with natural gas consumption before 2008. Since 2008, it has had a significant reverse relationship with oil consumption. According to the data, China has been affected by the global economic situation in 2008, exports account for a reduction in GDP after 2008, and this change has promoted China's oil consumption. After 2008, the proportion of imports in China's GDP improved, and it also promoted the increase in China's oil consumption. That is, the comprehensive trade factors in recent years have also shown that China's internal economic cycle has improved. This process may increase the demand for oil consumption during internal transportation.
The urbanization process (URR ) has a pulling effect on the increase in total energy consumption, but the role is significantly smaller after 2008. For other energy consumption, that is, non-fossil energy consumption, the urbanization process before 2008 shows its inhibitory effect, while after 2008, the inhibitory effect is no longer significant.
On the whole, there is a clear negative relationship between fossil energy security capabilities (REE) and natural gas and oil consumption. To a certain extent, it indicates that there is a clear geographic deviation between the distribution of oil and gas resources and the distribution of oil and gas consumption in China. After 2008, the deviation of oil resource distribution and consumption distribution has become more significant, while the deviation of natural gas became smaller. Changes in fuel prices did not restrain total energy consumption.
From the perspective of the fuel price index (EPR), China's fuel prices showed an upward trend during the study period. However, fossil energy consumption was not restrained by prices, and the price elasticity was positive and has increased since 2008. But the changes in fuel prices have also played a significant role in promoting the development of non-fossil energy and the consumption of natural gas, and there has been no significant change in time. Based on these results, the increase in the fuel price index cannot contribute to energy saving, but it can optimize the energy structure to a certain extent. During the study period, the improvement in energy efficiency had a restraining effect on energy consumption, but according to the staged regression, this effect showed a downward trend.  1998-2017 1998-2008 2009-2017 1998-2017 1998-2008 2009-2017 1998-2017 1998-2008 2009-2017 1998-2017 1998-2008 2009-2017 1998-2017 1998-2008 2009-

Results of regional study
Considering that there are strong geographical characteristics in both consumption and resource distribution of energy in China, a sub-regional analysis is conducted in this paper. Based on the division of NBSC in 2011, China's mainland can be divided into four major economic regions: east, central, west, and northeast, as shown in Table 9.
Based on the results of Table 10, the Hausman test of total primary energy, coal, and other energy consumption in the northeast is not significant, and the Hausman test of all types of energy consumption in various regions indicates that a fixed effect model should be adopted. Moreover, heteroscedasticity, serial correlation, and cross-sectional dependence tests are also presented in Tables 11, 12, and 13. For regional panel data in this study, as the data section unit (N) is less than the data time span (T), the Breusch-Pagan statistic is better for testing cross-sectional independence in the residuals of a fixed effect regression model, following Greene (2000). Based on the results of Tables 11, 12, and 13, as most equations of most energy consumption in most regions have heteroscedasticity, autocorrelation, and inter-group correlation problems, regression with the Driscoll-Kraay standard method is adopted. Results of Table 14 presented the significance test of coefficients of a dummy variable and its interaction terms with explaining variables and indicated that there was a structural change in the year of 2008. The coefficient estimates obtained by regressing the data from 1998 to 2017 and the two time periods of 1998-2008 According to the results of the regression of the eastern region, income increase is still the driving factor of per capita total energy consumption and coal consumption, while it is not yet a driving factor for non-fossil energy consumption, or even a restraining factor.
The increase in the proportion of household consumption in GDP has not become an opportunity for the development Table 9 Chinese economic regions division The data sources for this article include provinces in mainland China, and as there is no energy consumption data for Tibet, this region is not included in the analysis of this paper.

East
Beijing  of non-fossil energy and has even become a significant restraining factor after 2008, but it is still a driving factor that promotes the increase in total energy consumption, coal and oil consumption. The increase in the proportion of fixed asset investment significantly increases coal consumption and natural gas consumption in the region. The increase in the proportion of government expenditure has shown a certain inhibitory effect on total energy consumption, but has also shown a significant promotion effect on natural gas consumption after 2008. The trade changes in the eastern region are more consistent with the national situation; that is, the proportion of exports in GDP has declined since 2008, while imports have increased, and changes in exports and total energy consumption have an inverse relationship; that is, as the proportion of exports in the eastern region has fallen, per capita total energy consumption has increased, non-fossil energy consumption and natural gas consumption has increased, while coal and oil consumption showed a decrease change. An increase in the proportion of imports can play a role in restraining per capita total energy consumption. Urbanization is also one of the important factors driving total energy consumption in the eastern region, but its effect is also showing a downward trend. The development of eastern urbanization has even suppressed oil consumption, which may be related to the realization of the relatively higher public transportation network. Most provinces in the eastern region have relatively scarce fossil resources, and their own fossil energy consumption guarantee capacity is declining, which is beneficial to the development of non-fossil energy to a certain extent, but has not become a constraint to restrain its fossil energy consumption. The increase in the fuel price index has not become a factor that restrains total energy consumption. Although the consumption of oil and coal still rises with the increase in the fuel index, its promotion of non-fossil energy consumption and natural gas consumption is stronger, and this effect has become more effective after 2008, which shows that although the fuel price index increase cannot play a role in energy saving, it can promote the optimization of the energy structure of the region. In line with national estimates, the effect of energy efficiency improvement on energy conservation is diminishing. The increase in income levels (GDP) in the central region does not seem to be the main driving factor for energy consumption; during the study period, the proportion of household final consumption (SCON) in the region has risen and fallen. According to coefficient estimates, the increase in the proportion of household final consumption will reduce the total per capita primary energy consumption, but it will increase coal consumption. The increase in the proportion of fixed asset investment in GDP (SINV) will significantly increase per capita primary energy consumption and natural gas consumption in the region. The increase in the proportion of government consumption (SGOV) has a positive effect on non-fossil energy consumption, but the effect is no longer significant after 2008, and the promotion effect on natural gas consumption is still relatively significant.
Different from the development of trade in the nation and the eastern region, the proportion of exports (SEXP) and imports (SIMP) in this region increased during the study period. According to the coefficient estimates, the increase in exports promoted the development of non-fossil energy consumption and oil consumption, and since 2008, natural gas consumption has significantly increased. The increase in the proportion of imports has a certain inhibitory effect on primary energy consumption and oil consumption, but the effect is small, and since 2008, natural gas consumption has also significantly increased. Overall, the increase in trade level has positive significance for energy conservation and emission reduction in the region.
The promotion of energy consumption by urbanization (URR ) in this region is mainly reflected in natural gas consumption. As the central region's energy resource endowment (REE) of is also relatively lacking, it has not inhibited its primary energy consumption, and as its fossil energy security capacity has decreased, primary energy consumption has greatly increased after 2008. Compared with the whole country and the eastern region, the improvement of energy efficiency in this region still has a greater potential for energy conservation.
In terms of imports and exports, the proportion of imports (SIMP) and exports (SEXP) of western provinces has risen and fallen. The increase in exports has a very positive effect on increasing non-fossil energy consumption and reducing fossil energy consumption. The increase in the proportion of imports has a boosting effect on natural gas consumption.
The development of urbanization (URR ) does not seem to be beneficial to the consumption of non-fossil energy in the region, but the impact of this unfavorable factor is weakening. In addition, the development of urbanization has a significant role in promoting oil and natural gas consumption. The western region has strong fossil fuel security capabilities and is the main source of fossil fuel consumption in the country. However, the increase in resource security capabilities in the region has a clear reverse relationship with oil and natural gas consumption, indicating that more oil and gas have been shipped to the central and eastern regions. Rising fuel prices (EPR) have not become a factor in restraining energy consumption in the region, and even have a significant promotion effect on coal consumption, and have not promoted the development of non-fossil energy. The energy-saving effect of energy efficiency improvement in the region has also declined as a whole.
Increasing income (GDP) in the northeast is still one of the main driving factors for its energy consumption, but this effect is decreasing. An increase in income has a restraining effect on oil consumption, but this effect is also decreasing. The increase in the proportion of household consumption (SCON) will increase primary energy consumption and curb non-fossil energy consumption. The increase in the proportion of fixed asset investment (SINV) in the region was conducive to the development of non-fossil energy before 2008, but after 2008, it has significantly promoted the consumption of coal. The increase in the proportion of government expenditure (SGOV) in the region is conducive to the development of non-fossil energy and has a certain restraining effect on natural gas consumption.
Urbanization (URR ) in this area is the main driving factor of per capita primary energy consumption, and its effects seem to have not undergone significant structural changes. Historically, the region's fossil energy resource security capability was strong, and it has shown a downward trend in recent years. The decline in resource security capacity before 2008 has an inverse relationship with its per capita primary energy consumption. After 2008, the decline in resource security capacity had a certain inhibitory effect on its per capita coal consumption. The increase in fuel prices (EPR) has not suppressed the consumption of fossil fuels in the region, nor has it promoted the consumption of non-fossil fuels. The improvement in energy efficiency in this region is relatively small compared with other regions, and there is no obvious structural change as a whole.

Discussions
Some important reasons and implications emerge from the results presented.
First, from the perspective of income level, the results show that the role of income (GDP) in per capita total primary energy consumption has shown a decline after 2008, not only in China as a whole, but also in the middle, west, and northeast regions, showing the decoupling between energy consumption and GDP. However, income (GDP) growth continues to be a driving factor in per capita total energy consumption and coal consumption in China's east region. In fact, the above results are consistent with the literature, which finds that most developed countries have achieved relative decoupling between energy consumption and GDP to some extent and are now moving toward absolute decoupling, while most developing countries have not decoupled (de Freitas and Kaneko 2011;Roinioti and Koroneos 2017;Wang and Su 2020;Wu et al. 2018).
Second, from the perspective of income structure, overall, the effect of the proportion of gross capital formation in GDP (SINV) on total primary energy consumption has weakened, both at the national and provincial levels. Since 2008, China has undergone a structural change in aggregate investment from machinery and equipment-dominated nonresidential structures (Bai 2016). The capital formation from the investment of machinery and equipment would have a long-lasting energy consumption effect, while for the latter, there is much less direct energy consumption-driven.
Third, the impact of SCON, SINV, and SGOV on natural gas consumption has turned positive and significant after 2008 in China as a whole. As a result of the construction and operation of China's west-to-east gas pipeline, the country has boosted its natural gas supply across the country and nearly 700 billion cubic meters of natural gas have been transported to eastern China since it was officially put into operation in 2004. According to China's statistics, household use of natural gas nearly tripled in 2010 compared to 2005. Also, the natural gas used in power generation increased from 2.5 billion cubic meters in 2005 to 19.0 billion cubic meters in 2010, almost 8 times.
Overall, China's trade situation is deeply affected by the economic crisis. The comprehensive trade factors in recent years have also shown that China's internal economic cycle is improved. This process may increase the demand for oil consumption during internal transportation. Whether there are technological spillover effects (Dogan and Seker 2016) from other countries still need to be further explored. The urbanization process has a pulling effect on the increase in total energy consumption, but the role is significantly smaller since 2008. For non-fossil energy consumption, the urbanization process before 2008 shows its inhibitory effect, while since 2008, the inhibitory effect is no longer significant. On the whole, there is a clear negative relationship between fossil energy security capabilities and natural gas and oil consumption. To a certain extent, it indicates that there is a clear geographic deviation between the distribution of oil and gas resources and the distribution of consumption in China. After 2008, the deviation of oil resource distribution and consumption distribution has become more significant, while the deviation of natural gas became smaller. Changes in fuel prices did not restrain total energy consumption. From the perspective of the fuel price index, China's fuel prices showed an upward trend during the study period. However, fossil energy consumption was not restrained by prices, and the price elasticity was positive and has increased since 2008. But the changes in fuel prices have also played a significant role in promoting the development of non-fossil energy and the consumption of natural gas, and there has been no significant change in time. Based on these results, the increase in the fuel price index cannot contribute to energy saving, but it can optimize the energy structure to a certain extent. During the study period, the improvement of energy efficiency had a restraining effect on energy consumption, but according to the staged regression, this effect showed a downward trend.

Conclusions
Whether it is the study of national panel data or the research of panel data by region, the results show that the decomposition of per capita GDP from the perspective of the expenditure method has different effects on energy consumption. At the same time, the energy consumption of different type responses differently to the influencing factors, and the impact of the global financial crisis on the economic situation also makes it have a structural transformation effect.
Firstly, for the whole country, the role of income in energy consumption has shown a decline in recent years, which means that energy consumption is tending to decouple from GDP. The pulling effect on total energy consumption from urbanization has also become smaller after 2008. For non-fossil energy consumption, the urbanization process before 2008 shows its inhibitory effect, while after 2008, the inhibitory effect is no longer significant.
Secondly, regarding the results of the regression of different regions and periods, it is found that income increase is still the driving factor of per capita total energy consumption and coal consumption in the east and northeast, although the effect in the northeast is decreasing. However, it has not been the driving factor in central and western regions. Instead, in the western region, it started to restrain energy consumption.
Thirdly, from the energy commodity perspective, the increase in the proportion of gross capital formation in GDP significantly increases coal and natural gas consumption in the east and significantly increases per capita primary energy consumption and natural gas consumption in the central region. Since 2008, it has significantly promoted the consumption of coal in the northeast, in the western, and in the northeast regions. The increase in the proportion of gross capital formation was a driving factor for the development of non-fossil energy before 2008.
Finally, with regards to the impact of price, in the eastern region, although the fuel price index increase cannot play a role in energy saving, it can promote the optimization of the energy structure of the region. However, in the central, western, and northeast regions, rising fuel prices have not suppressed the consumption of fossil fuels in the region, nor have they promoted the consumption of non-fossil fuels.

Policy implications
The above results have the following policy implications.
Since fixed asset investment and government spending are the main driving forces for economic development, in the process of investment and government budgeting, attention should be paid to the impact of fossil fuels, and guidance should be given on how to promote the development of non-fossil fuels and energy-saving technologies. On the whole, an increase in the proportion of household consumer consumption in GDP is conducive to energy conservation and emission reduction. Although there are certain differences in the trade development models of various regions in China, the overall development is conducive to energy conservation and optimization of the energy structure. Regarding the trade model of the northeast, it is necessary to further study whether it follows the comparative advantage in the use of energy resources. China is still in the rapid development stage of urbanization, and its own development may be a driving force for the consumption of fossil fuels, which is not conducive to the development of non-fossil fuels. Therefore, it is of great practical value to explore the urbanization process of clean development.
For the whole country and many regions, the overall price increase and fossil fuel resource constraints seem to have little effect on fossil energy consumption, but they will also promote non-fossil energy consumption.

Limitations
We acknowledge the limitations of this study, particularly in terms of the availability and quality of sufficient data at a disaggregated level. At present, there is still a lack of micro-basic research, and the interpretation of its impact and in-depth policy recommendations are still subject to regional differences. Still, there is a lack of in-depth guidance on specific sectors.
Regarding to the limitations on research content, structural changes of the influential factors of energy consumption have not been explored in depth. This paper uses the expenditure method to explore the impact of the residential sector, private investment sector, government sector, and import and export sectors on the energy consumption structure. Due to the limitations of research methods and data, this paper does not explore the deep-seated structural changes in the influence of these factors before and after the economic crisis, and the lack of projections for the longer-term effects beyond the economic crisis. In particular, there is a lack of discussion on the impact of economic activities in other countries around the world on China's energy consumption before and after the economic crisis. Of course, these are also directions for further research and in-depth research in the future.
Author contribution All authors contributed to the study's conception and design. Jianghua Liu and Ting Wang are mainly responsible for material preparation, data collection, and analysis. The first draft of the manuscript was written by Jianghua Liu and mainly revised by Yongqiang Xu, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding This work was supported by the National Natural Science Foundation of China (Grant numbers 72173046, 72174113, and 71603086).

Data availability
The datasets are available from the corresponding author on reasonable request.

Ethics approval No ethics-related information.
Consent to participate I am free to contact any of the people involved in the research to seek further clarification and information.
Consent for publication This manuscript has not been published or presented elsewhere in part or in entirety and is not under consideration by another journal.

Competing interests
The authors declare no competing interests.